Stop buying the roadmap
Keynotes are masterclasses in the future tense
Vendor keynotes are masterclasses in the future tense. Coming soon. On the roadmap. Generally available later this year. The capabilities are dazzling, and a surprising number of them are not things you can use today, on the release you're actually running or selecting.
There's nothing dishonest about it. Roadmaps are how software companies sell. But there's a quiet sleight of hand in an AI demo: the line between what's live and what's promised gets blurred, and you leave the room having priced the promise as though it were the product.
So you sign. You pay from day one. And then you wait, for the release, for the feature to reach general availability, for your environment to be eligible. The value you bought is real, but it's months out. Meanwhile the meter's running.
The discipline almost no selection applies
The discipline almost no selection applies to AI is the simplest one: separate what's available now, on your release, from what's coming. A typical RFP interrogates functionality in forensic detail and never asks the one question that protects the budget: when, exactly, can our people use this?
This matters more with AI than with traditional software, because AI roadmaps move monthly and the temptation to oversell is enormous. Every vendor is racing the same race, and the keynote is where they show their lead, not their shipping schedule.
The questions that protect the budget
None of this means roadmaps don't matter. Direction of travel is a fair thing to weigh, as direction, not as something you're paying full price to use now. The questions are blunt and they work:
What's generally available today?
On our release?
What's early-adopter, and what's just a slide?
Frequently asked questions
Why do AI vendor demos oversell what's available?
Vendor keynotes lead with the future tense: coming soon, on the roadmap, generally available later this year. A surprising number of demoed capabilities aren't usable today on the release you're actually running.
What's the risk of buying on the roadmap?
You price the promise as though it were the product: you pay from day one, then wait for the feature to reach general availability or for your environment to be eligible, while the meter keeps running.
What discipline should selection apply to AI?
Separate what's available now, on your release, from what's coming. Ask when, exactly, your people can use it, not just what it can eventually do.
What questions protect the budget?
What's generally available today? On our release? What's early-adopter, and what's just a slide?
The question that follows
So when the demo dazzles you: how much of that can you actually use right now?