Have you noticed every vendor is now “agentic”?
We've been through this before. Remember a decade ago when every product was “cloud”? Some were. Most just tweaked their logo to make it look more "cloud-y". It’s happening again but this time the word is “agentic” and the stakes are higher because agents touch everything: customer data, systems, money, and more.
When you meet an "agentic" vendor, here are three tests that can help you cut through the BS and expose who is "agent washing" their product.
The kill test
Ask the vendor to kill their own demo mid-task, live, then restart it. Real ai agent orchestration resumes from the exact step that failed, without re-firing the steps that already happened. Agent washing restarts from zero and calls it “self-healing.”
The evidence test
Ask for the record of what the agent just did – not the chat transcript, the actions. Who did it act as, what did it call, in what order, and can the record be verified after the fact? If the answer involves the word “logs,” keep your checkbook closed.
The economics test
Ask what a failure costs. Durable workflows make the answer specific: nothing already completed is ever paid for twice. If the vendor can’t connect reliability to your token bill, they haven’t run this in production – they’ve demoed it.
One more tell, and it’s the big one: a real platform doesn’t make you abandon your current agent framework. The “rip everything out and use ours” pitch is pure vendor lock-in. Keep your framework. Add the reliability layer underneath. (We wrote up how that works: Keep Your Agent Framework, Add Durable Execution.)
Back in the day, cloud washing stopped working when people learned which questions revealed who was legit and who was not. For agentic vendors, you have 3 questions above.
Fore more detail on Why AI Agents Fail in Production, take a look!
Top comments (0)