Original Research · Q1 2026

We surveyed 139 CEOs on AI ROI in 2026.
Here's what we found.

More spending, more projects, less hype. And most still can't tell you whether any of it is working.

139 CEOs surveyed
86% Not measuring ROI
54% Had a failed AI project
14% Formally track returns
13 industries
$5M – $500M+ revenue
Jan – Mar 2026
Compared to Q2 2025 baseline (n=163)

Why we ran this

We work with mid-market CEOs every day on AI implementation, and the same questions keep coming up: what are other companies my size actually doing? Are they seeing results? What's going wrong? We ran this benchmark to answer those questions with real data instead of vendor marketing. This is the second year of the survey, which means we can now track how the landscape is shifting, not just where it stands.

What you'll get out of this

This report covers where 139 CEOs are on AI maturity, how much they're spending, whether they're measuring ROI (spoiler: most aren't), what's causing projects to fail, and where the gap is between where AI gets deployed and where it would actually save the most money. We've included year-over-year comparisons against our 2025 benchmark, sector-level breakdowns, cross-tab analysis that surfaces patterns the top-line numbers miss, and direct quotes from respondents throughout.

tl;dr

AI adoption has jumped significantly since 2025, but 86% of CEOs still aren't formally measuring whether their investments are paying off. The companies seeing real returns ($500K+) all share one trait: they've moved AI out of the chatbot phase and hard-coded it into specific operational workflows. More than half of respondents have had a failed AI project, and the reasons are consistently about planning (scope, data readiness, unclear metrics), not technology. The biggest untapped opportunity is in back-office operations like order processing and reporting, where labor costs are highest but AI deployment is lowest.

01

What the Data Actually Says

Before the full breakdown: eight findings that cut against common assumptions.

1

The companies making the most progress are also the ones taking the most bruises.

Companies with a dedicated AI hire have a 68% project failure rate, compared to 49% for CEO-led teams. They're also far more likely to have reached workflow-level maturity: 58% vs. 36%. Failure is correlating with implementation, not incompetence.

2

For many companies, AI licenses are becoming the new shelfware.

48 respondents have deployed enterprise tools like Copilot or ChatGPT Teams, and of that group just 2% formally track ROI while half aren't tracking outcomes at all. Only 53% have done enough real implementation to even encounter a failure. This spend is sitting in a dead zone between curiosity and operational change.

3

Budget buys learning faster than it buys ROI.

Among sub-$10K spenders, 37% have had a failure and only 2% formally track ROI. At $250K-$1M those numbers flip: 73% failure, 40% tracking. Higher spend doesn't translate into cleaner outcomes, it buys more at-bats and, eventually, the discipline to measure what's working.

4

The companies measuring ROI aren't avoiding failure. They're measuring through it.

65% of formal ROI trackers have had a failed AI project, while among companies with no tracking at all only 35% report failure. That doesn't mean trackers are worse at this, it means the non-trackers are often too early or too shallow to see failure clearly. Measurement is a sign of maturity, not cleanliness.

5

You can't prompt your way to a half-million dollars.

Of the 10 companies reporting $500K+ in verified AI ROI, every single one has moved past the "exploring" and "licensed" phases into workflow-specific or fully autonomous AI. Zero are still in the ChatGPT-as-a-chatbot stage. The money shows up when AI is hard-coded into operational workflows, not when it's available as an optional tool.

6

Companies are applying AI where it's easy, not where the labor pain actually is.

68% of organizations use AI for content creation, yet the costliest repetitive labor is order processing and data entry. Among companies that named order processing as their biggest drain, only 25% actually use AI in operations, and among those citing reporting only 32% use AI for financial analysis. Companies are automating the easy stuff, not the expensive stuff.

7

Your AI problem is actually a plumbing problem. And it doesn't go away as you mature.

"Data isn't ready / systems aren't integrated" is the #1 blocker at the licensed stage (35%), and it's still the #1 blocker at the workflow stage (37%). "Lack of internal talent" stays stuck near the top at both levels. Companies that haven't started think their problem is knowing where to begin (75%). Companies that are further along know the real constraint is their operating infrastructure.

