Salesforce wants companies to stop treating customers’ personal AI agents as a traffic problem. In an October 9 post, Mark Wakelin, EVP of Forward Deployed Engineering at Salesforce, argues that businesses should deploy agents of their own to receive them.
Wakelin’s post describes a split among businesses. Some block personal agents at the front door, and others open their websites and APIs to them. He says both choices lose something. Blocking keeps control but sends customers somewhere else. An open website and API wins the traffic but “forfeits the relationship,” because the personal agent weighs the options and decides what the customer buys.
His fix is a business agent that answers questions from company data, checks each request against company policy and uses deterministic logic for refunds, payments and escalations. The aim, he writes, is to stop a persuasive personal agent from reaching “an unauthorized refund or credit.” He names Agentforce as the way to build it. The post is a vendor’s argument for the product it sells, and it cites no customer results for the approach.
For a sales leader the useful part is the sequence it implies. If a buyer’s agent does the comparing, the first conversation in your pipeline may be with software, and whichever side answers it sets the terms. Our read is that this makes quoting and discount rules the first thing to settle: an agent can only hold a line that someone wrote down. We covered the same rules-layer idea in Sales AI Vendors Now Pitch a Rules Layer Between Agent and Action, and the data side in Salesforce White Paper Flags Shared Agent Memory as a Trust Risk.
The open question is who keeps the customer data when two agents transact. Wakelin says it should stay on the business side. The post does not say how a business would enforce that against a personal agent running on another company’s platform.
Source: Salesforce
