Is your product ready for AI agents?

Ten situations your product will face when customers send their agents instead of coming themselves. Two minutes, no sign-up.

Discovery1 of 10

A buyer asks their agent: “Can this product do what we need, and what will it cost us?” Where does the agent get the answer?

What it checks

Discovery
Can agents find your price and limits?
Access
Can customers limit what agents do?
Accountability
Can you tell who did it?
Doors
How do agents get work done?
Metrics and revenue
Do your numbers survive agent users?

Questions

What does AI agent readiness mean for a product?
A product is ready for AI agents when an agent acting for a customer can find out what it does and costs, get in with limited access, take actions that a person still approves when they matter, leave a record of what it did, and be counted separately from the person in your metrics. Most products today assume a person is looking at the screen.
Why does this matter now?
Personal agents such as Claude Duet, OpenAI’s Dots, Meta’s Muse and Grok Bot can act for people in the background. Sierra and Meta are drafting a Personal Agent Protocol for how those agents sign in to businesses, and Cloudflare now lets sites see and set rules for personal agent traffic. Nobody has published how much of this traffic exists yet, so this is the time to prepare.
Who is the scorecard for?
Product managers, founders and engineering leaders at software companies, whatever you sell. If customers log in to your product, their agents will too.
How long does it take?
About two minutes. Ten situations, three answers each, and your score appears as soon as you answer the last one.
How is the score calculated?
Ten questions, each worth 0, 1 or 2 points, for a total out of 20. 15 or more is Agent-ready; 8 or more is Agent-curious; 0 or more is Built for eyes. The bands are judgment calls, not industry benchmarks.
What happens to my answers?
We record which answers people choose, without your name or email, to see how product teams compare and to report the results. There’s no sign-up. See our privacy policy and cookie policy.
What should I do with my score?
Start with the three areas the scorecard flags. Each has one first step you can take this quarter. The full playbook, with what to build first in all five areas and how to tell it’s working, goes out free by email.
Who made this scorecard?
Shrey Malhotra, co-founder and Chief Product Officer of CambrianEdge.ai. It builds on his essay Your next user has no eyes.
See all ten situations
  1. A buyer asks their agent: “Can this product do what we need, and what will it cost us?” Where does the agent get the answer?
  2. The agent tries to start a free trial on the buyer’s behalf. What happens?
  3. A customer wants their agent to read invoices but never change billing. Can they set that up?
  4. The agent tries to cancel a subscription, delete a project or email every customer. What happens?
  5. A customer writes in: “Did I change this setting, or did my agent?” Can support answer in five minutes?
  6. One customer’s agent starts misbehaving at 2 AM. Can you stop it without locking the customer out?
  7. How can an agent actually get work done in your product today?
  8. The task needs back and forth, like a refund, a custom quote or a warranty claim. Who handles the customer’s agent?
  9. Customers who work through agents stop seeing your upgrade prompts, product tours and “what’s new” screens. What happens to expansion revenue?
  10. The person behind an account hasn’t opened your product in 60 days. Their agent logs in every morning. What does your customer health score say?