peekgenticBook a call

Analysis · AI in business

What mid-size companies are automating first in 2026

Four 2026 surveys, from JPMorgan, the Dallas Fed, the Census Bureau and RSM, on what mid-size companies are automating first, how deep it goes, what they're leaving alone and where they get stuck.

By Miguel GutierrezCo-founder, Peekgentic · Dallas
· 7 min read
Preview card: What mid-size companies are automating first in 2026.
Process work first. Most of it still shallow. Most of it stuck on data.Peekgentic

In brief

The surveys agree on the first target: routine process work. In JPMorgan's 2026 survey of 1,469 midsize business leaders, 62% said they use or plan to use AI for process automation, ahead of predictive analytics (44%) and market intelligence (42%). Most of it is still shallow. Among Texas manufacturers using AI in the Dallas Fed's May 2026 survey, only 15% said many employees use it regularly. Headcount is mostly being left alone: 60% of JPMorgan's respondents expect no headcount impact this year. And the projects that stall get stuck on the same two things, data quality and connecting to older systems, according to RSM's survey of 1,030 middle market firms.

  • Process automation leads: 62% of 1,469 midsize leaders use or plan to use AI for it (JPMorgan, 2026).
  • Most use is shallow: only 15% of Texas manufacturers using AI say many employees use it regularly (Dallas Fed, May 2026).
  • Data quality (34%) and legacy-system integration (28%) are top barriers; 51% of pilots fell short (RSM, 2026).
  • Headcount is mostly untouched: 60% expect no impact in 2026 (JPMorgan).
  • Self-reported surveys: read them as a map of where peers are starting, not proof they're ahead.

Four surveys, four different questions

Before the numbers, a warning. "AI use" means different things in different surveys, so the percentages don't line up. Here is who asked what.

The 2026 surveys in this article

SurveyWho answeredWhen
J.P. Morgan Business Leaders Outlook 20261,469 executives of midsize companies (40% CEOs or chairs, 29% CFOs)Nov. 4-25, 2025
Dallas Fed Texas Business Outlook Surveys, AI special questions313 Texas business executives, 74 of them manufacturersMay 12-20, 2026
U.S. Census Bureau, Business Trends and Outlook SurveyU.S. businesses of all sizesDec. 2025 to May 3, 2026
RSM Middle Market AI Survey 20261,030 managers at firms with $30 million to $10 billion in revenue (U.S.) that already use AIMarch 5-16, 2026

Keep the last row in mind. RSM only surveyed companies that already use AI, so its numbers describe the companies that have started, not the whole middle market.

What they're automating first: process work

JPMorgan asked midsize leaders which AI applications they use or plan to use. Process automation came first by a wide margin.

AI applications midsize businesses use or plan to use

Share of 1,469 midsize business leaders, J.P. Morgan Business Leaders Outlook 2026 (surveyed Nov. 2025).

  1. Process automation62%
  2. Predictive analytics44%
  3. Market intelligence42%
  4. Chatbots31%
  5. HR processes29%

Source: J.P. Morgan, 2026 Business Leaders Outlook.

Only 11% of JPMorgan's respondents said they don't intend to use AI applications at all. So the question for most mid-size companies isn't whether. It's which process goes first.

"Process automation" is broad. The survey doesn't break it down by department. But the Dallas Fed asked Texas firms to describe, in their own words, tasks that AI had replaced. One plastics and rubber products manufacturer wrote: "Order processing, accounts payable, engineering." That's one firm's answer, not a statistic. It does match where the routine typing lives in most offices: orders coming in, invoices coming in, and the emails and reports around them.

Advisers see the same order. In Cherry Bekaert's Middle Market CFO Survey report, as reported by CFO Brew, the firm's CFO advisory leader said it starts with core finance processes like order-to-cash and procure-to-pay, and moves to forecasting and reporting after that. The logic is simple: forecasts are only as good as the order and invoice data under them.

How deep it goes: mostly a few people, not the whole company

Many companies say they "use AI." Far fewer use it across the business. The Dallas Fed's May 2026 survey makes this clear for Texas.

