The State of
AI Readiness
in Texas
Eighteen voices, eighty billion dollars, and the quiet sound of an industrial economy reorganizing itself around intelligence.
A reader's map
through the readiness landscape.
This Report Contains
- Executive Summary
- The Texas Backdrop: A State Reorganizing
- Survey Methodology & the Honest Sample
- The Silence Signal: Reading the 9,982
- Key Findings Across Four Readiness Dimensions
- Leadership Intent vs. Operating Discipline
- Data Readiness, Governance & Risk
- People, Adoption & the Value Question
- Financial Expectations & ROI Discipline
- Energy Sector: Houston's Quiet Reckoning
- Healthcare: The World's Largest Complex Faces AI
- SMB Texas: The Capacity Trap
- The Infrastructure Tsunami: Stargate, Apple, ERCOT
- The New Arrivals: Tesla, SpaceX, Apple, Optimus
- The Workforce Question: Replacement & Reskill
- External Benchmarks: McKinsey, Deloitte, WEF
- Strategic Implications for Texas Executives
- Conclusion: Readiness as Capability
Texas is building
the largest AI infrastructure
buildout in modern history.
It is also, by the directional evidence in this survey, an economy whose people remain quietly unsure who owns AI, what it should change, and whether it is operational reality or innovation theater.
This report sets out to do something unusual for an AI readiness document: it tells the truth about a small sample. Eighteen Texas-connected professionals completed our State of AI Readiness Survey out of more than ten thousand invited to participate over a three-month field window. The instinct in our industry is to bury that number. We have chosen instead to lead with it, because the silence of the other 9,982 is, we believe, the most important finding in the entire study.
Around those eighteen respondents stretches a state that is, simultaneously, the global epicenter of AI infrastructure investment. Stargate — OpenAI, Oracle, and SoftBank's flagship $500-billion AI data center program — is rising on roughly one thousand acres outside Abilene. Apple is building a 250,000-square-foot AI server factory in Houston, the cornerstone of a $600-billion U.S. investment commitment. Tesla has relocated its global headquarters to a 2,500-acre campus on the Colorado River, launched its first robotaxi service in Austin, and intends to manufacture humanoid robots at a scale of one million per year. SpaceX has, this year, become its own city. And in Houston, the largest medical complex in the world — handling ten million patient encounters annually — is starting to ask what AI means at its scale.
The gap between these two pictures — the trillion-dollar buildout and the eighteen-respondent survey — is not an inconsistency. It is the story of this report.
all 18 respondents — between
"Inconsistent" and "Established"
on a list of 10,000+
Texas professionals
announced for Texas
data centers alone
ERCOT by 2030 — more
than 2× peak summer load
The five headlines you should walk away with
Leadership intent is high. Discipline is not.
Respondents scored "leadership has a clear point of view on why AI matters" at 4.0 — the strongest signal in the survey. Yet "structured decision frameworks when adopting new AI tools" came in nearly a full point lower. Texas leaders believe AI matters. They do not yet have a shared method for deciding what to do about it.
The bipolar sample tells its own story.
Seven respondents came from organizations of 1,000+ employees. Seven came from organizations of 1–10. Two Texases answered this survey — the enterprise Texas and the founder Texas — with strikingly similar composite scores (3.52 vs. 3.49) but radically different constraints.
Energy moves slower publicly than privately.
Among industrial respondents, physical-safety AI guardrails scored the highest (4.20) of any energy-specific item. The discipline exists. What is missing is the willingness to talk about it. Industrial Texas is experimenting more than its public posture suggests.
Healthcare is the loudest silence.
One healthcare respondent. From a city that hosts 10 million annual patient encounters. The healthcare-specific section of this survey is, by sample, almost entirely empty. That emptiness is itself a finding worth a hospital board's attention.
The infrastructure is arriving faster than the readiness.
Texas will host gigawatts of net-new AI compute by 2027. Apple's Houston server line is reportedly ahead of schedule. Tesla's Optimus production target is one million robots per year. None of these timelines wait for organizational readiness to catch up. The implication for Texas leaders is plain: the question is no longer whether AI will reshape your business model, but whether your operating model will be ready when the infrastructure surrounding you is finished.
The gap we observed is not a technology gap. It is an organizational gap, sitting in plain view, surrounded by the largest AI infrastructure buildout in American history. — Editorial Note, AI Nexus Summit 2026
A state quietly
becoming the physical layer
of the AI economy.
Before reading the survey, it helps to see the ground it sits on. Texas in 2026 is not merely an AI market. It is becoming the substrate.
For most of the past century, Texas was a place where energy was produced. For the next century, it appears poised to become the place where intelligence is produced — because intelligence, at scale, runs on energy, land, talent, and a permissive regulatory posture, and Texas has more of all four than any other state in the union.
The structural picture, as of early 2026, looks like this:
in the greater Houston region
across ~10 facilities
capacity sits in Houston
and its ship channel
at the Texas Medical Center —
the world's largest medical complex
the nation's #2 cluster,
concentrated around DFW
Houston alone produces a daily volume of refined product that, if it were a country, would rank among the top fifteen oil-refining nations on Earth. The Texas Medical Center, sitting fifteen minutes from those refineries, performs one surgery every three minutes and delivers one baby every twenty minutes, year after year. Just outside Austin, Tesla's 2,500-acre Gigafactory campus — the second-largest building by volume in the world — produces Model Ys and Cybertrucks alongside the company's global executive offices. And two hundred miles southwest, near the mouth of the Rio Grande, SpaceX has voted itself into being as the city of Starbase, Texas.
Onto this industrial substrate is now landing what is, by dollar value, the largest single technology buildout in American history.
The map of AI Texas, 2026
If you drew a single map of what is happening in this state right now, you would mark at least seven different things, and each of them would matter on its own:
For the first time in modern history, Texas is not simply selling energy to the rest of the country. It is selling intelligence — and intelligence is the most energy-hungry product the state has ever exported.
The honest sample,
and what it is not.
This is not a statistically representative survey. It is something more useful: a directional reading from a small, voluntarily self-selecting group of Texas professionals during a particular moment in the AI cycle.
- Instrument
- 56 closed-response readiness questions across four cross-cutting dimensions (Leadership & Decision Discipline, Data Readiness & Governance, People & Adoption, Financial & ROI) plus three sector-specific batteries (Energy, Healthcare, SMB). All items used a 6-point semantic scale: Not Present (1) · Informal (2) · Inconsistent (3) · Established (4) · Embedded (5) · Proven (6).
- Distribution
- The survey was distributed to a Texas-connected list of more than 10,000 professionals over a three-month field window, primarily through email and professional-network channels emphasizing energy, healthcare, and industrial services.
- Response
- n = 18 completed submissions. This is a 0.18% response rate. We do not hide that number. We treat the size of the non-response as the survey's most important data point, and we devote an entire section of this report to it.
- Respondent Profile
- Roles ranged from C-Suite Executive (2), CDAO (1), CFO (1), VP (1), Director/Head of Function (3), Manager (2), Owner/Founder (1), Advisor (1), Partner (1), Individual Contributor (2), and others. Functions skewed toward Technology/IT/Digital (6), Finance/Risk (3), Strategy/Transformation (3), and Operations/Delivery (2).
- Industry Mix
- Heavy representation from Oil & Gas, Chemicals, Consulting & Advisory, Software/SaaS, Engineering Services, and Energy Transition; one healthcare respondent; one academic.
- Company Profile
- Bipolar by design and by response: 7 respondents from organizations of 1,000+ employees; 7 from 1–10; 3 from 51–200; 1 from 201–500. Revenue range was similarly bipolar: 6 above $250M, 6 below $5M.
