The adoption and leverage of AI capabilities are advancing quickly across most organizations. It is an exciting time, a cautious time, and a challenging time, all wrapped up in the same AI journey. But many are finding that scaling requires a step back first, an honest assessment of organizational readiness. Adoption, it turns out, depends on a few conditions being in place to achieve the outcomes leaders expect.
Our AI adoption across the company was “on track.” Pilots were running in four departments. Developers were shipping faster. Marketing had a chatbot live. By every visible measure, things looked good.
But that isn’t the full story. Ask the broader question, though, and it forces a pause: an honest accounting of where the organization actually stands. If the board asked tomorrow whether the AI initiatives were governed, auditable, and tied to a number anyone would recognize as ROI, could you prove it in 90 days? Most leaders pause for a long moment before answering honestly: no.
That pause is the story of most AI adoption. It is not a story about whether organizations are using AI. Nearly all of them are. It is a story about the widening gap between activity and readiness, between deploying a tool and being organizationally prepared to scale it responsibly. IBM’s own diagnosis of the moment is direct: AI capability is advancing faster than organizational capability.
That single sentence describes more failed AI initiatives than any technical postmortem ever will.
Readiness, in this sense, is the organizational capacity to answer four questions honestly, at any moment, for any AI initiative running inside the company.
- Who owns this?
- What data is it built on?
- What does it cost, and what is it returning?
- Who is measuring it to know if something has gone wrong?
Organizations that can answer those questions have built governance deliberately rather than drifted into it. Those that cannot are accumulating a different kind of debt, one that does not show up on a balance sheet until a board asks that same question.
When readiness is missing, the gap does not stay abstract. It shows up as seven concrete, recurring problems: governance and oversight that cannot keep pace with deployment, data infrastructure that was never built for AI, cost that outpaces the value it was meant to create, individual productivity gains that never compound into organizational ROI, a workforce that leadership assumes is ready when it is not, a talent bench built for the wrong skills, and pilots that never leave the lab. These are symptoms of the same underlying condition: organizational readiness that has not caught up to organizational ambition.
Let’s unpack each of these, and what closing the gap requires.
When Readiness Is Missing: Seven Recurring Problems
Across recent industry research, the readiness gap shows up in the same seven places, regardless of sector or size.
Governance and oversight can’t keep pace.
EY found that over half of department-level AI initiatives are running without formal approval, and 78 percent of technology leaders say adoption is outpacing their organization’s ability to manage the risk it creates (1).
Data was never built for AI.
Teams routinely discover mid-project that the data they assumed they had is inaccessible, incomplete, or structured the wrong way. Gartner projects that 60 percent of AI projects will be abandoned through 2026 for exactly this reason (2).
Cost is outrunning visibility.
Rising token and tool spend has overtaken security and quality as the top adoption concern among engineering leaders, cited by 42 percent as their primary obstacle (3).
Individual gains aren’t compounding.
Ninety-seven percent of executives report their company has benefited from AI, yet only 29 percent see significant organizational ROI, a gap wide enough that 54 percent of C-suite leaders admit adoption is straining their organization (4).
Leadership overestimates workforce readiness.
CIOs and CTOs are five times more likely than COOs to believe their workforce is ready to adopt AI, and training remains the most underfunded investment area even as it stays disconnected from the workflows people actually use (5).
The talent bench is built for the wrong skills.
Organizations that spent the last two years hiring for prompt engineering now need people who can manage integration, security, and governance across live AI systems, a very different bench (6).
Pilots never leave the lab.
Across more than 120,000 enterprises surveyed, only 8.6 percent report AI agents running in production. The rest remain in pilot or have no formal initiative at all (7).
Each of the seven is what the same underlying condition looks like once it surfaces in a different department’s metrics.
Closing the Gap: What Readiness Requires
If the seven problems share a root cause, the fixes share root disciplines as well. None of them start with a new tool.
Governance as an operating model. The organizations closing the oversight gap, and the ones finally moving pilots into production, are treating governance as an organization design question rather than a policy memo: which decisions can be safely encapsulated inside an agent’s own boundary, and which must be orchestrated by a person, a team, or a process sitting above it. Get that decision boundary right once, and both the oversight problem and the pilot-purgatory problem start to resolve together, because a system with a clearly drawn boundary is also one a board can audit.