  • Manufacturers: 56.8% said they currently use AI. Of those, 30% were still testing or piloting, 52.5% said a small share of employees used it regularly, and 15% said many employees did.
  • Service firms: 69.2% said they use AI. Of those, 25.2% were testing, 52.1% had a small share of regular users, and 22.7% had many.

Two years earlier, in April 2024, only 28.2% of the Texas manufacturers surveyed said they used AI. So adoption has doubled, but most of it is one or two people using a chat tool, not a process that runs differently.

The Census Bureau's numbers are lower because it asks every kind of business, including very small ones, whether it used AI in any business function. In its May 2026 report, 32% of firms with 100 to 249 employees said they used AI, and 37% of firms with 250 or more. Across all businesses, the rate hovered between 17% and 20%.

RSM found the same gap at larger middle market firms. Of its respondents, all of whom use AI, 86% said AI is integrated into their operations, but only 36% said it is fully embedded across core processes. Among the 129 manufacturers in its survey, 32% reported full integration.

Where the money is going

Larger mid-size firms are spending real money. In RSM's survey, 58% plan to invest $1 million or more in AI this fiscal year, and 84% expect to raise AI spending next year. Keep in mind that RSM's firms start at $30 million in revenue and go up to $10 billion, so a 100-person company's budget will look very different.

What they're buying: generative AI is the most common (73% of respondents), followed by language AI (64%) and prediction AI (61%). Among manufacturers, 56% said they use agentic AI, meaning software that takes actions, not only answers questions.

One detail worth noticing. Among RSM respondents who expect to spend more on AI, the budgets most likely to be cut to pay for it were analytics projects (43%), cybersecurity (41%) and outside consultants (40%). RSM itself calls the cybersecurity trade-off a misalignment, since security is one of the top reasons projects stall.

Where they're getting stuck

The surveys are consistent about what goes wrong, and it's rarely the AI model.

Top barriers to wider AI deployment

Share of RSM Middle Market AI Survey 2026 respondents (1,030 firms already using AI).

  1. Data quality34%
  2. Security and privacy30%
  3. Connecting to legacy systems28%
  4. Talent or skills gaps28%

Source: RSM Middle Market AI Survey 2026: U.S. and Canada (PDF).

About half (51%) of RSM respondents who ran AI pilots in the past two years called the results moderate or limited. Among them, the leading reasons were data quality (53%) and integration challenges (47%). Manufacturers ranked security and privacy first (37%), then data quality (32%) and legacy systems (27%).

Accounts payable shows the same pattern from a different angle. A survey commissioned by Ottimate, an AP software vendor, of 225 finance leaders in retail, hospitality and healthcare found that only 4% had fully automated AP from invoice to payment, even though 93% used some automation. And 48% said their tools deliver minimal or no cost savings. It's a vendor survey with an interest in the answer, so read it as a direction, not a precise number. The direction fits: a tool that reads invoices but still leaves people retyping, chasing approvals and fixing mismatches doesn't save much.

Most AI projects that stall don't fail on the AI. They fail on messy data and systems that don't talk.

What they're leaving alone: headcount

For all the talk about AI replacing jobs, mid-size companies mostly aren't planning cuts. In JPMorgan's survey, 60% of midsize leaders said AI will neither increase nor decrease headcount in 2026, and 14% expected a decrease. Among Texas firms already using AI, 76.4% told the Dallas Fed it has not changed their need for workers so far, and 10.4% said it lowered it slightly or significantly.

The outlook shifts a little further out. Just under 30% of Texas AI users expect AI to reduce their need for workers over the next few years, and about 20% expect it to change the kind of workers they need. In practice, that looks more like slower hiring as volume grows than layoffs. We wrote about that choice in measure AI in hours, not headcount.

What this means for a 40-400 person company

If you run a manufacturer, distributor or service company in that range, the surveys point to a few plain lessons.