- Limitations
- Small sample. Voluntary self-selection. Self-reported maturity ratings. We do not generalize from these data to "the Texas market"; we read them as directional commentary from a thoughtful self-selecting minority, then triangulate against external benchmarks from McKinsey, Deloitte, the World Economic Forum, the U.S. Chamber, and others.
- How to read this report
- Every quantitative claim is sourced: survey, external benchmark, or editorial interpretation. We do not blur these categories. Where we are reading a pattern rather than measuring one, we say so.
What the sample looks like, visually
What the 9,982
were telling us
by not answering.
In a year when "AI" is the most-spoken word in every executive deck on Earth, ten thousand Texas professionals were invited to articulate their organization's posture on it. Eighteen did. The other ninety-nine point eight percent are, we think, the most interesting respondents in this study.
— of professionals invited to share their organization's AI readiness chose not to respond. That is not an apology. It is, on the directional evidence, the loudest signal in the dataset.
Each dot represents one of the 10,000 invited.
18 responded. The other 9,982 are this report's other dataset.
What might silence be saying?
It would be easy — and wrong — to read a 0.18% response rate as apathy. Our reading of the silence, supported by external research and triangulated against respondent commentary, is that it represents a far more textured set of organizational states. Many or all of these are likely operating simultaneously, in different organizations:
Reputational caution
To declare an AI maturity level publicly is to invite either being called overhyped (if you say "Embedded") or being called behind (if you say "Exploring"). The safest posture is to say nothing.
Governance ambiguity
Many organizations do not yet have a clear answer to who is even authorized to speak about our AI position. CIO? CDAO? COO? CEO? Communications? In the absence of an answer, no one speaks.
AI fatigue
The third major AI survey of the quarter receives the lowest response rate. The market is saturated; legitimate research signals through the noise of vendor questionnaires.
Internal political sensitivity
In industrial sectors especially, AI is a workforce question, a union question, a safety question, and a capital-allocation question. Senior leaders are not eager to commit early positions to writing.
Confidentiality posture
For many enterprises — particularly in oil & gas, healthcare, and defense-adjacent industries — sharing operational maturity ratings, even pseudonymously, runs afoul of legal and communications conventions.
Genuine uncertainty
And then, perhaps most importantly: many recipients honestly did not know how to answer. Their organizations have AI activity, but no shared map of what "readiness" even means.
Silence, at this scale, is not the absence of a signal. It is a signal about the absence of organizational consensus on what to say. In a state spending hundreds of billions on AI infrastructure, that absence is the headline.
What this means for the survey's interpretation
We did not get the typical "industry pulse" sample of an enterprise survey. We got the eighteen organizations that were willing and able to put a number next to a maturity claim. They are not representative of Texas as a whole. They are, almost by definition, the more reflective end of the distribution — the operators who have done enough internal thinking to feel comfortable scoring it.
That is a feature, not a bug. The signal in this small dataset is unusually high-quality because the respondents are unusually deliberate. The implication is the reverse: the other 9,982 may be, on average, less ready than our eighteen, not more.
What they said
when no one was writing it down.
A short editorial reflection on the conversations that ran alongside the survey — what we heard off the page, and why it deserves a place in this report.
Surveys ask the questions the survey designers want answered. People say what they are willing to say in writing. In the months we were fielding the State of AI Readiness Survey, members of the editorial team spoke informally with a number of Texas professionals who had received the instrument but, for one reason or another, did not return it. We promised them — and we maintain here — that nothing about those conversations would identify anyone. What follows is not data. It is the candid texture the data, by its nature, could not capture.
The reluctance, named
Several recipients told us, in private, what we already suspected from the response rate: putting an AI maturity rating in writing felt risky. Some worried about being seen as over-claiming. Others worried about being seen as behind. A few were waiting for their organization to issue an internal posture before committing personal answers to anything external. The pattern was less apathy than caution — caution that has not yet found a vocabulary it trusts.
ROI was real — but it lived in "pet projects"
The most striking pattern across the conversations was this: many of the people we spoke with had personally seen AI deliver measurable return. Almost always, those returns had come from what they themselves called pet projects — small, independent, sometimes under-the-radar initiatives that solved a specific problem for a specific team. Hours saved on reporting. Faster anomaly detection on a single line. A new way of summarizing inspection findings. The ROI was honest. What was missing was the bridge from those wins to enterprise-scale practice. The wins existed inside the organization; the institutional mechanism to scale them did not.
The sectors that were scaling — and the ones that were not
Manufacturing and data analytics consistently came up as places where AI was working at a level beyond individual experimentation. Process maturity already existed there. The data was structured. Outcomes were measurable. Adoption fit naturally into existing operating disciplines, and results showed up quickly enough to attract more investment.
Field services, back-office functions, and customer-facing operations told a different story. Pilots were happening. Some were excellent. But scaling was elusive — not because the technology was not ready, but because the operating model around it was not. Standard work, change management, training, and the documented playbooks that let a good idea travel across a workforce were the things that turned out to be limiting, not the tools.
Shadow AI is real
Several conversations surfaced the same fact in different vocabulary: people are already using AI in their daily work, whether or not their organizations have officially sanctioned it. Employees are running their own ChatGPT, Copilot, and Gemini sessions to draft, summarize, analyze, and decide. Their managers may or may not know. Compliance and security teams may or may not know. In organizations without a clear position, shadow AI fills the vacuum — and brings real productivity gains alongside real, undiscussed risks.
The readiness gap, when you actually listen
The composite finding from the conversations was unambiguous. The hard part of AI readiness was not the technology. It was people, process, mindset, and culture. The organizations whose people we heard described as adapting well had two things in common: existing process maturity, and people who treated the new tools as extensions of how they already worked rather than as threats to it. Where one of those was missing, AI adoption stalled regardless of investment. Where both were present, returns showed up quickly.
Where existing process was good and people were adaptable, AI worked. Where either was missing, no amount of tooling closed the gap. The conversations did not reveal a technology shortage. They revealed a readiness shortage with a human face. Editorial Synthesis · Off-the-Record Conversations · 2025–2026
Why this belongs in a survey report
The eighteen respondents on the record built the numerical scaffolding of this report. The conversations off the record gave us something the numbers could not: the texture of why the responses look the way they do, and why so many recipients chose not to respond at all. We have kept all sources anonymous and no individual organization, project, or person has been identified. The picture they collectively painted is consistent enough with the survey data and with the external benchmarks that we believe it deserves a place in the document — not as evidence, but as honest commentary alongside the evidence.
Where Texas actually sits
across four readiness dimensions.
An overview of the directional patterns observed across our eighteen respondents, before we look at each dimension in detail.
The headline composite finding across our sample is this: Texas organizations sit, on average, between Inconsistent and Established on the maturity scale — a composite of 3.5 out of 6. They are doing real things with AI. They have not yet built the institutional muscle to do them consistently or to know whether they are working.
The shape of the curve
Read across the four cross-cutting dimensions and you see something interesting: scores do not vary as much as you might expect. People & Adoption (3.72) edges out Leadership (3.51), Financial Discipline (3.42), and Data Readiness (3.40), but the spread is narrow. This is the signature of organizations that are roughly equally mature — or equally immature — across the board, rather than organizations that have made disproportionate progress in one area.
External benchmarks agree with this directionally. McKinsey's State of AI 2025 reports that 88% of organizations now use AI in at least one function, but only about one-third have begun to scale it across the enterprise. Deloitte's 2026 report frames the same gap differently: 66% of organizations report productivity gains from AI, but only 34% are using AI to "deeply transform" their business.
