- Adi Singh of AgentMail argued that agents will gain identity on surfaces like email, not through a new issued standard.
- He also warned that these old surfaces may break under agent load, and that giving agents a person's own inbox widens the damage when things go wrong.
Adi Singh, co-founder of AgentMail, argued on the AI Engineer show that AI agents will not get their identity from a new standard. In his view, agents will build it up on surfaces people already use, starting with email. Trust, he said, is something that accumulates rather than something handed out.
What was said
Singh said the systems we use today rest on trust that grew over time, including the errors and fallbacks along the way. He applied the same idea to agents: "it's not so that we get issued one, I think it might be that we accumulate one." He compared this to the early car, which still ran on roads built for horses.
He was cool on the new specs. These include agent cards (JSON files that verify each agent), passports and decentralised identifiers. Singh said they work like a key card: they say who you are and what you may do, then expire. Tighter scoping and faster expiry make sense for security, he said. But they leave less room for adoption.
His example was a phone carrier whose agent sign-up flow assumed the user had an email address. Agents often do not. The carrier added a call to AgentMail so an agent could get an inbox. The agent then received a verification code and an API key. The inbox became a permanent contact point.
He made a similar point about payments. An agent with a credit card still needs an email address to finish checkout. For refunds and disputes, someone needs a record of who the agent was and why it signed up.
Why it matters
Our reading: if agents become regular visitors, an inbox may matter as much as a login for anyone running a sign-up flow. Flows written only for humans can fail when the visitor has no email. Buyers could also ask a plain question of any agent tool: if something goes wrong, can we trace who the agent was and why it acted? Singh's account suggests that record may live in old channels, not only in new identity papers.
The other side
Singh himself said this will not last. He expects existing services to buckle under the strain as they get overcrowded with agents. He also warned about people giving agents their own email or WhatsApp access. A wrong message can go out, and he described "the blast radius for all of these things expanding and exposing."
There is a further caution. Singh runs a company that sells email to agents, so he gains if his view is right. He also said his talk was not a claim about exactly how agents will identify themselves. The excerpts do not show how a service would tell an agent from a person, or how trust would be withdrawn once it is earned.
Written by the WebPulse Newsroom with AI assistance, and checked by our editorial review: every quotation was verified against the recording's transcript. How we use AI.
The conversation this talking point comes from
- AI Engineer: The Next Trillion Internet Users Still Don’t Have an Identity — Adi Singh, AgentMail (2026-10-12)
