If personal AI agents start handling consumer errands at scale, some of them will also do the first round of shopping for a buyer, and a sales team that has no policy for that caller will find out what its policy is by accident. I think sales leaders should write the policy this quarter.
What prompted this
Kishan Chetan, EVP and GM of Agentforce and Service Cloud at Salesforce, published a post on October 19 arguing that personal agents will drive more contact center traffic than they offset. His reasoning is about patience. Personal agents “don’t mind waiting on hold,” in his words, and an agent asked to keep an eye on a delayed order can call back every few hours until it ships. Because an agent keeps no business hours, he writes, its calls can arrive at 2 a.m. as easily as at noon.
One of his examples is a sales-shaped one. An agent shopping for a better phone plan, he says, can contact several providers in the time it takes a person to dial one. His prescription for contact centers: “The best way to prepare is to meet personal agents at the front door with your own business agents.” I have no quarrel with that for service. I think revenue teams face a second question that the post does not take up.
Advertisement
My argument
Take Chetan’s phone-plan example and move it to a software purchase. A buyer asks a personal agent for three vendors that meet a requirement and a price range. The agent contacts all of them in the same hour. The vendor whose pricing page, qualification form and response queue can answer a machine caller gets into the comparison, and the vendor that routes the request to an inbox for Thursday does not. Salesforce’s post covers calls to a service line, and I am applying its logic to the front of the funnel. Nothing I have read says buyers are doing this at volume today, so this is a bet on a direction and not a report on current traffic.
That is why I think the work belongs to sales leaders and RevOps as well as the service organization. Four decisions need an owner:
- What a machine caller can learn without a human, such as list pricing, packaging and implementation timelines.
- What a machine caller must hand to a human, such as discounts, custom terms and security reviews.
- How inbound agent contacts are labeled in the CRM, so they do not count as a person’s intent or poison a lead score.
- Who is accountable for the response-time target on those contacts.
Our earlier column on speed to lead as a routing problem covers the routing side of response time. A rule written in advance costs less than a routing fix written after a quarter of odd data.
The strongest objection
The best counter-argument comes from Chetan’s own post. He notes that a personal agent may hold a user’s inbox and whatever the user has shared, but likely lacks the full picture of the relationship with a business, and that agents run into security checks and questions only a human can answer. If agents stall that often, the objection runs, a B2B sales team can ignore them for now. Large purchases involve committees, procurement and legal, and no agent signs a contract.
I accept the premise and disagree with the conclusion. Stalling at the contract stage tells us little about who is on the shortlist at the first stage. If an agent can compile a comparison before any human talks to a vendor, the vendors it could not get answers from are missing from the table when the human arrives. Our report on Salesforce’s argument that businesses need their own agents to greet personal agents described the service side of this. The sales side has the same shape and a higher cost per missed contact.
What I would do this quarter
Start with an audit of what a stranger, human or machine, can learn from the public site without filling in a form. Then write the handoff list from the four decisions above and put it where the sales development team and the service team both read it. Add a field in the CRM for source type so agent contacts can be counted separately. Review the count in 90 days. If the number is zero, the field cost an afternoon. If it is not, the team already has a baseline and a rule.
Source: Salesforce
