Will This AI Category Still Exist in 2027? A Durability Test for Revenue Buyers

Buying a product that disappears is not the risk. Vendors get acquired constantly and customers survive it. The risk is buying a category that disappears, because then there is no acquirer with a reason to maintain your integration and no replacement that works the way you built your process.

Durability is assessable in advance. It comes down to five factors, and none of them are about the current feature set.

Factor one: proprietary data

Does the product accumulate data that a competitor could not reconstruct? A relationship graph built from years of your own activity is durable. A wrapper around a general model with no data accumulation is not, because the moat is rented and the rent is falling.

Test: if the vendor lost every model provider relationship tomorrow, what would still be valuable in their product? If the answer is nothing, score zero.

Factor two: workflow depth

Does the product own a workflow end to end, or does it produce an output that someone pastes elsewhere? Categories that sit inside a daily operating rhythm survive. Categories that produce an artefact get absorbed into whichever tool already owns the rhythm.

Test: name the recurring meeting or daily task that would break if this product disappeared on Monday. No answer means no depth.

Factor three: switching cost that is not lock-in

There is a difference between a product you cannot leave because your data is trapped and one you do not want to leave because leaving means rebuilding a process that works. The first is fragile: it fails the moment a credible alternative offers migration. The second is durable.

Test: would a competent competitor be able to migrate you in under two weeks? If yes, this is a feature-priced category and your contract should reflect it.

Factor four: distance from the platform roadmap

The clearest durability signal in this cycle is how close a category sits to what the large platforms are already shipping. When a platform makes its own interface optional and pushes agent execution into its own stack, adjacent point solutions lose their reason to exist as separate purchases. We tracked that shift in the headless GTM stack and in the Agentforce pricing pieces.

Test: is the vendor's core function on a published platform roadmap, in beta, or already shipped as a bundled feature? Each step closer cuts the durability score.

Factor five: evidence class

Categories where customers publish results survive scepticism. Categories where all evidence is vendor-produced collapse the moment one high-profile buyer publishes a negative result, because there is nothing to weigh against it. The autonomous SDR category demonstrated exactly that, as covered in 11x after the walk-back.

Test: can you find two customers who published results independently of the vendor, with names attached? The compiled state of evidence across categories sits in the proof gap research.

Scoring and what to do with the score

Score each factor zero to two. Ten is maximum durability.

• 8 to 10. Infrastructure. Multi-year term is reasonable. Negotiate on price and roadmap influence, and build process depth deliberately.
• 5 to 7. Contested. Twelve-month term, kill criteria written, and an annual re-score. Do not build irreplaceable process on it yet.
• 3 to 4. Feature in waiting. Monthly or quarterly commitment. Assume absorption within eighteen months and keep your data exportable in a format you have actually tested.
• 0 to 2. Do not buy as a platform. Use it, pay per use, and never make it a dependency for a reported metric.

Note that a low score is not a reason to avoid the tool. It is a reason to change the contract. Plenty of low-durability products are worth using this quarter. They are not worth a three-year commitment for a twelve percent discount.

The three questions to ask the vendor directly

1. What do you do that the platform we already pay for cannot do within twelve months, and why is that gap structural rather than a matter of their priorities?
2. What data do you accumulate that gets more valuable the longer we stay, and can you show us it?
3. If you were acquired next quarter, which parts of this product would the acquirer keep, and which would they fold in?

The third question is uncomfortable and the good vendors answer it well, because they have thought about it more than you have. An evasive answer is itself the score.

Re-score annually

Durability is not fixed. A platform ships a feature and a seven drops to a four in one keynote. Put the re-score on the same calendar as renewals, hold it in the same review as your supplier shortlisting process, and let the score set the term rather than the other way round.

Related: AI vendor diligence questions · Headless GTM stack · AI tech landscape · Skill: testing AI category durability