Research Report
AI at Work
Where AI is delivering, and where it's quietly falling short in 2026
by
Scroll
Where AI is delivering, and where it's quietly falling short in 2026
Introduction
The "AI at Work" report starts with a simple question: where is AI actually delivering value, and where is it falling short?
01
Adoption is accelerating, but most organizations are still working through the realities of implementation. Leadership expectations are rising, ownership is often unclear, and few teams have reliable ways to measure success.
02
At the same time, AI is delivering results, just not at the scale many expected. The most effective use cases are narrow, team-specific, and closely managed, while broader initiatives struggle to produce consistent outcomes.
03
Across every dimension in this research, one pattern stands out: the limiting factor is rarely the technology itself. It is the surrounding conditions, including data quality, governance, and how AI is introduced into the organization.
This report explores where those gaps exist, and what separates the organizations seeing results from those that are not.
Key Takeaways
of IT leaders report a gap between leadership expectations and what teams can realistically deliver. Organizations with strong alignment are 4.5x more likely to keep initiatives on track.
lack formal, outcome-based ROI metrics. Only 17% say most of their initiatives have delivered measurable results. Without measurement, underperforming projects linger in uncertainty.
of organizations skipped a full documentation audit. 32% say their knowledge base is less than 50% accurate. When inputs are unreliable, outputs will be too — no matter how advanced the model.
of organizations verify AI outputs before acting. Only 2% trust AI without human review. Combined with limited visibility and emerging security risks, trust in AI is still being earned.
How This Research Was Conducted
SaaS · Professional services · IT consulting · Manufacturing · Financial services · Healthcare · Cybersecurity · Government · Retail
Enterprise
1,000+ employees
Mid-Market
200–999 employees
Medium
50–199 employees
Analyses by company size is based on limited sample sizes. Percentage differences between groups should be interpreted with caution, as they are indicative rather than statistically conclusive. Readers are encouraged to focus on directional trends rather than precise percentage differences between groups.
The Reality Gap in AI
Expectations are moving faster than execution. 61% of IT leaders report a gap between what leadership expects and what teams can realistically deliver today.
How aligned is leadership's AI expectations with what your team can deliver today?
Moderate gap — leadership expects faster progress than we can deliver
43%Well aligned — leadership understands the potential and the limitations
26%Significant gap — leadership expectations are disconnected from operational reality
18%Leadership doesn't have specific AI expectations yet
13%The gap looks different by company size
Enterprise leads in perceived gap
65% report a moderate or significant gap, and only one respondent reported leadership has no specific expectations.
Mid-market shows similar misalignment
64% report a perceived gap, and 20% say expectations have not been set, leaving IT to define direction independently.
Medium companies show the lowest gap — highest ambiguity
50% report an expectations gap, but 27% say leadership has no AI expectations at all, meaning teams are adopting AI without a clear mandate or framework.
Key Finding — Alignment Predicts Outcomes
3X
Companies that are misaligned on AI expectations are 3X more likely to stop or scale back initiatives. 33% of initiatives get affected in misaligned orgs vs. only 12% in well-aligned ones.
Why Measuring AI ROI Remains a Challenge
73%
lack formal metrics
Most organizations still do not have a clear way to measure whether AI is working. They rely instead on adoption rates or subjective feedback rather than clear business impact.
How does your organization measure the ROI of AI initiatives?
We haven't established a measurement approach
32%Formal metrics — hours saved, tickets deflected, cost reduction
27%We track usage/adoption rates but not outcomes
20%Informal feedback ("it feels faster")
17%We tried to measure, but the results were inconclusive
4%The Measurement Gap by Company Size
Enterprise
18%
no measurement approach. Lead in formal metrics at 35% — but that still leaves nearly two-thirds without outcome-based measurement.
Mid-Market
40%
have no measurement approach — a significant share operating without visibility into whether AI investments are delivering.
Medium
50%
the most exposed — half have no measurement approach at all, making it nearly impossible to justify or improve AI investments.
Where AI Initiatives Are Falling Short
Despite early momentum, many AI initiatives are struggling to sustain results. More than a quarter of organizations have paused, scaled back, or fully abandoned at least one AI initiative in the past 12 months. Another 23% report initiatives that are underperforming and at risk.
In the past 12 months, has your organization paused, scaled back, or abandoned an AI initiative that was already underway?
All initiatives are progressing as planned
32%Initiatives not delivering, may be at risk
23%Haven't had formal initiatives long enough
18%Fully stopped at least one initiative
14%Significantly scaled back scope or budget
13%27%
have paused, scaled back, or fully abandoned at least one AI initiative in the past 12 months.
23%
report initiatives that are underperforming and at risk of being stopped.
The Role of Data Readiness
90%
skipped full audit
Most AI accuracy problems are not AI problems. They are documentation problems that existed long before anyone deployed a model.
90% of organizations deployed AI without extensively auditing the internal documentation that those tools would access.
Before deploying AI tools, did your organization audit or clean up the internal documentation AI tools would access?
No, we deployed AI on existing documentation as-is
45%Partially, we cleaned up some areas
38%Yes, extensively
10%We started, but didn't finish the process
7%The audit gap by company size
Medium companies are the most exposed. 62% deployed AI on existing documentation as-is, and only 4% audited extensively. Enterprise organizations fare somewhat better at 14% extensive audit rate, but 37% still deployed without any cleanup.
Enterprise
37%
deployed as-is
14%
audited fully
Mid-Market
48%
deployed as-is
8%
audited fully
Medium
62%
deployed as-is
4%
audited fully
What This Means
The Minority — On Track
Teams that started with governance, cleaned up their documentation, set expectations with leadership, and defined what success looks like before deploying anything. Their initiatives are on track, their failures are caught early, and their AI investments are producing measurable results.
The Majority — Working Backward
Teams that deployed first and are now trying to bolt on governance, clean up data, and justify ROI after the tools are already live. Not failing because of the wrong technology — failing because the organizational scaffolding was never built.
This is not a call to slow down. The pressure to adopt AI is real, and the organizations that figure it out will have a significant advantage. But the data is clear that figuring it out starts with the work nobody wants to do: auditing 20 years of Confluence pages, defining what a successful outcome actually looks like, and having an honest conversation with leadership about what is possible with current resources.
The full breakdown behind every stat in this report. Data readiness benchmarks, ROI measurement gaps, and what the top performers are doing differently.
We built a short AI assessment that scores you against the same benchmarks in the report: leadership alignment, ROI measurement, data readiness, and rollout maturity.
About Leebry
Leebry
Leebry is a Work AI platform designed to help organizations make better use of the knowledge and tools they already have. By connecting systems like Confluence, Slack, Jira, Okta, and HiBob, it provides secure, citation-backed answers and enables everyday workflows through a single interface.
Available Q2 2026. Learn more at leebry.com
Built by
MacPaw Products:
© 2026 MacPaw Inc. All rights reserved. Leebry is a trademark of MacPaw Inc.
AI AT WORK · RESEARCH REPORT · 2026
We won't spam. But we would like to send some valuable content we produce in the future.
By submitting the form, you agree to receive updates and marketing communications about Leebry launch. You acknowledge that your personal data will be processed by MacPaw in accordance with our Privacy Policy.
Thank you! The report will be in your inbox shortly.
Open now