I can't remember the last client conversation where having access to frontier AI was the interesting part. Everyone has it now. A couple of years ago, that alone would have been the whole pitch.
The models themselves have more or less converged: similar capability, similar pricing, open to whoever's willing to pay. Which makes me wonder why so many organisations I meet are still spending heavily on AI without much to show for it.
My take is that the advantage has simply moved somewhere else. It isn't the model anymore, it's everything built around it: how you govern it, route work to it, what it remembers, how well people are coached to use it. That surrounding system is what actually decides what your AI produces, day after day, at scale. At Cognizant we call it Cognizant AI Nexus, and most of the enterprises I talk to haven't built one yet.
Here's what that's costing you
Here's the odd part: the price per token keeps falling, yet most AI bills keep climbing anyway. A few reasons come up again and again when I dig into it:
- Different teams end up solving the same problem separately, and paying for it separately, because nobody has a way of knowing someone already cracked it.
- When one engineer works out a better way of using AI, that knowledge usually stays with them. It rarely makes its way into anyone else's workflow.
- Most governance tooling was built before AI existed, so it has no real way of tying spend back to what that spend is actually buying.
It's a fairly simple problem underneath all this: you can't manage what you can't see, and right now almost nobody can see it clearly.
This gap doesn't close. It widens.
I don't think this is a temporary lag that closes once everyone else catches up. It's structural. Organisations already running a governed AI system are quietly building up reusable patterns, coached skills and routing intelligence with every quarter that passes, and that's the kind of advantage that gets genuinely hard to catch up on, not just more expensive to match.
Waiting a quarter doesn't just push the start date back. That's ground you don't get back.
What we do about it
Cognizant AI Nexus sits between every request, whether it comes from a person or an agent, and the AI that answers it. It tracks every token that gets spent, routes each task to whichever model actually fits the job rather than the priciest one on offer, and feeds what your best people learn back to everyone else on the team.
We break it down into five moves: reveal, coach, route, distil, multiply; and each one makes the next a bit sharper.
For most leaders, governing AI spend at some point isn't really in question. What's less certain is whether they do it while the gap is still closeable.
Click here to read the full white paper
Want to talk it through? Get in touch with us to arrange a short, free discovery session on how to harness your AI
Arvind Pal Singh leads Cognizant's Consulting and AI Unit across EMEA. Arrange a discovery session.