Our Texas sample looks like a microcosm of that broader pattern: broadly adopted, narrowly transformative.
The five strongest signals
The questions that scored highest across our 18 respondents (on the 1–6 scale) tell a coherent story:
These are intent metrics. They measure what people say about AI, how they feel about it, and what they expect it to do. Texas respondents score themselves well on intent.
The five weakest signals (within the cross-cutting dimensions)
These are discipline metrics. They measure governance, frameworks, data quality, and the willingness to kill failing initiatives. Texas respondents score themselves substantially lower on discipline than on intent.
The Texas readiness profile, in one sentence: high intent, established adoption, inconsistent discipline. We know we want AI. We have some of it working. We do not yet know how to govern it, when to stop it, or whose data it should run on.
The bipolar sample, revisited
One of the most striking findings in our data is how similar enterprise Texas (1,000+ employee firms) and founder Texas (1–10 employee firms) look on a composite basis — 3.52 vs. 3.49 — despite having radically different constraints. That number is a useful illusion. What it hides is where the two ends of the spectrum struggle.
The clearest gap
in the dataset is between
intent and method.
Texas leaders broadly know why AI matters. Far fewer have a shared method for deciding what to do about it.
The leadership dimension was our largest battery — twelve questions covering everything from "leadership has a clear point of view on why AI matters" to "we use structured decision frameworks when adopting new AI tools." The pattern within it is consistent and, for executives reading this report, instructive.
What the curve says
Read top-to-bottom, this is a near-perfect leadership maturity ladder, and Texas organizations are sitting unevenly on it. They have strong conviction about AI ("leadership has a clear POV" → 4.0). They have moderately developed operating discipline ("decisions are made deliberately" → 3.61). They have the weakest performance on institutional methodology — structured frameworks, criteria for when AI should and should not be used, and the discipline to avoid defaulting to AI for speed alone.
This matches what McKinsey identifies as the defining trait of AI high performers: workflow redesign and structured leadership ownership. McKinsey's 2025 data finds that high performers are 3.6× more likely than peers to be aiming for transformational change, and 55% have fundamentally reworked processes when deploying AI — almost three times the rate of other firms.
Our Texas sample's profile — strong intent, weaker frameworks — is consistent with the broader market's "ambition outpacing operating model" pattern. The Deloitte 2026 enterprise report puts it more starkly: only 34% of organizations are using AI to "deeply transform" their business.
Leadership conviction is necessary and insufficient. The Texas leaders in our sample have the conviction. What they describe in the data is an organization that has not yet built the muscle of method around that conviction.
The "AI should not be used here" question
One leadership item in our survey asks whether leaders understand where AI should not be used. It averaged 3.5 — squarely in the "Inconsistent / Established" band. We single this out because, in our reading of the data and the external research, it is one of the most underweighted questions any organization can ask itself.
Deloitte's 2026 report finds that only one in five companies has a mature governance model for autonomous AI agents — yet 85% of companies expect to customize agents to their business in the near term. The "where should we not use AI" question is, increasingly, where competitive advantage lives. Texas leaders score themselves cautiously here, and they are right to.
The dimension where Texas
scored itself lowest.
Across eight items spanning data accuracy, access, ownership, regulatory understanding, and incident response, Texas respondents averaged 3.40 — the weakest of the four cross-cutting dimensions.
Why this matters disproportionately
Every external benchmark we triangulated against names the same culprit when AI initiatives fail to scale: data foundations, not models. McKinsey's 2025 report identifies "fragmented data and legacy tech" as one of the three persistent blockers preventing organizations from moving out of "pilot purgatory" — the state in which AI experiments never graduate to production. The Salesforce SMB Trends Report finds that 85% of IT professionals confirm "AI outputs are only as good as data inputs."
Our respondents understand this. The single highest-scoring item in this dimension is "we understand where data gaps limit AI effectiveness" (3.78) — meaning Texas organizations have, on average, accurately diagnosed the problem. The substantially lower scores on data accuracy (3.22), data ownership (3.17), and a clear approval process for new AI tools (3.28) suggest the diagnosis has not yet translated into action.
For an industrial state — where operational data sits in SCADA systems, historians, ERP instances, and a long tail of paper, where physical-asset data is often locked inside vendor systems, and where regulated data sits behind strict legal walls — the data-foundation problem is harder than in software-native economies. This is not a Texas-specific failing; it is a Texas-specific difficulty multiplier.
The incident-response question
One item in the data battery sticks out for executives in regulated industries: "We know how we would respond if an AI system caused harm or failure." Texas respondents scored themselves at 3.50 — squarely in the "Inconsistent / Established" band.
That is, on average, not yet a credible answer. In an industrial state where an AI-influenced decision could plausibly affect a refinery flare, a pipeline pressure setpoint, a clinical recommendation, or a robotaxi route, the absence of a rehearsed incident-response posture is one of the more sobering findings in this dataset. This is one of the most actionable items in the report.
The infrastructure to build AI in Texas is arriving at a rate the data foundations underneath it cannot match. That mismatch is the single largest organizational risk in our sample.
The strongest dimension —
and the one where Texas
looks most like its workforce.
Texas respondents scored their organizations highest on the People dimension (3.72), driven primarily by employee openness to AI and visible early efficiency gains.
The two best-scored items, in context
"Employees feel safe raising concerns about AI use" averaged 4.17 — the highest score in our entire survey. "AI is improving efficiency, quality, or speed in at least one area today" averaged 4.06. Read together, these are encouraging signals about the bottom-up reality of AI in Texas workplaces. People are using it. People are willing to talk about it.
This matches a striking finding from Deloitte's 2026 enterprise report: worker access to AI rose by 50% in 2025, with the share of workers equipped with sanctioned AI tools growing from under 40% to around 60% in a single year. AI has, in a meaningful sense, already arrived in the workforce. The question now is what executives do with that arrival.
The middle scores, where the work lives
"Resistance to AI adoption is understood and actively managed" averaged 3.39. "Teams adapt effectively when AI changes workflows or roles" averaged 3.39. "AI initiatives are reviewed based on outcomes, not novelty" averaged 3.50. These are the items that distinguish using AI from operating with AI, and they are exactly where Texas organizations show the most variance and the most opportunity.
The PwC AI Jobs Barometer found that industries with higher AI adoption have seen productivity growth rates four times higher than less-AI-intensive sectors. The opportunity is not theoretical — it is measurable, and it is currently uneven. Workforce strategy is, in our reading, the single most underdeveloped lever in the Texas executive toolkit.
Most Texas organizations
can describe AI's expected value.
Few can describe its delivered value.
The Financial & ROI dimension averaged 3.42 — second-weakest after Data — and the shape of that average is more interesting than the number.
The expectation-to-realization gap
"AI investments are expected to produce measurable business return" scored 3.89. "Actual outcomes of AI initiatives are reviewed against expected value" scored 3.39. "AI initiatives are adjusted or stopped when expected value is not realized" scored 3.11 — among the bottom five scores in the entire survey.
Translation: Texas leaders expect AI to pay back. They are not yet measuring whether it is. They are even less willing to stop it when it isn't.
This is consistent with — and arguably worse than — the global pattern. McKinsey's 2025 data finds that only 39% of organizations report any measurable effect on enterprise-level EBIT from AI, and the small group it calls "AI high performers" — organizations where more than 5% of EBIT is attributable to AI — represents only about 6% of all respondents.
The hardest organizational muscle to build, in our reading of the data, is not starting AI initiatives. It is stopping them.
