Last week Sierra raised $950 million. $15.8 billion valuation. Co-founded by Bret Taylor — former Salesforce co-CEO, current OpenAI board chair.
Background done. Here's the number that actually matters.
Nordstrom deployed Sierra's AI customer agent in five weeks. Cigna in eight. Cigna cut patient authentication time by 80%. The company counts 40% of the Fortune 50 as customers and crossed $150M ARR in just eight quarters.
Five weeks.
Last November, we were in a meeting about a follow-up feature for a clinic scheduling tool. A patient sends "can we reschedule?" — the system should respond automatically. Reasonable feature. It went into the backlog. Sprint planning came around. "No capacity." Moved to next sprint. And the next. And the next.
In the same period, Sierra took one of the world's largest insurers to production in eight weeks.
PMs usually ask: "When should we add the AI feature?"
Wrong question.
The right one: "How long before an AI company solves this need better and faster than us?"
Sierra's model isn't complex. They enter large enterprise accounts, collect domain data, integrate the system, and go live in 5–8 weeks. It works. $150M ARR in eight quarters is not hype — it's evidence.
While you're triaging backlog items, someone else already shipped.
"But our product is different." Domain-specific. Smaller user base. Sierra is enterprise-only.
Partly true. But think about this: Sierra's core value proposition is a formula — integrated AI + domain expertise + fast deployment. That formula scales down. It shows up in every vertical, every size.
Dental clinic communication automation? Exists. Dietitian patient follow-up? Exists. Small clinic appointment management with purpose-built AI? Someone is shipping it right now.
And you've been trying to prioritize this in the backlog for six months.
Here's the question you actually need to ask: Are my users waiting for us to solve this, or are they already looking elsewhere?
Three users raised the same complaint in four weeks? That's a signal. The moment you route that signal into next quarter's backlog, you've effectively passed it to a competitor.
Sierra didn't teach us that big AI companies will steal your customers.
It taught us that speed is a product moat. And PMs keep being the slowest link in that chain.
To be clear: you can't fit everything into a sprint. Tradeoffs are real. But "when should we add AI?" gives you the wrong frame — it makes you feel like you have time.
You don't.
The actual question is: "Who is going to solve this user need, and do I want it to be us?"
If the answer is yes, the sprint conversation looks different.