Telcoms already ran this experiment, and it's now impacting their P&Ls. VMO2's ARPU is in decline, debt is high, and the business is set to be asked to make £600m in cost savings. BT's consumer revenue has fallen 2% year on year, and it's running a multi-billion pound cost transformation programme.
Every Telco has had the same pitch: unlimited data, unlimited calls, unlimited everything. Different logos, same networks underneath. When every player can do the same thing, the product stops being the product. Price becomes the product.
It should have already been obvious where this ends. Airlines did it first — strip away the loyalty points and a seat is a seat. Web hosting did it next — same servers, same stack, price the only lever left. Generic pharma did it too — patent expires, ten manufacturers make the same pill, margin gone. Telcoms weren't writing a new playbook. They were just late to read the old one.
Same models, same playbook, same fate
SaaS now risks walking the same worn path.
Most AI features run on the same handful of foundation models, and prompt-engineering often follows best practice. Same underlying engine, different logo on top. Sound familiar? It's the network all over again — different pipe, same commodity. The only thing left to compete on is price.
Why does this happen so fast in AI?
Software used to take years to copy. Build a UI, ship an integration, earn some trust — that used to be the moat.
AI breaks that timeline. Model APIs are commodities. Agent frameworks are open source. A competent team can clone an "AI feature" in a sprint, not a year.
Feature parity used to be a lagging indicator. In AI, it's basically real-time.
So what actually survives?
Outcomes. Not access.
Selling "we called an API well" is selling something anyone can rebuild. Selling "the ticket got closed," "the meeting got booked," "the revenue got recovered" is selling something the buyer already budgets for, already values, and can't get from a cheaper wrapper.
Outcome pricing doesn't compete against cheaper AI. It competes against the cost of doing it the old way.
Telcoms learned this too late. The ones who tried to escape the price war with more bundles, more loyalty points, more unlimited-everything — they just competed harder on the same commodity. The ones who moved to selling connectivity outcomes (uptime guarantees, managed service contracts, SLAs with teeth) found one of the few exits that wasn't a straight race to zero.
But don't get comfortable
Outcomes aren't a permanent shield. The easy ones commoditize too. Basic summarization-as-a-service? That's heading the same direction as the chatbots.
The real moat isn't "priced on outcomes." It's the outcome being genuinely hard to deliver — compound workflows, proprietary data, someone willing to own the risk when it goes wrong. That's what a small team with API access can't clone over a weekend.
The uncomfortable summary
Model access → commodity, price war, race to zero.
Features → temporary lead, closing fast.
Outcomes → real moat, but only if the outcome is actually hard.
Access is easy to copy. Outcomes are easy to talk about. Hard outcomes, reliably delivered — that's the only thing left that stays defensible.
Telcoms already showed the industry how this movie ends. SaaS still has a chance to skip the sequel.