The healthiest item in the ROI dimension is the most basic: "we can clearly explain how AI is expected to create business value" (3.83). The least healthy is the most disciplined: "we intentionally distinguish between experimental AI efforts and ROI-driven investments" (3.06). The bridge between these two scores is, in our view, where executive attention will produce the most return in 2026.
Houston refines
a quarter of America's crude.
It is also refining its AI posture.
Five of our 18 respondents answered the Energy-specific battery. Their scores, surprisingly, were the strongest of any sector-specific battery — averaging 3.88 on the 1–6 scale.
To set the stage on what "energy in Texas" actually means at scale: the Houston metropolitan region operates roughly 10 refineries with a combined ~2.6 million barrels per day of crude processing capacity. The Texas Gulf Coast accounts for more than 87% of the state's refining capacity and more than a quarter of the entire United States' refining capacity. Houston alone holds ~42% of U.S. base petrochemical capacity. Texas is the leading U.S. natural gas producer, accounting for roughly 28% of total national production.
Onto this physical economy, AI is arriving in two simultaneous and very different ways:
AI as customer of energy
Texas data centers — Stargate, the Apple Houston AI server plant, and the wider DFW hyperscale cluster — are about to consume gigawatts of new electricity. ERCOT is forecasting data center demand to reach 77,965 MW by 2030, up from ~29,600 MW in the 2024 projection. The state's wholesale market is restructuring around this new load.
AI as tool of the energy operator
Inside the refineries, chemical plants, and utilities themselves, AI is being applied to maintenance prediction, throughput optimization, safety monitoring, well-log interpretation, and trader-desk decision support. This is the AI our energy respondents were scoring.
What the energy respondents told us
The single strongest score in the entire Energy battery was on "AI is not used to automate decisions that could impact physical safety without defined human oversight" — 4.20 of 6.0. The single strongest score on financial discipline anywhere in the entire survey was "AI initiatives are expected to improve uptime, throughput, cost efficiency, or risk exposure in measurable ways" — also 4.20.
This profile is, frankly, encouraging. The Texas energy respondents in our sample appear to have internalized the right rules early: humans stay in the loop where physical consequence is at stake, and AI investment is justified against the same operational improvement criteria as every other capital project. That is a more mature posture than many software-native industries demonstrate at the same scale of adoption.
The reliability question
One Energy-specific item asks whether "AI solutions are designed to function reliably in constrained or low-connectivity environments." It scored 3.40 — the weakest in the battery. This is the industrial reality check: a large amount of operational AI has to work on a wellhead in west Texas, on a platform in the Gulf, or inside a refinery control room where the latency budget is unforgiving and the network is not always there. The score reflects that this is, plausibly, where the work still is.
Our Texas energy respondents scored higher on physical-safety discipline than on every other readiness dimension we measured. The most important thing about that finding is that it is almost impossible to detect from the industry's external posture.
The grid is the constraint
One more piece of context belongs in this section. The bottleneck on AI's near-term growth in Texas — and therefore on the energy industry's exposure to AI both as a buyer and as a participant — is not chips. It is power.
requests filed with ERCOT
for connection by 2030
the state's record 2025
summer peak demand
requests are data centers,
per ERCOT
in 2023 to finance new gas plants
backstopping the AI grid load
The Stargate I campus in Abilene, when fully built, will require approximately 1.2 GW of power — enough to supply roughly one million Texas homes for a year. ERCOT has added about 23 GW of new generation capacity between 2024 and 2025; another 9 GW is slated for early 2026; and the long-term load forecast suggests peak demand could reach 139 GW by 2030. The Texas grid is, in 2026, being rebuilt at speed around the load profile of AI compute.
One healthcare voice
from the world's largest medical complex.
In a city that hosts ten million patient encounters per year, our survey received exactly one healthcare-specific response. That fact is, in itself, the section's most important finding.
The Texas Medical Center, fifteen minutes from downtown Houston, employs more than 120,000 people across 21 hospitals, 8 academic and research institutions, 4 medical schools, and 60+ member institutions. It performs more heart surgeries than any complex in the world. It delivers a baby every 20 minutes. It is home to the world's largest cancer hospital (MD Anderson) and the world's largest children's hospital (Texas Children's). And our survey's Healthcare-Specific battery — covering AI clinical decision support, patient safety, PHI governance, bias evaluation, clinician trust, and AI-related harm response — received exactly one valid response.
at the Texas Medical Center
2.1-square-mile campus
complex — TMC alone
of the Texas Medical Center
What one response tells us
The single healthcare respondent in our sample scored every question in the Healthcare-Specific battery at 1 — Not Present. We do not, for one second, believe that is representative of the actual state of AI in Houston's medical complex. We believe it is representative of the state of declarable AI in Houston's medical complex — the AI that an organization is willing to publicly attest to having governed.
Healthcare is, by far, the most regulated of the sectors covered in this survey. HIPAA, FDA AI/ML device guidance, state medical-board scope-of-practice rules, payer-side reimbursement constraints, and institutional IRB processes all combine to make speaking publicly about AI maturity — even in a confidential survey — a non-trivial governance act in itself. The silence here is not absence; it is institutional caution.
What we know from external research
External signals strongly suggest that AI activity inside the Texas Medical Center is substantial and accelerating. The complex has been advancing $3 billion in construction projects, including the TMC3 / Helix Park innovation campus, which opened in 2024 and has been actively recruiting partnerships with AI-in-health programs (including Baylor College of Medicine's AI in Health Lab). Notable milestones include BiVACOR's first-in-human Total Artificial Heart implantation, conducted at the Texas Heart Institute in collaboration with Baylor St. Luke's and Baylor College of Medicine — a procedure that depends heavily on real-time machine learning.
McKinsey's 2025 data identifies healthcare as one of the leading industries in scaled AI adoption — alongside technology, media, and telecommunications. The pattern globally suggests that healthcare's caution is operational, not strategic. The institutions are moving; they are just not advertising it in surveys.
A single response is not a statistic. It is an institutional signature — and the signature here reads "we are not yet ready to discuss this on the record."
What healthcare leaders should be asking
If we were preparing an executive briefing for a hospital system's board in 2026, the questions we would put on the agenda — drawn from the items we could not score in this survey — would include:
Clinical decision boundaries
For each AI system in clinical workflow, is it explicitly used to support a clinician's judgment, or to make one? Has that boundary been documented and tested?
Bias evaluation cadence
Are AI tools evaluated for demographic, socioeconomic, and clinical bias both before and after deployment, with the data of our specific patient population?
PHI governance
Does AI governance explicitly address protected health information at the level of architecture, not just policy? Where does PHI sit in our AI pipelines?
Override and escalation
Do clinicians and staff know when and how to override or escalate an AI-generated recommendation? Has that pathway been rehearsed at the bedside?
Cognitive load
Is the AI we have deployed reducing cognitive and administrative burden on clinicians, or adding to it? Have we measured that, post-deployment, with the clinicians themselves?
Adverse outcome response
If an AI system contributed to a patient adverse outcome tomorrow, do we know who decides what to do, what we tell the patient, and how we report it?
Founder Texas:
fast adopters,
capacity-trapped.
Nine respondents answered our SMB-specific battery. Their scores tell a coherent story: small Texas firms have outsized AI ambition, almost no AI strategy, and absolutely no time to fix the gap.
The capacity trap
The single lowest-scoring item in our entire survey, after the unanswered Healthcare battery, was the first SMB question: "We have sufficient time and capacity to pursue AI initiatives we believe are valuable." Average: 3.00. This is what we are calling the capacity trap, and it is the defining constraint of founder Texas.
