Companies are on track to spend $2.7 trillion on AI this year, and most of them can't yet tell you what it bought. The tools are real and the budgets are approved, but the return keeps arriving late, if it arrives at all. Some companies have closed that gap and pulled ahead; most are still standing at the edge of it, looking for a first move. And the first move, more often than not, is the same: they convene.
A steering committee forms, a center of excellence gets a charter, an AI council starts meeting on a standing invite. The apparatus goes up quickly, and it reads as momentum. Yet a company can stand up every one of those structures and still watch the return stay flat.
Philip Samuelraj has watched that gap from inside a lot of engagements. He's the Founder and CEO of Techjays, the firm he founded in 2020 to redesign how companies operate on AI end-to-end, and before that a Program Manager at Google working on generative AI for Google Support. He's pointed about the distance between running a transformation and staging one.
"The pace of the committee becomes the pace of their laggards," Samuelraj says. As members work to catch up, the field keeps moving. "By the time they've even caught up, they're already behind."
Motion, or progress?
Samuelraj is careful to say the structures themselves aren't the problem. Security matters, compliance matters, and a company moving fast on AI needs rails. "Those are guardrails," he says. "But what problems are you solving, and how are you truly taking the company to succeed in the agentic era?" The trouble starts when the guardrails become the activity, when the meeting about AI stands in for the work of it. He's watched plenty of companies reach for a committee as their first move and mistake the convening for the change. "A lot of them are using corporate frameworks. 'All right, we'll start an AI committee.' There's this AI governance theater. 'We need to be secure, we need to be compliant,' and instead of solving real problems, that becomes the work."
For Samuelraj, the test is what changes in the work itself. He asks leaders to identify tasks agents now handle, workflows they have eliminated, and improvements they can measure.
Three questions that cut through
His diagnostic consists of three questions. First, what were we doing by hand a quarter ago that's now handled fully by agents, and can you name it? Second, which workflows have we retired because they no longer make sense in an agentic setup? Third, are we getting more done, at better quality, in a fraction of the time, and can you measure it? "Can you really name things," he asks, "and what is the impact that you had?"
That does not mean people disappear from the workflow. Samuelraj also describes spending more time on verification and applying human judgment to the output. The question is which work agents can take over, and where people can make the result better.
He points to one client whose operational team, he says, could handle 5.5 times its previous workload after Techjays redesigned the operation. The company did not plan to lay people off because of the change. Instead, it intended to use the additional capacity to support rapid growth, leave some vacancies unfilled through natural attrition, and move people into other departments, including sales, where they could help generate revenue.
The questions are unforgiving by design. Each one asks for a named thing rather than a status update, and a number rather than a sentiment. That's a higher bar than most oversight clears. Boards are approving the AI spend without setting expectations for what it should return, which is how a program can look governed and still go unmeasured.
Map the spend to an outcome
Samuelraj starts by plotting potential AI projects on a matrix of cost and expected return, then prioritizing those with lower costs and higher returns. "We tend to target at least a minimum 3x ROI the first year," he says. That is the threshold Techjays aims for when selecting opportunities.
For each initiative, he wants leaders to connect the investment to increased margins, increased revenue, or an improved customer experience—and quantify the expected impact. Starting with a focused project also gives his team a chance to understand the company's people, processes, and technology before expanding into other departments.
Samuelraj has a warning for the companies that can't. "If you're not able to clearly articulate it," he says, "then you may be just wallowing in governance, in theater."
