The Texas middle market is adopting AI faster than most people think — and getting less out of it than almost anyone admits. This report puts numbers to both sides of that gap, looks at it through a Texas-sector lens, and explains what separates the operators who turn pilots into production from the ones who don’t.
It’s written for owners, CEOs, COOs, and CFOs at middle-market companies — founder-led, family-owned, and PE-backed — and for the operating partners pushing AI across their portfolios. Every statistic below carries a source and a year. We did not invent any of them.
Executive summary
- AI adoption in the small and mid-market has roughly doubled year over year. The U.S. Chamber of Commerce reported 58% of small businesses using generative AI in 2025, up from 40% in 2024 (U.S. Chamber of Commerce, 2025).
- The gap between mid-market and enterprise is closing — and is smaller than it has been for any prior technology wave. The U.S. Census Bureau’s most rigorous data shows firms with 100–249 employees adopting at ~32% versus ~37% for firms with 250+ (U.S. Census Bureau, May 2026).
- Adoption is not the problem. Conversion to results is. MIT found that about 95% of generative-AI pilots delivered no measurable P&L impact (MIT Project NANDA, 2025).
- Private equity is forcing the issue. EY found roughly two-thirds of PE clients had implemented at least one AI initiative across their portfolios by 2024 — but Alvarez & Marsal found only 8% of PE firms describe themselves as “leading” on AI value creation (A&M, 2026).
- Texas is where this plays out at scale: a ~$2.7 trillion economy, the 2nd-largest state economy in the U.S. and 8th-largest in the world if it were a country (BEA, 2024; Texas Comptroller, 2024).
How fast is the middle market actually adopting AI?
Two families of data tell the story, and they don’t agree — for a good reason.
Self-reported surveys ask owners whether they “use AI.” That sweeps in someone pasting a draft into ChatGPT. The U.S. Chamber of Commerce put small-business generative-AI use at 58% in 2025, up from 40% in 2024 — more than double its 2023 level (U.S. Chamber of Commerce, 2025). In the same survey, 96% of small business owners plan to adopt emerging technologies including AI.
Government transaction data is stricter. The U.S. Census Bureau’s Business Trends and Outlook Survey measures whether a firm actually used AI to produce goods or services. That number is lower and, when broken out by size, reveals the real shape of the adoption gap: firms with 250+ employees adopt at ~37%, firms with 100–249 employees at ~32%, and firms with fewer than 20 employees at under 20% (U.S. Census Bureau, May 2026).
Both are true. A firm can honestly say “yes” to the Chamber and “no” to the Census. What matters for a mid-market operator is the direction and the gap. The SBA’s reading of the Census data is the headline: small firms have closed the AI adoption gap with large enterprises to roughly a one-year lag — historically fast (U.S. SBA Office of Advocacy, 2025).
So the comfortable story — “we’re too small, we’ll wait for the big companies to figure it out” — no longer holds. The 100–249 employee band, the heart of the middle market, is already adopting at close to enterprise rates.
The harder truth: most projects never reach production
Adoption is the easy part. Value is the hard part, and the evidence here is brutal.
- ~95% of generative-AI pilots delivered no measurable P&L impact, according to MIT — “high adoption, low transformation” (MIT Project NANDA, The GenAI Divide: State of AI in Business 2025).
- At least 30% of generative-AI projects will be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls, escalating costs, and unclear business value (Gartner, July 2024).
- More than 80% of AI projects fail — roughly twice the failure rate of IT projects that don’t involve AI (RAND Corporation, 2024).
- 42% of companies expected to scrap most of their AI initiatives in 2025, up from 17% the year before — and organizations abandoned an average of 46% of AI proofs-of-concept before production (S&P Global Market Intelligence, March 2025).
- Looking forward, Gartner expects over 40% of agentic-AI projects to be canceled by the end of 2027 (Gartner, June 2025).
Read those together and a pattern appears. The failures cluster around a small set of root causes, not around model quality.
Why projects stall
RAND’s interviews with 65 senior AI practitioners ranked the causes, and the top one isn’t technical — it’s that stakeholders misunderstand or miscommunicate what problem AI is supposed to solve (RAND, 2024). After that come the foundation problems: not enough good data, infrastructure that can’t manage data or deploy models, and chasing the newest technology instead of the real bottleneck.
MIT named the same thing from a different angle: the issue is the learning gap — tools that don’t adapt to how the business actually works. MIT also found that more than half of generative-AI budgets went to sales and marketing while higher-ROI back-office automation stayed underfunded (MIT Project NANDA, 2025).
