Agent washing
Agent washing is presenting ordinary automation (a scripted chatbot, a rules engine, an assistant that needs a person to finish the job) as an autonomous AI agent, or claiming results an agent does not produce. Gartner used the term in June 2025, describing vendors that rebrand assistants, chatbots and robotic process automation as agentic AI without substantial agentic capabilities (Gartner, 25 June 2025).
Regulators#
- In September 2024 the US Federal Trade Commission announced a sweep of cases over deceptive AI claims, saying there is no AI exemption from the laws on the books (FTC, 25 September 2024).
- In August 2025 it sued a company that advertised "conversational AI" it said could replace human customer service representatives, alleging false earnings claims and refund guarantees made to small businesses (FTC, 25 August 2025).
Why it matters to a buyer#
Agent washing shows up in the numbers as outcomes that a person finished: a "resolution" a support agent completed after the AI, a meeting a salesperson actually set, an appointment staff entered from the AI's request. Each of Attify's standards has a check for exactly this: handoff for support, "person stepped in" for sales meetings, and "not booked by the AI" and "staff fixed it" for bookings.
What to ask for#
Evidence from your own systems, not the vendor's dashboard: who created each record, who wrote each reply, and what happened afterwards. See how to check an AI vendor's claims before buying.
Questions
How do I tell a real AI agent from agent washing?
- Ask what it completes without a person and how that is counted, then check the count in your own systems on a trial: did the outcomes it claimed happen, and did anyone on your team have to finish them.
Is overstating an AI agent's results illegal?
- In the US, the FTC treats unsupported AI claims like any other deceptive claim. Its 2024 sweep and its 2025 case over AI agents sold to small businesses are on ftc.gov. This is not legal advice.