Small Texas businesses are, on average, perfectly capable of identifying AI initiatives that would be valuable to their organizations. They have no spare hours in the week to actually pursue them. The strongest score in the SMB battery — "AI investments are expected to buy back owner or leadership time, not just reduce costs" (4.11) — is the mirror of the weakest. SMB owners know exactly what they need: their time back. They cannot find the time to get it back.
External benchmarks broadly agree
The U.S. Chamber of Commerce's 2025 small business report finds that 58% of small businesses now use generative AI, up from 40% in 2024 and from just 23% in 2023. The SBA's longitudinal analysis shows the large-vs-small adoption gap shrinking from 1.8× in early 2024 to near-parity by August 2025. The momentum is real.
But the same external research surfaces the consistent SMB barriers: integration friction (72% report it as a challenge), data and privacy concerns (70%), and — most striking — that 82% of the smallest SMBs (under 5 employees) cite "AI isn't applicable to my business" as their reason for non-adoption. The U.S. Chamber's data calls this an "education rather than applicability issue."
The strategic gap
"AI tools are adopted as part of a defined strategy rather than opportunistically" averaged 3.00 in our SMB respondents — tied for the lowest score in the battery. This is the second face of the capacity trap. Founder Texas is adopting AI, but without strategy. The result is an accumulation of tools — chatbots, transcript summarizers, marketing copy generators, image makers, scheduling agents — that improve individual workflows but rarely add up to a coherent business advantage.
SMB Texas's AI story, in one sentence: I know I need it, I'm using some of it, I have no idea if it's working, and I have no time to find out.
The implication for SMB-serving institutions — banks, CPAs, business associations, and the major SaaS platforms operating in Texas — is significant. The bottleneck on SMB AI value is not access to tools. It is the cost of integration, the cost of strategy, and the cost of time. The vendors that solve those costs first — particularly through agentic AI that takes work off the owner's plate — will likely capture disproportionate share.
Stargate, Apple, ERCOT —
and the question of who is ready
for what is being built.
If our survey is a thermometer for organizational readiness, the AI infrastructure being installed in Texas is a tidal wave. The two should be read together.
Stargate — the flagship
The Stargate Project, announced from the White House in January 2025 by OpenAI, Oracle, and SoftBank with President Trump, is the largest AI infrastructure venture in history. Its flagship campus sits on roughly 1,000 acres outside Abilene, Texas, at the Lancium Clean Campus. The first phase — two buildings totaling over 200 MW — was energized in mid-2025 on Oracle Cloud Infrastructure, running Nvidia GB200 racks. The eight-building campus, when complete, will support hundreds of thousands of GPUs on a single integrated network fabric, drawing approximately 1.2 GW of power.
In September 2025, Stargate announced five additional sites including two more in Texas — Shackelford County and Milam County. By that announcement, the program had crossed $400 billion in committed investment and nearly 7 GW of planned capacity, putting it on track for its $500 billion / 10 GW commitment by the end of 2025. In March 2026, Microsoft announced it was taking on additional buildings on the Abilene campus, becoming a Stargate neighbor.
Apple's Houston AI server plant
In February 2025, Apple announced a $500 billion U.S. investment commitment over four years (later expanded to $600 billion). Its centerpiece manufacturing project: a 250,000-square-foot AI server factory in Houston, built in partnership with Foxconn. The factory is producing servers for Apple Intelligence and Private Cloud Compute. By October 2025, Apple confirmed the facility was building and shipping American-made AI servers ahead of schedule. The company has stated it will employ "thousands" at the facility.
Beyond Apple itself, the broader Houston AI hardware corridor is taking shape: Applied Optoelectronics broke ground in Sugar Land, Texas in February 2026 on a 210,000-square-foot manufacturing facility for optical transceivers used in AI data center networking. AOI's CFO publicly remarked that "the state of Texas has done a phenomenal job in positioning itself to be the leader in AI."
The DFW hyperscale cluster
As of September 2024, Texas already hosted 279 operating data centers — the nation's second-largest concentration. More than half of those sit in the Dallas-Fort Worth area. ERCOT is forecasting data center demand growth from approximately 29.6 GW (2024 projection) to 77,965 MW (2030 projection) — and that is the conservative forecast. The "extreme" interconnection-request pipeline as of late 2025 was 220+ GW, of which approximately 70% were data centers.
The grid response
Texas is restructuring its electricity market around this load profile in real time. The 2023 Texas Energy Fund — $9 billion in low-interest loans and grants for new gas plants — was created specifically to backstop the new demand. Senate Bill 6 (2025) imposes performance requirements on data centers and other large loads during emergency grid conditions. ERCOT has added approximately 23 GW of new generation capacity between 2024 and 2025; another 9 GW is slated for early 2026. New 765-kV transmission lines, approved by the PUCT, can carry more than twice the voltage of current infrastructure. Private partnerships — Dow + X-energy for 320 MW of small modular nuclear, Oklo + Diamondback Energy for Permian Basin microreactors, Last Energy for 30 microreactors across ERCOT — are rewriting what Texas's energy mix will look like by 2030.
The infrastructure for AI in Texas is being built faster than the organizational readiness to use it. That is not a criticism. It is a strategic warning to every executive in this state.
Tesla, SpaceX, Apple —
and the vertical integration
of AI in Texas.
In addition to the data center buildout, three of the most aggressive vertically integrated AI companies on Earth have all made Texas their physical center of gravity. The implications for the state's labor market, supply chain, and competitive landscape are non-trivial.
Tesla — the AI car company and its humanoid robot
Tesla relocated its global corporate headquarters from California to Gigafactory Texas, just outside Austin, in December 2021. The campus now spans 2,500 acres along the Colorado River with more than 10 million square feet of factory floor — equivalent to roughly 100 football fields. It is the second-largest building by volume in the world. The factory produces the Model Y for the Eastern United States and is the global production hub for the Cybertruck. Documents filed in 2024 indicated an additional 5-million-square-foot expansion underway, expected to complete by the end of 2025.
In June 2025, Tesla launched its Robotaxi service in Austin — initially with safety monitors in modified Model Ys. By January 2026, Tesla began removing safety monitors from selected rides. By March 2026, Tesla announced expansion to Dallas and Houston with unsupervised vehicles, and plans to expand to seven additional cities in the first half of 2026. As of December 2025, Tesla operated approximately 135 robotaxis. The Cybercab — Tesla's purpose-built, steering-wheel-free robotaxi — is scheduled for volume production starting in 2026.
And then there is Optimus. Tesla's humanoid robot program is, per company statements, preparing for first-generation mass production in 2026 with an eventual planned capacity of one million units per year. The Gen 3 version, the first design intended for mass production, is being unveiled in Q1 2026. The robot is positioned as Tesla's biggest product ever, with Musk forecasting that by 2040 there would be more humanoid robots than people, and that Optimus alone could anchor a $25 trillion company valuation.
SpaceX — a company that incorporated itself into a city
On May 3, 2025, voters near Boca Chica Beach approved the incorporation of Starbase, Texas, the first new city in Cameron County since 1995. The vote passed 212 to 6. Both commissioners and the mayor are SpaceX executives. The city now includes the SpaceX launch facility, manufacturing complex, and company-owned land covering a 1.6-square-mile area, home to roughly 500 permanent residents. In September 2025, Cameron County formally turned over portions of Boca Chica Beach to Starbase's jurisdiction.
The Starship program is, increasingly, not just an aerospace platform. Its development pipeline depends heavily on machine learning for trajectory optimization, engine telemetry, and autonomous catch-and-recover operations. Starbase is, in many practical ways, the first incorporated city in U.S. history whose civic infrastructure is co-designed with an AI-and-robotics research program.
