Why AI Strategy Needs a Reality Check

Most management is still stuck on "AI strategy", and honestly, this has been going on for like 2 years now.

You know those conference posters, the ones with faces of people who are gonna talk about AI, and you personally know that guy knows shit about AI. I always laugh and go, "This guy is talking about AI? What's he gonna say?" But here's the thing: if you actually pull these people aside and ask them real questions about AI, you'll just hear buzzwords, nothing more. You won't find the depth or insight you thought you'd get.

Now you might say, Prathamesh, this was happening before too; leadership always talked in buzzwords while the real work happened somewhere else. And yeah, fair. But the difference now is, it's costing money. AI isn't cheap, and if everything just stays at the strategy level and nothing actually converts, that spend is just gone.

You can see it playing out already. Uber apparently burned through its entire 2026 AI budget by April. And a bunch of companies are quietly pulling back on AI contracts because token costs are eating whatever value the AI is supposed to be creating.

Then there's the Xbox thing, which I think is the cleanest example. Asha Sharma becomes Xbox CEO in Feb 2026. Before that, she ran Microsoft's CoreAI division for 2 years, so this isn't some random exec; this is someone who actually knows AI inside out. In May, she just kills Copilot for Gaming on consoles, a feature Microsoft had been pushing as a big AI flex, and basically says it wasn't solving any real problem for players; not every AI thing is something people actually want.

That one hit different for me. The person who actually understands the tech is the one saying, "We don't need this here."

Honestly, I think this is the real issue. Somewhere along the way, "AI can solve this" became the default answer to literally everything, which is kinda delulu when you think about it. That was fine when AI was cheap, and it was all upside. It's not fine anymore when every "let's bolt AI on here" is real money out the door, and someone in the room needs to be able to ask: does this problem even need AI, or are we just paying more for the same answer?

I get it, business folks don't need to know the tech they're talking about; they care about revenue, clients, growth, all that. But in the age of AI, you might end up selling a client a dream that costs both of you in the end. So maybe you do need to know this stuff in depth now, not just enough to sound like you do.