8

Scale is not showing up as an AI execution advantage.

Companies under $5M in revenue are at 54% workflow-level maturity, while companies above $500M sit at 33%, with the middle market clustering around 38-40%. The top-end sample is small, so treat this directionally. The trend still holds: bigger companies are not executing their way past the same problems everyone else has.

The story this year's data is telling: Enterprise AI has moved past the experimentation problem and into the execution problem.

02

The Year-Over-Year Shift

In mid-2025, we surveyed 163 CEOs on AI readiness (see the full 2025 report). The top three blockers were all about knowing what to do: distinguishing hype from reality (45%), building a roadmap (43%), picking a first ROI project (42%). Seventy-two percent were still researching or piloting.

Twelve months later, the hype question is gone. Only 26% of the 2026 cohort is still in "exploring" or "not started" mode. The blockers didn't disappear, though, they just matured. The new #1 is "data isn't ready / systems aren't integrated" (35%), followed by "lack internal talent to execute" (27%).

What changed: The conversation moved from "is AI worth it?" to "our plumbing can't support it." That's progress, even if it doesn't feel like it.

2026 blockers at a glance

n = 139 respondents, single select

Data isn't ready / systems aren't integrated 35%
Lack internal talent to execute 27%
Can't identify which processes to automate 17%
Don't know where to start 11%
Don't trust the ROI projections 8%

These blockers barely change as companies mature. "Data isn't ready" is the top blocker at the licensed stage (35%) and at the workflow stage (37%), and "lack of talent" stays near the top at both levels. The bottleneck is not AI, it's plumbing and people.

"Too many tools, too many ways, unclear which road to take!"

Real Estate CEO, $25M–$100M
03

The Maturity Picture

Most organizations have moved past experimentation. 37% now have AI integrated into specific workflows. Another 35% have deployed enterprise licenses. Only 3% are fully autonomous.

n = 139 respondents

37%
Workflow-specific: AI is part of defined processes
35%
Licensed tools deployed but not embedded in workflows
23%
Exploring: employees experimenting on their own
3%
Autonomous: AI running end-to-end

The averages hide two patterns worth paying attention to.

Framework: Cogs & Grease

"Grease" is AI that makes existing work a little faster: writing emails, summarizing docs. "Cogs" replace entire workflows end-to-end, running without a human prompt. Most of the 35% in the licensed group are still Grease. The 37% in workflow-specific have started building Cogs. That's where ROI lives. (More on this framework)

AI maturity is splitting by sector

75% of Tech & Software respondents have reached workflow-level maturity or beyond, and Professional Services and Financial Services are at 55%. Real Estate/Construction, the largest cohort in this survey, sits at just 16%, with Manufacturing at 14%. Not tech-averse industries by nature, but becoming that by default.

% at workflow-specific or autonomous maturity. Sectors with n ≥ 5 shown.

Tech & Software 75%
Professional Services 55%
Financial Services 55%
Retail / Consumer Goods 33%
Real Estate / Construction 16%
Manufacturing 14%

Scale isn't the advantage you'd expect

Companies under $5M in revenue hit 54% workflow maturity, while those over $500M land at 33%, with the middle market clustering around 38-40%. Small sample at the top, so treat directionally, but bigger is not translating to faster.

"Last year was exploration. This year is adoption. It's go time. Biggest challenge will be getting mass buy in. Need some demonstrated wins on the scoreboard to get broad momentum."

Manufacturing CEO, $100M–$500M
04

The ROI Gap

86% of CEOs are not formally measuring the return on their AI investments. 46% rely on gut feel. 40% aren't tracking at all.

n = 139 respondents

14%
Formally tracking ROI
46%
Informal / anecdotal only
40%
Not tracking at all

In 2025, 62% of budget-holders reported unclear ROI, and a year later spending went up while measurement didn't follow.

The licensed-tools group is where this gets sharpest. Of the 48 companies that have deployed tools like Copilot or ChatGPT Teams, just 2% formally track ROI and half don't track outcomes at all. They've bought the tools, they just haven't done the work to know if the tools are worth it.