Apple — the world's most valuable company picks Houston
Apple's $600 billion U.S. investment commitment is the largest in the company's history. The geographic centerpiece is in Houston: a 250,000-square-foot facility partnered with Foxconn to produce AI servers for Apple Intelligence and Private Cloud Compute. The investment includes expansion in Michigan, Texas, California, Arizona, Nevada, Iowa, Oregon, North Carolina, and Washington, but Houston is the manufacturing tip of the spear. As of October 2025, Apple confirmed the Houston facility was operating ahead of schedule.
And the second tier of arrivals
Beyond the three flagship AI companies, the supplier and adjacent-technology base is densifying around them. Applied Optoelectronics, headquartered in Sugar Land, broke ground on its expanded 210,000-square-foot facility in February 2026, targeting the AI and data center transceiver market with a planned $300 million investment by end of 2026. Cummins, Chevron Phillips Chemical, Bureau Veritas, Dow, Southern Company, and dozens of other major industrial players surveyed in this study are all building AI capability in or via Texas operations. The list of established Texas companies actively deploying AI is longer than the list of new arrivals — and that, too, is part of the story.
Texas has, in a remarkably short time, become both the infrastructure layer of the AI economy and the vertical integration layer — where the chips are made, the data centers are run, the cars drive themselves, the humanoid robots are assembled, and the rockets fly. No other state is positioned this way.
Replacement,
creation, and the part
nobody is willing to say out loud.
No serious AI readiness report can avoid the labor question. Texas — with its industrial base, its medical complex, its energy workforce, and its incoming wave of humanoid robots — will be one of the country's most visible test beds for what happens when AI meets work.
Our survey was not designed to measure workforce displacement directly. But the People dimension surfaced a piece of data worth taking seriously: "Employees feel safe raising concerns about AI use" scored 4.17 — the highest score in the entire survey. Workers in Texas organizations are, by their managers' reports, willing to talk about their concerns. Whether anyone is listening is a different question.
What the global data says, applied to Texas
170M jobs created globally
The World Economic Forum's Future of Jobs Report 2025 projects 170 million new roles created globally by 2030, driven by technology, the green transition, demographics, and the rebalancing of the global economy. Fastest-growing roles include AI and data specialists, software engineers, FinTech roles, and renewable energy engineers.
92M jobs displaced globally
The same report projects 92 million roles displaced. Net positive (+78M) globally — but the gross displacement is real, painful, and concentrated. The IMF's 2024 assessment found ~40% of jobs globally face meaningful AI exposure; in high-income economies, ~60%.
40% of employers plan reductions
The WEF reports that 40% of employers expect to reduce their workforce in areas where AI can automate tasks. Among SMBs specifically, the SMB Group's 2025 research found 16% have already replaced jobs with AI, and 25% expect to do so within twelve months.
The Texas-specific exposure
Texas's industrial workforce is uniquely exposed in both directions. The state's largest employment sectors — energy, healthcare, construction, retail, transportation, food service — each have their own AI-exposure profile. Where Texas is unique is in the concentration of its industrial workforce in cities that are also the front lines of AI deployment.
Consider the Texas labor markets most exposed to specific AI vectors:
Indicative exposure assessment · synthesized from WEF Future of Jobs 2025, IMF AI Exposure, PwC AI Jobs Barometer, Veritone Q1 2025 labor data. Bottom row shows job growth rate.
The robotaxi case study, in real time
Tesla's Austin robotaxi launch is, in 2026, the most visible AI-vs-labor experiment running in the United States. Tesla operated approximately 135 robotaxis in Austin as of December 2025, with safety monitors. In January 2026, the company began removing safety monitors from selected rides. By April 2026, the service had expanded to Dallas and Houston with unsupervised vehicles.
The honest picture is more complex than "robots replace drivers." Tesla's Austin service has been involved in 14 crashes since June 2025, with five concentrated in December 2025 and January 2026, raising questions about whether the service is, at this stage of development, more accident-prone than human drivers. Waymo, in contrast, was delivering substantially more rides per day in more limited operational zones at the same time. Robotaxi is happening; it is not happening cleanly; and the labor market implications are still being measured.
The Optimus question
Tesla's publicly announced near-term production target for the humanoid Optimus robot is one million units per year, with pricing projected at $20,000–$30,000 per unit. During a May 2026 tour of Gigafactory Texas, however, Tesla personnel told one of the authors of this report that the company's internal ramp targets are substantially more aggressive: roughly ten million units per year as a medium-term goal, with an eventual long-range ambition of one hundred million units annually. Those numbers should be read as company-stated ambition rather than delivered capacity — Tesla has a long history of production timelines that slip — but even a meaningful fraction of them, manufactured in Texas, would have implications for the state's warehouse, fulfillment, light manufacturing, and basic-service workforces that are difficult to overstate. The Deloitte 2026 enterprise report finds that physical AI adoption is projected to reach 80% within two years across manufacturing, logistics, and defense — and Texas is the U.S. center of all three.
The most candid finding in our People dimension was the highest score: employees feel safe raising concerns about AI. The most candid finding in the external research is that 40% of employers expect to reduce headcount because of AI. Those two facts deserve to be read together.
What the upside actually looks like
The growth story is real and, in Texas, particularly real. Veritone's Q1 2025 labor market data recorded 35,445 AI-related job openings in the U.S., up 25.2% year-over-year. AI engineer roles surged 143.2% year-over-year in demand. Median pay for AI roles reached $156,998. The PwC AI Jobs Barometer found that workers with demonstrable AI skills earn on average 25% more than peers without them.
The state's universities — UT Austin, Rice, Texas A&M, UH, UT Dallas — are positioned to feed this market disproportionately. The community college system, especially in DFW and Houston, has been pivoting curricula toward AI-adjacent technical roles. The Texas Energy Fund and the new manufacturing investments come with workforce-development commitments. The infrastructure is being built, and so is the talent pipeline. The bridge between the two — reskilling for incumbent workers in industries about to be reshaped — is the public-policy question that will define Texas's next decade.
What McKinsey, Deloitte,
and the WEF tell us
around our 18 respondents.
The directional readings from our survey are not generalizable to "the Texas market." But they sit usefully alongside the world's most rigorous AI enterprise studies — McKinsey's State of AI 2025, Deloitte's State of AI in the Enterprise 2026, and the WEF Future of Jobs Report 2025.
Where we overlap with global enterprise surveys, our numbers agree.
Eighteen respondents cannot represent Texas. But where this study's questions overlap with comparable items in much larger global enterprise surveys — McKinsey's 1,993-respondent State of AI 2025, Deloitte's 3,235-respondent State of AI in the Enterprise 2026, and the WEF's Future of Jobs 2025 — the directional readings agree within a few percentage points.
The convergence is not coincidence. It suggests our small sample is reading the same underlying enterprise reality measured by surveys three orders of magnitude larger — and that the directional findings in this report, while not statistically representative, are not idiosyncratic to our eighteen respondents.
Three observations worth reading carefully
1 · Adoption is broader than transformation, globally and locally
McKinsey reports 88% of organizations now use AI in at least one function — up from 78% in 2024. But only about one-third have begun to scale it across the enterprise; two-thirds remain in what observers call "pilot purgatory." Deloitte's 2026 data finds the same gap: 66% of organizations are realizing productivity gains, but only 34% are using AI to "deeply transform" their business. McKinsey's small group of "AI high performers" — organizations attributing more than 5% of EBIT to AI — represents just 6% of all respondents globally.
Our Texas sample, with its composite score of 3.5 out of 6 across cross-cutting dimensions, sits in exactly the same band: adoption is real, transformation is rare.