The mid-market evidence is consistent. RSM’s survey of 966 middle-market firms found 92% of those using generative AI hit challenges during rollout — led by data quality (41%), data privacy and security (39%), and insufficient internal skills (35%). More telling: 70% recognized they need outside support to get the most from AI, and only a third felt more than “somewhat prepared” to implement it (RSM US, 2025).
This is the core of it. AI doesn’t fail because the model is weak. It fails because the data isn’t ready, the systems aren’t connected, the problem was never sharply defined, and nobody owned the path to production. Precisely and Drexel found only 12% of organizations believe their data has sufficient quality and accessibility for effective AI, with 64% citing data quality as their top integrity challenge (Precisely / Drexel LeBow, 2024).
The private-equity push
If you’re PE-backed, the pressure is no longer optional.
- Roughly two-thirds of PE clients had implemented at least one AI initiative in their portfolio by 2024 (EY, 2024).
- 84% of PE funds expect AI to have a significant, transformative impact on their business (EY, 2024).
- 86% of corporate and PE firms have adopted generative AI in their M&A workflows — 65% within the prior year (Deloitte, 2025). AI now shows up in diligence (31% of firms have it somewhat or fully integrated — the highest-adoption stage), valuation, and deal sourcing (S&P Global Market Intelligence, 2026).
- Bain found that by late 2024, a majority of portfolio companies were testing generative AI and nearly 20% had operationalized use cases with concrete results (Bain & Company, Global Private Equity Report 2025).
But the same firms admit the value isn’t showing up yet. Alvarez & Marsal found 73% of PE investors expect AI to increase portfolio value over the next 12 months — but only 8% describe their firm as “leading,” meaning AI is actually moving EBITDA or the exit story (A&M, 2026). Accenture’s portfolio diagnostics put it bluntly: more than 60% of mid-market companies claim an AI strategy, fewer than 15% track its EBIT or revenue impact, and nearly 90% of pilots never advance past the pilot stage (Accenture, 2025).
That’s the “everyone’s trying, few are winning” gap — the same one MIT measured — playing out inside the portfolios that are supposed to be furthest ahead. The upside is real when it lands: Bain cites portfolio examples with 22% code-productivity gains and 40% content-cost reductions (Bain, 2025), and EY points to margin improvements above 10% from targeted AI automation (EY, 2024). The catch is that those are the cases that reached production.
A Texas-sector lens
Texas is the natural place to look at the middle market under stress. It’s a ~$2.7 trillion economy — the 2nd-largest in the U.S. and the 8th-largest in the world if it were a country (BEA, 2024; Texas Comptroller, 2024). Real GDP grew about 4.8% in 2024, roughly double the national rate (Texas Comptroller, 2024). It leads U.S. exports for the 23rd straight year, gained more residents than any other state in 2023–24, and has drawn 314 corporate HQ relocations since 2015 (Texas Economic Development Corp., 2024–25). The U.S. middle market overall is about 200,000 companies, a third of private-sector GDP, and roughly 48 million jobs (National Center for the Middle Market, Year-End 2024). Texas holds a heavy share of it. Here’s where AI has the clearest line to value, sector by sector.
Logistics and distribution
Texas international trade topped $1 trillion in 2024, with $547.9 billion (51.5%) crossing the Mexico land border and Laredo alone handling $339.7 billion (Texas Comptroller, 2024). Nearshoring is pushing more volume through these corridors. For logistics and supply-chain operators, the AI value is operational and unglamorous: demand forecasting, load and route optimization, dock scheduling, exception handling, and document automation across customs and freight. These are decision-automation problems on data the business already generates — exactly the back-office category MIT found underfunded and high-ROI. We built one carrier a freight-operations intelligence system on exactly this pattern.
Energy and oilfield services
Oil and gas directly employs 492,000+ Texans at an average wage 76% above the state private-sector average and supports about 1.4 million jobs in total (TXOGA, 2024). For mid-market oilfield- and energy-services operators, AI’s near-term wins are in predictive maintenance, scheduling and dispatch, anomaly detection on equipment telemetry, and bid/quote acceleration — turning sensor and operational data already being captured into fewer failures and faster cycle times. Our energy-optimization work with ESL is one example of pulling value out of telemetry the business was already capturing.
Manufacturing
Texas produces roughly 11% of U.S. manufactured goods and has the nation’s second-largest manufacturing workforce, with more than 1 million workers (Texas Comptroller; Texas Economic Development Corp.) — though the Dallas Fed’s manufacturing surveys note activity and employment softened through 2025. For mid-market manufacturers, the operational AI targets are quality inspection, yield and throughput optimization, demand and inventory forecasting, and quote-to-order automation. The foundation problem bites hardest here: legacy MES and ERP systems that don’t talk to each other.