What the 14% who measure look like

They aren't a different species, but they do have a few things in common:

n = 20 respondents who formally track ROI metrics

80%
Have reached workflow-specific or autonomous maturity
65%
Have had at least one failed AI project
100%
Of $500K+ ROI reporters are at workflow or autonomous stage

That 100% is not a rounding artifact. All 10 companies reporting $500K+ in returns have reached workflow or autonomous maturity, and none are still in the chatbot stage.

"AI is delivering a demonstrable ROI. It's automating most of the analyst job and has geometrically increased our speed. I suspect our ROI is at least 20:1 at this point."

Financial Services CEO, $5M–$25M

"Can't find the ROI beyond time savings and thought generation help."

Real Estate CEO, <$5M
05

Where It's Going Wrong

54% of respondents have had at least one AI project that didn't deliver, and the failure rate climbs with spend: 37% for sub-$10K companies, 75% at $50K+. More money doesn't buy smoother outcomes: it buys more attempts.

Why projects failed

n = 75 respondents with at least one failure, multi-select

33
Scope was too ambitious
28
Data wasn't ready
28
No clear success metrics defined upfront
20
Project ran too long / lost momentum
18
Couldn't get internal adoption
15
Vendor overpromised

This list looks like every failed ERP rollout and CRM migration. The technology isn't the problem, the project management is.

Framework: Atomic Units

"Automate our sales process" is not a project. It's a wish. AI works best as discrete atomic units: one trigger, one task, one output. Email comes in, CRM gets updated. PO submitted, line items validated. Stack enough atomic units and you have a workflow. Start too big and you get a stalled project. (More on this framework)

Progress and failure rise together:

Dedicated AI Hire (n=19)
68%
have had a failed project
58%
have reached workflow maturity
CEO-Led (n=85)
49%
have had a failed project
36%
have reached workflow maturity

They fail more because they attempt more. And they end up further ahead because of it.

"Expected easy identification of solutions and execution but that hasn't happened. Also lack internal execution leads while the leadership at top is clear in terms of adopting AI."

Manufacturing CEO, $500M+

"We've come to find some of the simplest tasks which AI should 'knock out of the park' come up short, which then deters our employees from diving in further."

Manufacturing CEO, $25M–$100M
06

The Alignment Problem

Where AI is being deployed today:

n = 139 respondents, multi-select

Content creation / marketing 68%
Internal knowledge search 57%
Code / software development 50%
Sales prospecting / outreach 46%
Data entry / document processing 45%
Financial reporting / analysis 35%
Customer service / support 33%
HR / recruiting 28%
Operations / supply chain 25%

Now compare that to where CEOs say repetitive labor actually costs them the most:

n = 139 respondents, single select (excludes "none/unsure" and "other")

Order processing / fulfillment 17%
Data entry / document handling 17%
Proposal / quote generation 14%
Reporting / reconciliation 14%
Customer support tickets 9%
AP / invoice processing 6%

The gap is specific. Among companies that named order processing as their biggest labor drain, only 25% use AI in operations. Among those citing reporting, only 32% use AI for financial analysis.

Content creation is a fine starting point. Stay there, and you're optimizing for convenience, not margin. The highest-ROI processes (order processing, data entry, proposals, reporting) all require plugging AI into the systems your business runs on, which is harder, and exactly why most companies haven't done it.

Framework: The Linear Ceiling

If doubling your revenue means doubling your support team, ops team, or sales team, that's your Linear Ceiling. The function where growth is physically constrained by headcount is where AI capital goes first. That's where the return is highest and where your team already feels the need. (More on this framework)

"We are focused on internal processes that have direct and quantifiable cost savings through a reduction in employee time. If we stick to this, we are seeing a mix of direct cost savings, saved hours, increased employee engagement, and better overall results on quality."

Real Estate CEO, $25M–$100M
07

The Ownership and Spending Problem

61% of CEOs say they personally own AI initiatives at their company. Only 14% have a dedicated AI hire. 6% say nobody owns it.

n = 139 respondents, single select

CEO / leadership team 61%
Internal IT / engineering 14%
Dedicated AI / innovation hire 14%
No one (ad-hoc) 6%
External consultants / vendors 5%

CEO-led teams reach workflow maturity at 36%, while companies with a dedicated AI hire reach 58%. Having someone whose actual job is driving AI forward makes a measurable difference.

