Texas Executives Using AI Say It Hasn't Cut Their Headcount - Yet

Ask 203 Texas executives who _actually use AI_ in their companies what it has done to their staffing, and 76% give the same answer: nothing. Not "not yet," not "we're evaluating" — no impact on the number of workers they need.

That is the most under-reported number in the entire AI-and-jobs debate right now. But the same survey contains a second number that tells you exactly why the anxiety hasn't gone anywhere — and why it might be pointed at the wrong thing.

The number that contradicts the headlines

Between May 12 and May 20, 2026, the Federal Reserve Bank of Dallas asked 313 Texas business executives — 74 manufacturers and 239 service-sector firms — a set of supplemental questions about artificial intelligence. The 203 respondents who said their firm currently uses AI were then asked, plainly: on net, how has AI affected employment at your firm?

The distribution is not what the headlines would lead you to expect. 76.4% said AI has not impacted their need for workers at all [Fact]. 8.9% said it slightly decreased that need, and 1.5% said it significantly decreased it — a combined 10.4% reporting any reduction at all [Fact]. Another 4.4% said AI changed the _type_ of workers they need but not the number, and about 1% said their need for workers went up. The remaining 7.9% simply didn't know.

Ten percent. That is the realized, on-the-books, executives-signing-the-payroll figure for AI-driven headcount reduction across a 313-firm sample in one of the largest state economies in the United States, in mid-2026.

AI adoption has quietly hit a ceiling

Here's the context that makes that number credible rather than merely comforting. 66.2% of responding Texas firms say they currently use AI — up sharply from 38.3% in April 2024 and 59.1% in May 2025, but essentially flat against the 66.0% recorded in December 2025 [Fact]. Adoption didn't stall because AI stopped working. It stalled because the firms that were going to say yes have mostly already said yes.

Service-sector firms are further along (69.2% using AI) than manufacturers (56.8%) [Fact]. But the more revealing question is _depth_, not breadth. Among AI-using firms, only 21.2% say the tools are used regularly by many employees. A majority — 52.2% — say AI is used regularly by only a small share of staff, and 26.1% are still testing or piloting without regular use [Fact].

Read those two facts together and the 76.4% stops being a mystery. Most Texas firms are at the "a handful of people use it a lot" stage. That is precisely the stage at which headcount does not move [Estimate]. Among the small group where many employees use AI regularly (43 firms), 62.8% say the tools are already integrated into core business processes — so the deeper stage exists. There just isn't much of it yet.

The gap between the ledger and the forecast

Now the second number. Asked how they expect AI to affect employment over the next few years, the same executives shift hard. 25.7% of all respondents (current users plus firms planning to adopt within 12 months) expect AI to decrease their need for workers — 21.8% slightly, 3.9% significantly [Fact]. Among firms _already using_ AI, that expected-decrease share rises to 29.3% [Fact].

So: 10.4% realized reduction, 25.7% expected reduction. A re-processing of the Dallas Fed's own figures puts the forecast at roughly 2.5× the reality [Estimate]. The fear isn't in the ledger. It's in the forecast.

And here's the twist that should make you sit up: the _less_ experience a firm has with AI, the _less_ disruption it predicts. Among firms that haven't adopted yet but plan to within a year, 65.5% expect no impact on staffing and 0.0% expect a significant decrease [Fact]. The executives who use AI most intensively are the ones bracing hardest. Experience is not producing relief — it's producing forecasts.

One more line worth holding onto: 19.5% expect AI to change the _type_ of workers they need without changing the number [Fact]. That's roughly one in five firms planning a reskilling problem, not a layoff.

Productivity is up. Pay is not. That's a 10-to-1 problem.

This is the finding that deserves to be on every worker's radar. 71.4% of AI-using firms report that AI has raised productivity for employees who use it, relative to similar employees who don't [Fact]. Asked whether AI use has led to _higher wage growth_ for those same employees: only 7.4% said yes. 76.5% said flatly no [Fact].