Healthcare services
Texas hospitals employ more than 445,000 people — roughly one in twelve Texas jobs — and the sector was one of the few driving statewide job growth in 2025 (Texas Hospital Association; Dallas Fed). For mid-market healthcare-services providers and services firms, the practical AI value is in revenue-cycle automation, prior-authorization and claims workflows, documentation, and scheduling. We have shipped this kind of workflow in regulated clinical settings — for example, an equine practice-management platform. Data privacy and security — the second-most-cited barrier in RSM’s survey — is non-negotiable here, which makes human-in-the-loop design and governance a precondition, not an afterthought.
Financial and professional services
Dallas–Fort Worth is now the 2nd-largest U.S. hub for finance jobs, and DFW’s professional-and-business-services and financial-activities employment ran about 19% above its pre-pandemic peak in late 2024 (Federal Reserve Bank of Dallas). Professional, scientific, and technical-services employment in Texas reached roughly 1.06 million jobs in 2025 and continued setting records, with 55 straight months of annual growth through June 2025 (Texas Workforce Commission). These document- and judgment-heavy sectors — financial services and professional services — are where generative AI’s productivity evidence is strongest, but also where unmanaged accuracy and explainability risk shows up fastest. The path to ROI runs through narrow, well-governed workflows, not blanket deployment — the kind of governed data foundation behind our AML data-warehouse build.
Location-based entertainment and tourism
Texas travel and tourism set records in 2024: $97.5 billion in direct visitor spending, $199.5 billion in total economic impact, and 1.3 million jobs supported (Travel Texas, 2024). For location-based entertainment operators, AI value concentrates in demand and dynamic pricing, staffing and scheduling, and customer-service automation — operational levers tied directly to margin in a high-fixed-cost, variable-demand business.
What separates the operators who win
The data points in one direction. The winners aren’t the ones with the best models or the biggest budgets. They’re the ones who avoid the four failure modes that RAND, MIT, Gartner, and RSM all keep naming. In practice, that means four disciplines.
Start from the outcome, not the technology. RAND’s number-one cause of failure is misframing the problem (RAND, 2024). The operators who win pick a specific, measurable bottleneck — a quote cycle, a dispatch decision, a claims queue — and define what success looks like in dollars or hours before any model is built. MIT’s underfunded back-office automation is where this discipline pays. (Our AI assessment exists to find that first bottleneck before anyone writes code.)
Integrate; don’t replace. MIT’s “learning gap” is what kills most pilots: tools that don’t fit how the business actually runs (MIT, 2025). Nearly 60% of AI leaders cite legacy-system integration as a primary barrier (Deloitte, 2025). Winners build AI into existing systems and workflows rather than asking the organization to bend around a new tool.
Ship to production. The entire failure story — Gartner’s 30% abandoned after PoC, S&P’s 46% of proofs-of-concept scrapped, Accenture’s ~90% of mid-market pilots stuck — is a failure to cross the line from demo to daily use (Gartner, 2024; S&P Global, 2025; Accenture, 2025). The operators who win treat “in production, measured, owned” as the only definition of done. That’s the bar our value sprints are built to clear, and what running an ongoing AI Office is for.
Put real skin in the game. With 70% of middle-market firms saying they need outside help and only 8% of PE firms calling themselves AI leaders, the gap is execution, not ambition (RSM, 2025; A&M, 2026). The work that lands is structured around shared accountability for the outcome — not a deck, not a pilot that quietly dies, but a result someone is on the hook for.
This is the approach Frogslayer is built around — outcome-first, integrate-don’t-replace, ship-to-production, and shared risk — and it’s why our work shows a 96.5% success rate, 93 NPS, and over $1B in measured client ROI across 100+ companies. The point of this report isn’t the pitch. It’s that the evidence and the field results agree: in the Texas middle market in 2026, the constraint is no longer access to AI. It’s the discipline to turn it into a result.
Sources cited inline include the U.S. Census Bureau, U.S. SBA Office of Advocacy, U.S. Chamber of Commerce, MIT Project NANDA, Gartner, RAND Corporation, S&P Global Market Intelligence, Bain & Company, EY, Deloitte, Alvarez & Marsal, Accenture, RSM US, Precisely/Drexel, the U.S. Bureau of Economic Analysis, the Federal Reserve Bank of Dallas, the Texas Comptroller of Public Accounts, the Texas Economic Development Corp., TXOGA, the Texas Hospital Association, Travel Texas, and the National Center for the Middle Market. Figures reflect the most recent data available as of June 2026.