Business process as bounded context. Data readiness is rarely a data problem first. It is usually a business process mapping problem wearing a data costume. Data is a byproduct of process. If a business process is fuzzy, ad hoc, or exists differently in five people’s heads, the data it throws off will be fuzzy too, not because anyone failed at data management, but because there was never a clean process generating clean data in the first place. Most of “our data isn’t AI-ready” complaints are actually “we never wrote down how this part of the business works” complaints. Fix the second thing, and the first thing tends to resolve on its own. McKinsey’s research bears this out directly: across 25 factors tested against companies’ financial returns from AI, workflow redesign, rebuilding how the process runs, has the single largest measured effect on whether an organization sees real bottom-line impact (8).
Redefining skill and competency alignment. The workforce and talent gaps both live in the same uncomfortable space, the gray area between the role someone was hired to do, and the role AI has quietly started to reshape. Training that only teaches a tool leaves that gray area undefined. Training that redesigns the role around it, and rebuilds the talent bench alongside it, is what closes both gaps at once.
Scaling with ROI intact. Much of the ROI translation gap comes from business and technology working as separate silos, each with its own view of what “valuable” means. The strongest AI strategies close that gap by aligning business and technology on a shared definition of value; that means giving the business team that will use an AI workflow real ownership over it, not just sign-off rights, while IT holds the oversight layer that keeps it governed. Aligned this way, ROI stops being something IT claims and business doubts, or business demands, and IT can’t deliver.
Budgetary guardrails. Full cost visibility takes time to build. Until it exists, usage caps, tiered access, and budget guardrails function as compensating controls, guarding against runaway spend while the underlying metering catches up. The point is to make sure enthusiasm and accountability travel together.
Conclusion
These seven problems trace back to a systems issue rather than a tool or the new capability itself: a lack of clarity around decisioning (governance) and bounded context (structure). From the outset, our point of view has been direct: a well-constructed governance and operating model is a readiness condition, not just for adopting AI, but for any new capability an organization adds to create value for its customers.
“A well-constructed governance and operating model is a readiness condition for any new capability that creates value.”
The scale of the gap is why this matters more now. Separate research from MIT found that 95 percent of enterprise AI pilots deliver no measurable bottom-line impact, and that only the small share reaching production create real value (9). That number belongs alongside the seven problems above, further evidence of the same underlying condition: organizations that have not yet built the capacity to know what to pursue, where to focus, how to deliver, and how to capture the value once it exists.
“95 percent of enterprise AI pilots deliver no measurable bottom-line impact”
In practice, closing that gap tends to take a consistent shape:
- Value Aligned: A prioritized view of where the value lives
- Governance & Structure: Ownership and decisions sitting with named roles and business owners rather than left implicit
- Measure: A KPI framework that can defend the investment when the board asks, and a validated path that moves an idea from concept to proof before it is asked to scale.
Organizations that build that discipline once tend to reuse it. It’s how they evaluate any new capability meant to create value, whether or not AI is involved. Organizations that treat readiness this way, as an ongoing practice rather than a gate, are the ones building AI capability that holds up under scale.
LiminalArc CEO Mike Cottmeyer has described the alternative bluntly: most stalled AI programs are “a structure problem wearing AI clothes.” Fix the structure once, prove it in production, and the same pattern repeats faster with each cycle that follows.
This post comes from our management consulting practice, which specializes in designing and implementing operating models that align governance, processes, and technology to drive measurable business outcomes.
References
1. EY, “EY survey: autonomous AI adoption surges at tech companies as oversight falls behind,” 2026.
2. Gartner, 2025 Survey on Data Management Practices for AI, cited in Cygnet.One, “AI Adoption Challenges: Enterprise Guide 2026.”
3. Jellyfish, “State of Engineering Management,” 2026.
4. Habib, M. (WRITER), 2026 AI Adoption in the Enterprise survey, conducted with Workplace Intelligence.
5. Grant Thornton, 2026 AI Impact Survey.
6. IBM, “The Biggest AI Adoption Challenges for 2026.”
7. Recon Analytics survey (120,000+ enterprise respondents, March 2025-January 2026), cited in TechRepublic, “AI Adoption Trends in the Enterprise 2026.”
8. McKinsey & Company, “The State of AI: Global Survey,” 2025. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
9. MIT NANDA, “The GenAI Divide: State of AI in Business,” 2025.