Roughly a 10-to-1 ratio between productivity gains reported and wage gains transmitted [Estimate]. In manufacturing the spread is even wider — 70.0% report higher productivity, 2.5% report higher wage growth [Fact].

If you use AI at work, this is your number. Your employer is very likely getting measurably more output from you, and very likely not paying you for it. That is not a technology problem. It's a negotiation problem — and negotiation problems are the kind you can actually do something about.

Which tasks are actually going away

The survey's write-in responses are the closest thing we have to a task-level X-ray, and they show something important: firms are not replacing occupations. They're replacing the formulaic layer _inside_ occupations.

A securities and investment firm listed "graphics designers, coders, mid-level managers — processes that are redundant or repetitive, reminders, outreach tasks." A professional-services firm listed "lower-level contract review, summarizing lengthy but somewhat formulaic documents." A plastics manufacturer listed "order processing, accounts payable, engineering." An auto dealer cited marketing, planning, and automated customer follow-up. A restaurant group listed exactly one thing: "accounts payable entry."

That pattern maps cleanly onto occupations we track. See paralegals, graphic designers, bookkeeping clerks, order clerks, customer service representatives, and software developers for the underlying exposure and automation-risk data.

And the new work? A metal manufacturer reported AI creating demand for financial summary and management-discussion writeups. A professional-services firm reported "the need for software support people" _increasing_ because AI drove more software sales. Small, but real.

The warning, and the counterpoint

One respondent — a professional-services executive — offered the survey's bluntest prediction: "AI will significantly decrease demand for first-year to mid-career associates in large law firms fairly quickly (within two years)" [Claim]. That is one executive's judgment, not a forecasting model, and it should be read as such. But it is a specific, falsifiable, time-bounded claim from someone with skin in the game, which is more than most AI predictions offer.

The counterpoint came from a real estate executive: "If anything, AI is solidifying the need for talented, highly skilled white-collar professionals who know how to develop, mentor and maintain strong working relationships with colleagues and clients" [Claim]. And from a small professional-services firm that is _hiring because of_ AI: "AI is benefiting smaller and medium companies like us more... we can hire more people in coming months, as our sales and profits are increasing" [Claim].

What this actually means for your job

Four things you can act on this quarter.

Get into the 71.4%. The productivity gap between AI-using and non-AI-using employees is now something executives measure and report. Being on the wrong side of it is a visible position.

Convert productivity into evidence. With a 10-to-1 gap between reported productivity gains and reported wage gains, nobody is going to hand you the raise. Document what you shipped, how fast, and what it replaced. That is the only currency in this conversation.

Audit the formulaic layer of your own job. Not "will my job be automated" — that framing is failing the data. Ask instead: which parts of my week are formulaic document work, repetitive processing, or routine follow-up? Those are the parts under pressure. What's left is the part worth deepening.

Watch the 19.5%. One in five firms expects to change the _type_ of workers they need, not the number. That is a reskilling window, and it opens before the layoff window does.

The honest read of this survey isn't that AI is harmless. It's that the disruption is currently running about 2.5× hotter in executives' expectations than in their actual payrolls — and that gap is where your leverage lives. The people who use the next two years to move up the value chain will not be the ones the forecast was about.

Limitations worth stating: this is a self-reported survey of 313 firms in the Eleventh Federal Reserve District (Texas and parts of New Mexico and Louisiana). Several questions have subsamples under 50 responses. It is a strong signal, not a national census.

Sources

• Federal Reserve Bank of Dallas, "Texas Business Outlook Surveys — Special Questions," May 26, 2026: https://www.dallasfed.org/research/surveys/tbos/2026/2605q
• Federal Reserve Bank of Dallas, Special Questions underlying data file (XLSX), May 2026: https://www.dallasfed.org/-/media/Documents/research/surveys/tbos/2026/2605sq-data.xlsx

_AI-assisted analysis. Survey figures were read directly from the Dallas Fed's published results page and its underlying data file; interpretation, ratios and derived comparisons are our own. Occupation exposure and automation-risk figures come from our own dataset covering 1,016 O\*NET occupations._