2 · Leadership and workflow redesign are the differentiators
The McKinsey 2025 report's most useful finding is that AI high performers are 3.6× more likely than peers to be aiming for transformational change, and 55% have fundamentally redesigned workflows when deploying AI — almost three times the rate of other firms. Workflow redesign, not technology selection, is the single highest-correlated driver of AI-attributable EBIT impact.
Our survey's strongest signal — leadership conviction (4.0/6) — is also the necessary condition for that redesign to happen. Our survey's weakest signal — structured decision frameworks (3.06/6) — is what high performers actually do differently.
3 · Agentic AI is coming faster than governance is
Deloitte's 2026 report finds that only one in five companies has a mature governance model for autonomous AI agents — yet 85% of companies expect to customize agents to their business in the near term. McKinsey reports that 23% of organizations are scaling an agentic AI system somewhere in their enterprise, with an additional 39% experimenting. Yet within any specific function, no more than 10% of organizations say their agents are fully scaled.
Our survey did not measure agent governance directly, but the patterns it did measure — weak data ownership, inconsistent approval processes, low scores on "we know how we would respond if AI caused harm" — are the same patterns that produce poor agent governance at scale.
On every cross-cutting dimension where comparable benchmarks exist, our Texas sample looks like a slightly more cautious, slightly more industrial version of the global enterprise picture. That is not a weakness. It may be an advantage, if the state's leaders use the next 24 months to convert caution into method.
What Texas executives
should be doing differently
in the next 24 months.
Six implications drawn directly from the survey, the external benchmarks, and the infrastructure context of this state.
Convert intent into method
The single largest gap in our data is between leadership conviction (4.0) and structured decision frameworks (3.06). The first move is not another pilot — it is a documented method for evaluating AI use cases, killing the ones that don't work, and prioritizing the ones with the cleanest connection to operational improvement.
Name an AI accountable executive
Across our sample and the external benchmarks, the single sharpest dividing line between high performers and the rest is senior executive ownership — not delegated to a steering committee, but owned. The role title matters less than the singular accountability. If no one in the C-suite can answer "who owns AI here?" with their own name, that is the first problem.
Treat data work as the precondition
The lowest cross-cutting dimension was Data (3.40). The most-cited blocker in McKinsey's global data was the same. AI does not work on the data your organization wishes it had. It works on the data it actually has. The unglamorous data work — ownership, accuracy, access, lineage — is the work that AI returns on.
Rehearse the failure mode
"We know how we would respond if an AI system caused harm" scored 3.50. In an industrial state, that is not yet a credible answer. Run the tabletop exercise. Write the playbook. Identify the regulator, the spokesperson, the operational override path. Do it before the incident, not after.
Distinguish experiments from investments
"We intentionally distinguish between experimental AI efforts and ROI-driven investments" scored 3.06 — one of the lowest in the survey. Mixing the two creates organizational confusion in both directions: experiments are over-measured, investments are under-measured. A separate budget line, separate review cadence, and separate exit criteria for each.
Make the workforce conversation explicit
Our highest-scoring item was employees feeling safe raising concerns about AI (4.17). That latent psychological safety is rare and valuable. Spend it. Talk publicly, inside your organization, about which roles you expect to change, which to grow, and what investment in reskilling looks like. The companies that handle this conversation well will keep their best people through the transition.
The infrastructure is being built. The talent pipeline is being built. The question every Texas executive should be asking is whether their operating model will be ready when the rest of the state's AI substrate is finished. On the directional evidence in this survey, the answer in most organizations is: not yet.
Readiness is not a score.
It is an institutional habit.
A final reflection on what our 18 respondents, the 9,982 who did not respond, and the trillion-dollar buildout around them collectively suggest about the year ahead.
The temptation, in a report like this, is to end with a thunderous call to action. We will not. The honest reading of our data is that the call to action has, for most Texas organizations, already been made internally. People know AI matters. Leaders have a point of view. Employees are willing. The early efficiency gains are visible. The infrastructure is arriving on a scale that is genuinely difficult to absorb. What is missing is not will. What is missing is method.
Method is the quiet, slow, unglamorous work of building an organization that can decide which AI initiatives to start, which to stop, which to scale, and which to never attempt. Method is data ownership, governance rituals, ROI discipline, incident response, and the small everyday acts of asking where AI should not be used as carefully as the question of where it should. Method, in our reading of every benchmark we triangulated, is what separates the 6% of AI high performers from the rest of the market.
Texas is, in 2026, simultaneously two things at once. It is the substrate of the AI economy — the place where the compute, the chips, the cars, the rockets, the robots, and the energy are being produced. And it is a state full of organizations whose readiness to use that substrate, by their own self-assessment, is still uneven. The mismatch is not a crisis. It is an opportunity. The companies that close the readiness gap — that turn intent into method — will own a disproportionate share of the value that flows out of this state's AI buildout over the next decade.
The other thing worth saying, in closing, is about the 9,982 who did not respond to this survey. We did not, in the end, find them irresponsible. We found them honest. They did not have the institutional consensus to answer a 56-question maturity instrument, and they did not, in the moment we asked, pretend to. The first step in becoming a high-performing AI organization is, perhaps, the willingness to admit you are not yet one. The eighteen people who answered our survey did exactly that. The other 9,982 are, in their own quiet way, doing the same.
The future of AI in Texas will not be decided by the size of its data centers or the speed of its robotaxis. It will be decided by whether tens of thousands of organizations, in the next 24 months, build the operating habits that allow them to use the infrastructure being built around them. The answer to that is not in this report. It is in the rooms the readers of this report are about to walk into.
Who built this —
and what we do not claim.
The people, the principles, and the boundaries behind the State of AI Readiness in Texas.
The authors
Citation, version & review
- Recommended citation
- Donaleski, C., & Guerre, S. (2026). The State of AI Readiness in Texas — 2026 Edition. AI Advisory Group / AI Nexus Summit. Version 1.0.
- Version
- 1.0 · First edition · Published May 2026.
- Editorial review
- Internal editorial review by the AI Advisory Group team. Findings, framing, and interpretations are the authors' own and do not necessarily reflect the positions of sponsoring organizations.
- Next wave
- This is the first edition of a planned longitudinal program. The next wave is targeted for late 2026, with an expanded respondent base, deeper sector-specific batteries, and the addition of public-sector and municipal voices. Readers interested in participating in the next wave can contact the editorial team via AI Advisory Group.
- Use & distribution
- This report is distributed in confidence to executive participants of the AI Nexus Summit. Excerpts may be cited with attribution using the citation above. For redistribution or republication inquiries, contact the editorial team.
- Conflicts of interest
- The authors retain advisory and operating relationships with organizations referenced in the Texas industrial, energy, and AI ecosystem. No sponsor, advisor, or referenced organization had editorial control over the findings, framing, or recommendations in this report.
What this report does not claim.
- It does not claim statistical representativeness of the Texas economy or of any sector within it.
- It does not generalize from eighteen respondents to broad populations of executives, organizations, or industries.
- It does not infer causation from any score; all readings are observational and directional.
- It does not compare sectors quantitatively where the sample falls below five respondents — notably Healthcare, where n = 1.
- It does not equate intent metrics (what leaders say or expect) with delivered outcomes (what AI has actually produced).
- It does not assert sponsor, partner, or affiliated-organization positions; findings and framing are the authors' own.
- It does not treat any single number in this report as definitive in isolation from the triangulating external benchmarks cited throughout.
This report was
made possible by our sponsors.
Our gratitude to the organizations whose support made the State of AI Readiness research, fieldwork, and distribution possible.
