Quick answerRoute a support conversation to human sales when the customer's underlying need has shifted from resolving a problem to evaluating an expansion, upgrade, or new purchase decision that requires pricing authority or relationship context a support agent does not have. Keep it in support when the conversation is still about making the current product work as purchased, even if a sales-relevant signal (a mention of growth, a competitor, a budget cycle) comes up along the way.
Two related posts this is distinct from
This is specifically about routing an inbound conversation that started as support and reveals a sales-relevant need partway through, not about the format of any handoff package itself, which is a separate topic covering what information should travel with an escalation regardless of which team receives it. It is also not about a company proactively reaching out to a customer, covered in deciding when an AI agent should reach out first, which is about outbound-initiated contact, not the routing of an inbound conversation the customer already started.
The core test: what does the customer actually need next
The cleanest signal is whether resolving the customer's stated problem is enough, or whether the conversation has moved into questions only someone with pricing authority and account relationship context can actually answer: an upgrade to a higher tier, adding seats or usage capacity, or interest in a capability that requires a different contract. If a support agent (human or AI) could fully resolve the need by fixing, explaining, or configuring something within the current plan, it stays in support. If the resolution requires a decision about what the customer buys next, it belongs with sales.
From the team
We build production AI systems for startups.
LLM pipelines, RAG, and agent workflows that hold up under real traffic — not just in the demo.
Signals worth routing on, and signals that are just noise
A customer mentioning that their team is growing, in the course of an otherwise ordinary support question, is not by itself a routing trigger. It becomes one when it is paired with an actual question the support flow cannot answer: "we're adding twelve more users next quarter, what would that cost us" is a sales question. "We're adding twelve more people, can you also give them logins" is a support question with a sales-adjacent detail in it, and can often be resolved entirely within support with a note passed to the account team for awareness, not a live handoff.
This distinction matters because over-routing to sales on every loosely sales-adjacent mention creates two problems: it interrupts the customer's actual support need with an unrequested sales conversation, and it floods the sales team with low-quality routed conversations that dilute their attention on the ones that are genuinely ready. The same discipline applies to ambiguous or multi-intent request routing generally: route on the actual need, not on the loudest keyword in the message.
Designing the handoff itself
Once the routing decision is made, the handoff should preserve the full support context so the customer never has to re-explain what brought them into the conversation in the first place. It is also worth giving the customer a clear, low-pressure option rather than silently rerouting them: something like offering to connect them with someone who can talk through pricing for their growth plans, so the shift from support to sales feels like a service, not a redirect they did not ask for. Where the routing decision itself is genuinely unclear, defer to a human-in-the-loop check rather than guessing, since a wrong sales handoff is more disruptive to the customer than a short delay.
FAQ
Should the AI agent ever try to close the sale itself instead of routing to a human?
For most companies, no. Routing to a human with appropriate context is safer and more effective than having the support-context agent attempt a sales conversation it was not designed or authorized for.
What if the customer explicitly says they just want their support issue fixed and do not want to talk to sales?
Respect that directly. Resolve the support issue and, at most, leave an internal note for the account team, rather than pushing a sales handoff the customer explicitly declined.
How do you measure whether this routing decision is working well?
Track both under-routing (sales-ready conversations that stayed in support and went nowhere) and over-routing (routed conversations sales closes as not-yet-ready), and tune the routing criteria based on which error is more common.