Release 03 ranks the 25 most urgent barriers stalling enterprise AI.
The top two, Runaway Buildout and Copilot Ceiling, trace to organizations investing in AI faster than they're set up to use it.
Risks range from unchecked AI decisions and overloaded employees to write-offs and walked-back claims.
NEW YORK, Nov. 18, 2026 /PRNewswire/ -- Andus Labs today published Release 03 of its Ground Truth Index, ranking the 25 most urgent barriers standing between enterprise AI investment and return. Runaway Buildout ranks first, with companies running dozens of unmeasured pilots that look productive and change nothing. Copilot Ceiling ranks second, as companies roll out AI assistants that get wide use for drafting and summarizing while the business runs the way it did before.
The previous Index found that seven of the top 10 barriers traced back to leadership decisions. This release shows what happens when companies move ahead without addressing them. Four of the top eight patterns fall under Institutional Coherence, which tracks whether a company's structure and spending support the changes it expects AI to deliver. All four are new to the Index, including the top two. The other two are Artifact Inflation (No. 6), where finished-looking AI work floods reviewers who can no longer tell what's ready from what's half-baked, and Swim Lane Sprint (No. 8), where multiple teams run their own AI projects without coordinating.
"Companies were sold AI on productivity. What many of them bought was risk," said Chris Perry, founder and CEO of Andus Labs. "Wall Street is debating whether AI is a bubble, but that debate is about twenty or so companies. Everyone else buying AI has a different problem. The technology is moving faster than they can rebuild the work around it, and that gap is where the risk builds."
The top-five highest-ranked patterns:
Runaway Buildout. Teams get a mandate and budget to experiment with AI before leaders agree on how to measure success. The pilots look productive, but until someone sorts out which ones work, real change stalls.
Copilot Ceiling. Companies roll out AI assistants and report near-universal adoption, but use stops at drafting and summarizing, because going further means rethinking how the business runs.
Empty Chairs. Companies give AI systems more decision-making authority without naming anyone responsible for overseeing and, when needed, overturning those decisions.
Learning While Drowning. Organizations add mandatory AI training to already full workloads without making room for employees to learn new ways of working.
Shadow Agency. Employees route client data and proprietary strategy through AI tools the company hasn't approved, and keep quiet because working faster invites scrutiny and rarely earns credit.
The Index also identifies the five forms of risk building inside enterprises: capital tied up in pilots with no defined goal, AI decisions without oversight, employees stretched thin, work duplicated across teams, and credibility lost when claims about AI have to be walked back. None of this arrives at once. It builds over quarters, showing up as write-offs, incidents, departures, and retractions.
Nine of the 25 patterns are new, and 16 return from prior releases. Leadership & Decision-Making remains the largest category, with eight of the 25 patterns, down from 12 in Release 02. Together, Leadership & Decision-Making and Institutional Coherence account for 14 of the 25, centering on two questions: who owns decisions about AI, and whether the company can follow through on them. Empty Chairs and Learning While Drowning, previously first and second, now rank third and fourth. Shadow Agency, Pilot Graveyard, Skill Atrophy, and Unchecked Output, have appeared in all three releases. (Note: Unchecked Output appeared as Fluency Divide in Release 01.)
"Every company I've seen that's getting AI right goes through a stretch where the work gets harder," said Mike Connery, Partner, Intelligence at Andus Labs. "It takes time and effort, and the people doing the work feel it first. They're reviewing and refining what AI produces, deciding who answers for its mistakes, and building new ways of working instead of automating old tasks."
Methodology
Andus Labs' Ground Truth Index draws on more than 2,100 conversations with business leaders at 700 companies, through client engagements, prospect meetings, and executive forums. From those conversations, Andus Labs has documented 332 recurring patterns. For each release, five independent AI analyst agents score every pattern against five dimensions set by the Andus Labs team. The agents cannot see each other's results, and each pattern's score is the median of the five. Andus Labs researchers then review the top 50, examining any pattern where the agents disagree. News coverage, financial filings, and regulatory action add context, but the final rankings rest on what the team hears in the field. The final 25 become the Index.
Read the latest Ground Truth Index at https://humanreadiness.org/ground-truth-index.
About Andus Labs
Andus Labs builds the Human OS for AI, the operating layer that orchestrates work across people and agents. The work spans operational redesign, decision simulations, and agentic products that change how enterprises run, with human judgment and accountability intact. Andus Labs delivers services-as-software to Fortune 500 clients and global IT leaders. The Ground Truth Index is the company's public diagnostic, drawn from a library of 332 barriers to enterprise AI observed in the field.
Media Contact
[email protected]
SOURCE Andus Labs

Share this article