Your AI Roadmap Is Really A Planning Problem
Monisha Somji | Senior Digital Transformation & AI Executive.
gettyWalk into almost any large enterprise today, and you will find AI everywhere: copilots in the inbox, pilots in every function, a steering committee or two. Yet the financial results rarely match the activity.
In my experience, the pattern is consistent. The demo works, the pilot succeeds and the initiative stalls anyway because nobody owns the step between a working prototype and a changed way of working.
The data agrees. McKinseyβs latest global survey found that 88% of organizations now use AI in at least one business function, but only 39% report any enterprise-level EBIT impact, and just 6% qualify as βhigh performers.β Gartner predicts that βover 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.β
The usual explanation is that the technology is not ready or that employees need more training. Having led digital transformation and product programs across technology, retail and mobility, I see a different culprit: the planning and funding machinery wrapped around the model.
Across industries, leaders often avoid saying out loud that this is a leadership problem, not a technology problem. Pilots are popular because they are politically safe. They let every executive show AI progress without consolidating a team, giving up a budget or killing a sponsored project. Redesigning a workflow forces decisions about who owns it, who loses control of it and what stops.
Those decisions are hard because incentives point the other way. Leaders are rewarded for launching initiatives, not retiring them. Budgets belong to functions, not workflows, so no one gains by consolidating. Risk teams are measured on avoiding incidents, so they review late and cautiously. Everyone behaves rationally inside a system that quietly rewards the wrong thing.
Most large organizations still run on annual budgets, quarterly planning and intake queues designed for software that ships a few times a year. AI capabilities shift in months. A use case approved in one cycle can reach production two cycles later on a model already a generation behind.
That is why McKinseyβs research keeps landing in the same place. Of the 25 organizational attributes it tested against EBIT impact, workflow redesign had the strongest effect. The companies capturing value are not adding a tool to an old process. They are rebuilding the process.β
Every function runs its own AI queue, with its own vendors and its own definition of success. It is common to find three teams funding near-identical summarization tools from three vendors, each unaware of the others, until a budget review puts the requests side by side.
A pilot gets a budget and a deadline. When the money ends, the initiative ends, whether or not it worked. Pilots routinely hit their targets and still die because the funding line closes the same quarter the results arrive and no one owns the workflow they were meant to change.
Risk and privacy reviews arrive after the build, as a gate. A team builds for months, meets risk review for the first time at launch and a two-week review becomes a two-month rework. According to a Gartner analyst,β many current agentic efforts βare mostly driven by hype and are often misapplied,β which is what happens when use cases are chosen before anyone tests them against value and risk.
Consolidate the front door. Create one intake and prioritization path for AI work, shared by product, technology and the business. One view of demand makes it possible to sequence work, reuse components and kill duplicates early.β
The next phase of enterprise AI will not be won by whoever licenses the most capable model, because those models will be available to everyone. It will be won by organizations that decide quickly, fund the right workflows, govern as they build and change how work actually gets done.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