Sources, Citations & Methodological Notes
All survey statistics are drawn from the State of AI Readiness Survey field instrument administered between mid-2025 and early 2026 by the AI Nexus Summit research team. The n=18 sample is a self-selecting subset of approximately 10,000+ Texas-connected professionals invited to participate. Statistics derived from this sample are directional and presented as such throughout. External benchmark statistics are sourced as follows:
- McKinsey & Company. "The State of AI in 2025: Agents, Innovation, and Transformation." McKinsey QuantumBlack, November 2025. Survey of 1,993 participants across 105 nations.
- Deloitte. "The State of AI in the Enterprise — 2026 AI Report: The Untapped Edge." Deloitte AI Institute, January 2026. Survey of 3,235 leaders across 24 countries, conducted August–September 2025.
- World Economic Forum. "The Future of Jobs Report 2025." January 2025. Survey of more than 1,000 employers representing 14 million workers across 22 industry clusters and 55 economies.
- U.S. Chamber of Commerce / Teneo Research. "Empowering Small Business: The Impact of Technology on U.S. Small Business 2025."
- U.S. Small Business Administration, Office of Advocacy. "AI In Business: Small Firms Closing In." Research Spotlight, September 2025. Based on Census Business Trends and Outlook Survey (BTOS).
- JPMorgan Chase Institute. "Understanding the Use of AI Among Small Businesses." April 2026.
- OpenAI / Oracle / SoftBank. Stargate Project announcement (January 21, 2025) and subsequent expansion announcement (September 23, 2025). Site location announcements: Abilene (TX), Shackelford County (TX), Milam County (TX), Doña Ana County (NM), Lordstown (OH), Wisconsin.
- Apple Inc. "Apple Will Spend More Than $500 Billion in the U.S. Over the Next Four Years." Press release, February 24, 2025. Expanded to $600 billion per Apple subsequent disclosure, October 2025.
- Tesla, Inc. Form 8-K filings, FY2025 and FY2026 Q1. Earnings call statements regarding Robotaxi (June 2025 Austin launch), Optimus production targets, and Gigafactory Texas expansion.
- Electric Reliability Council of Texas (ERCOT). 2025 Long-Term Load Forecast (LTLF). Aurora Energy Research, "Assessment of Resource Adequacy Needs in ERCOT," November 2025.
- Texas Medical Center. Statistics from TMC.edu Facts & Figures; "About TMC" institutional summary. Patient encounter and surgical volume data per institutional disclosures, 2024–2025.
- Houston Ship Channel / Greater Houston Partnership. Petrochemical capacity statistics, 2018–2025. Economy of Houston, Wikipedia compilation citing primary regional economic data.
- U.S. Energy Information Administration (EIA). Refinery Capacity Report 2025 (January 1, 2025 data) and "U.S. Refining Capacity Largely Unchanged as of January 2025," June 30, 2025.
- Veritone, PwC. Veritone Q1 2025 Labor Market Analysis; PwC AI Jobs Barometer 2025.
- SpaceX / Cameron County, Texas. Starbase, TX incorporation vote (May 3, 2025), formal incorporation (May 20, 2025), beach jurisdiction agreement (September 24, 2025).
- Crusoe Energy Systems. "Crusoe Announces Flagship Abilene Data Center is Live," September 30, 2025. Stargate campus financing and construction disclosures.
Editorial note: This report distinguishes between original survey findings (presented with explicit n=, scoring scale, and limitations), externally sourced statistics (cited above), and editorial interpretation (which is identified as such throughout). All readiness scores are presented on a 6-point semantic scale defined within the survey instrument; we have not converted them to percentages or implied statistical significance beyond the sample.
The survey instrument,
in plain sight.
A representative selection of the items in the State of AI Readiness Survey, organized by the four cross-cutting dimensions and the three sector-specific batteries. The scoring rubric appears below.
The six-point readiness scale
Every item in the instrument used the same semantic scale. Respondents were asked to rate each statement as it applies to their organization today.
The items
Dimension 1 · Leadership & Decision Discipline
- Our leadership has a clear point of view on why AI matters to this organization.
- Our leaders critically evaluate AI-generated recommendations rather than accepting them at face value.
- We have defined where human judgment is required when AI outputs are uncertain.
- Decisions about AI are made deliberately, not reactively.
- Our leaders understand where AI should not be used in our organization.
- AI initiatives are tied to specific business outcomes.
- Our organization avoids defaulting to AI for the sake of speed alone.
- Leadership alignment on AI is consistent across functions.
- We use structured decision frameworks when adopting new AI tools.
Dimension 2 · Data Readiness, Governance & Risk
- We understand where gaps in our data limit AI effectiveness.
- AI outputs in our organization are monitored for unintended consequences.
- We understand the regulatory and legal risks associated with our use of AI.
- We know how we would respond if an AI system caused harm or failure.
- Teams can access the data they need to use AI without excessive friction.
- We have a clear process for approving and deploying new AI tools.
- The data used for AI in our organization is accurate and reliable.
- Data ownership is clearly defined across our organization.
Dimension 3 · People, Adoption & Value
- Our employees feel safe raising concerns about how AI is being used.
- AI is improving efficiency, quality, or speed in at least one area today.
- Our employees are open to learning AI-enabled tools.
- We can clearly articulate where AI will create value for our organization.
- AI initiatives are reviewed based on outcomes, not novelty.
- Resistance to AI adoption is understood and actively managed.
- Our teams adapt effectively when AI changes workflows or roles.
Dimension 4 · Financial Expectations & ROI Discipline
- AI investments are expected to produce measurable business return.
- We can clearly explain how AI is expected to create business value.
- Expected return is discussed before AI tools are approved.
- AI initiatives have an understood time-to-value or payback horizon.
- Leadership is aligned on what financial success looks like for AI investments.
- Actual outcomes of AI initiatives are reviewed against expected value.
- AI initiatives are adjusted or stopped when expected value is not realized.
- We intentionally distinguish between experimental AI efforts and ROI-driven investments.
Sector Battery · Energy
- AI is not used to automate decisions that could impact physical safety without defined human oversight.
- AI initiatives are expected to measurably improve uptime, throughput, cost efficiency, or risk exposure.
- Our AI strategy incorporates physical, operational, and enterprise data.
- We know how to override or shut down AI systems if they behave unexpectedly.
- Operators and field teams trust AI outputs enough to use them in daily decisions.
- AI investments compete for capital using the same rigor as other operational projects.
- AI initiatives have an understood deployment timeline and time-to-value.
- AI solutions are designed to function reliably in constrained or low-connectivity environments.
Sector Battery · SMB
- We have sufficient time and capacity to pursue AI initiatives we believe are valuable.
- AI tools are adopted as part of a defined strategy rather than opportunistically.
- We can prioritize AI alongside our day-to-day operations.
- AI adoption is distributed beyond the owner or a small leadership group.
- We know how to measure business value or ROI from AI in our organization.
- We understand where AI tools could increase our dependence on specific vendors.
- AI investments are expected to buy back owner or leadership time, not just reduce costs.
Sector Battery · Healthcare
The Healthcare-Specific battery — covering clinical decision support boundaries, bias evaluation cadence, PHI governance, override and escalation pathways, cognitive load assessment, and adverse-outcome response — was issued as part of the field instrument but received only one valid response. Section 11 of this report discusses the implications of that silence. The full Healthcare battery is available on request from the editorial team.
Want to participate in the next wave —
or talk about your own readiness?
Whether you're interested in joining the next survey wave, contributing as a respondent or partner, or exploring an advisory conversation about your organization's AI readiness, the editorial team would welcome the dialogue.
Inquire with AI Advisory Group → aiadvisorygroup.com/contact