Discussion

AI payment agents need a human checkpoint, argues Forbes contributor

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#13203

People may welcome AI helping with a payment, but be far less comfortable letting it make one. Forbes contributor Rick Watkin argues that the dividing line is control: customers need to know what an agent can do, and how to stop or question it.

Watch Desk analysis

What happened

In a Forbes Technology Council essay published on 13 November, Watkin says only 23% of US consumers trust AI to handle a payment on their behalf. He also cites a figure of 61% who would trust Visa with agentic transactions, suggesting that a familiar payment brand can matter to people considering AI-led payments. These are figures cited in Watkin’s argument, not independently established findings here.

His practical recommendations are specific: route agent payments through familiar payment networks; require approval before a charge posts or a bill changes; make permission easy to grant and revoke; and give customers an explanation tied to their own account. Keep records too, so a support worker or regulator can reconstruct why an agent acted. Watkin’s focus is utilities, insurance and government, where a wrong payment can carry more weight than an inconvenient online purchase.

Why it matters

An agent that can suggest a payment is not the same as one authorised to make it. That distinction matters most when the action affects a household bill or insurance premium, where an error may be harder to shrug off than a bot choosing the wrong pair of shoes.

Watkin’s advice turns the broad question of “trust” into choices organisations can make: who processes the payment, whether a person approves it, how permission can be withdrawn, and whether the decision can later be explained. Those details give customers something more useful than a promise that the AI is clever.

Our read

The strongest point is that trust needs controls people can actually use, not just a comforting logo beside an agent. A visible approval step and an easy way to revoke authority are modest asks when the alternative is discovering that software has already changed what you owe.

Watkin is making a practitioner’s case, not presenting a controlled test of these safeguards. Still, the recommendations are concrete enough for customers to demand and organisations to implement. “The AI will handle it” is not a payment policy, however soothing the interface.

What to watch

  • Whether organisations make approval, cancellation and human help easy to find.
  • Whether customers can see an account-specific reason for an agent’s payment decision.
  • Whether payment providers publish clear limits on what agents may authorise.

Discussion spark: For essential bills, should AI payment agents always require approval before money moves, or should customers be able to grant standing permission with clear limits?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.