Spending is flat

69% spent less than $50K on AI in the past twelve months, barely changed from 2025.

n = 139 respondents, single select

$1M+ 3%
$250K – $1M 11%
$50K – $250K 17%
$10K – $50K 35%
Less than $10K 35%

One pattern worth noting: 40% of $250K-$1M spenders formally track ROI. Among sub-$10K spenders, 2% do. Budget doesn't guarantee results, but it seems to force the discipline of measurement.

Framework: Data Eligibility & Legibility

"Data isn't ready" stays the #1 blocker at every maturity level. Two questions to ask before building anything: Eligibility: can the AI access the system? Is there an API, a webhook, a connection? Legibility: can the AI read what's in there? Your major SaaS tools probably pass both tests. Your legacy ERP and the spreadsheet on someone's desktop? Probably not. Map your stack, score on both, start where everything is green. (More on this framework)

"AI adoption has been difficult with a small core team. Project implementation is difficult to manage due to a mix of internal resistance and lack of time and resources to drive process implementation."

Diversified CEO, $25M–$100M
08

In Their Own Words

Unedited responses from survey participants on how AI is working — or not working — inside their companies.

The Optimists

"It's blowing us away. Very rapidly becoming native on everything we do."

Tech CEO, $5M–$25M

"AI has saved us millions of dollars and also made us millions."

Tech CEO, $25M–$100M

"It is a game changer and I have challenged my teams to 4x their weekly production using AI."

Professional Services CEO, $25M–$100M

"It's freed up a lot of time for employees to focus on tasks that AI can't do."

Retail CEO, $25M–$100M

The Pragmatists

"Will be working on quick return, low cost initiatives, as opposed to big ticket items."

Professional Services CEO, $5M–$25M

"We are more experienced now with AI. We know what works and doesn't work. Our internal teams have learnt from failures."

Education CEO, $5M–$25M

"It hasn't failed, it's just a process of discovery and implementation."

Diversified CEO, $100M–$500M

"It is clearer what AI can and cannot do and that knowledge allows us to build systems that will deliver value."

Tech CEO, $5M–$25M

The Frustrated

"Too many tools, too many ways, unclear which road to take!"

Real Estate CEO, $25M–$100M

"Too much hallucination and context loss."

Tech CEO, $5M–$25M

"Salesforce requires you to spend far too much money to implement AI into their CRM, so this is holding us back."

Financial Services CEO, $25M–$100M

"The biggest impediment for adoption has been mind shift from older developers."

Tech CEO, $25M–$100M

The Aspirational

"A roll out of standardized AI Agents who perform tasks with humans in the loop. Also complete digitization of our value stream."

Real Estate CEO, $100M–$500M

"AI becomes 24/7 coach for frontline workers to get knowledge and know-how very fast."

Manufacturing CEO, $100M–$500M

"I want to see process specific agents deployed across workflows outside of finance and technology."

Professional Services CEO, $25M–$100M

"Strategies adopted by front line teams, not just living in leadership meetings."

Healthcare CEO, $100M–$500M
09

Who Responded

139 CEOs from our peer CEO community, surveyed January through March 2026, majority mid-market with $5M to $500M in revenue.

By industry

n = 139 respondents

Real Estate / Construction 23%
Tech & Software 17%
Professional Services 16%
Manufacturing 10%
Financial Services 8%
Retail / Consumer Goods 6%
Healthcare, Energy, Logistics, Education, Other 19%

By annual revenue

n = 139 respondents

$500M+ 4%
$100M – $500M 23%
$25M – $100M 31%
$5M – $25M 32%
Less than $5M 9%

Participated in this benchmark?

Look up your personalized Cog vs. Grease Score to see where you landed.

Were you a respondent?

If you participated in this benchmark, enter your email to retrieve your personalized Cog vs. Grease Score and see where you landed.

/ 100

Want help closing your AI ROI gap?

We work with mid-market CEOs to find the highest-ROI AI opportunities in their business, build the plan, and execute it.

Talk to Your AI Department