  1. Start with process work, where your peers are starting. The work that moves orders, invoices and status updates between email, PDFs and your ERP is where "process automation" pays first. Our list of office processes CEOs never see is a good place to look for candidates.
  2. Go deep on one process instead of wide on chat tools. The Dallas Fed numbers show most firms stuck at "a small share of employees use it." One process that runs end to end, from inbox to ERP, beats twenty people with a chatbot.
  3. Fix the data first. Data quality is the top barrier in RSM's survey and the top reason pilots fall short. Customer and item records, part number cross-references and vendor lists need to be right before automation can use them. See clean data before AI.
  4. Plan the connection to your ERP on day one. Integration with legacy systems is the other big reason projects stall. A tool that reads a document but can't write the result into Sage, Epicor, NetSuite or Business Central only moves the typing somewhere else.
  5. Don't cut security to pay for it. RSM found firms trimming cybersecurity budgets to fund AI, at the same time security was a top barrier. An AI tool that reads your mailboxes needs more attention to access, not less.

Where the surveys have limits

These are self-reported surveys. Executives tend to describe plans more generously than results. RSM's headline that 97% of respondents are satisfied with AI's business value comes from firms that already chose to adopt it, and RSM's own report cautions that many are judging success on isolated wins rather than company-wide change. JPMorgan's 62% is "use or plan to use," not confirmed deployment. The Dallas Fed's Texas sample is small, 74 manufacturers, so a few answers can move a percentage.

So don't read these numbers as proof your competitors are far ahead. Read them as a map. The companies that have started are mostly early, mostly on routine process work, and mostly held back by the same data and integration problems you'd face. That's a fair starting line.

Common questions

What are mid-size companies automating first with AI in 2026?

Process work. In JPMorgan's 2026 Business Leaders Outlook, 62% of midsize business leaders said they use or plan to use AI for process automation, ahead of predictive analytics (44%) and market intelligence (42%).

How many mid-size companies use AI?

It depends on the survey. The Census Bureau found 32% of firms with 100 to 249 employees and 37% of firms with 250 or more used AI in a business function (data period ending May 3, 2026). In the Dallas Fed's May 2026 survey, 56.8% of Texas manufacturers and 69.2% of Texas service firms said they use AI.

Is AI use at mid-size companies widespread or limited?

Mostly limited. Among Texas manufacturers using AI, the Dallas Fed found 30% still piloting, 52.5% with a small share of regular users, and 15% with many employees using it regularly.

Why do AI projects at mid-size companies stall?

Data quality and integration. In RSM's 2026 survey, 51% of firms that ran pilots called results moderate or limited, and the leading reasons were data quality (53%) and integration challenges (47%).

Are mid-size companies cutting jobs because of AI?

Mostly not yet. 60% of JPMorgan's midsize respondents expect AI to have no headcount impact in 2026, and 76.4% of Texas AI users told the Dallas Fed it hasn't changed their need for workers.

Where should a 100-person company start with automation?

With one routine process that moves data between email, documents and the ERP, such as order entry or invoice processing. Clean the underlying records first and plan the ERP connection from the start.

Sources

  1. J.P. Morgan, 2026 Business Leaders Outlook (midsize businesses, surveyed Nov. 4-25, 2025)
  2. Federal Reserve Bank of Dallas, Texas Business Outlook Surveys: special questions on AI, May 2026
  3. U.S. Census Bureau, AI use by businesses (America Counts, May 2026)
  4. RSM, Middle Market AI Survey 2026: U.S. and Canada (PDF)
  5. RSM, news release: The middle market has embraced AI. Now comes the hard part (2026)
  6. RSM, AI survey 2026 industry snapshot: manufacturing
  7. Ottimate, Only 4% of finance teams have fully automated AP (vendor survey, Feb. 25, 2026)
  8. CFO Brew, Midsize businesses have big AI plans this year (Jan. 14, 2026)

More from Peekgentic

Published November 19, 2026 · Written by Miguel Gutierrez, Co-Founder, Peekgentic (Dallas, TX). Every statistic links its source.