6sense Runs Its Own AI Email Agent on Guardrails, Not Autonomy

6sense’s own AI email agent asks a new inbound lead three questions and hands the thread to a human rep once the prospect replies. In the example 6sense’s Dan Hieb shares, the prospect replied within three minutes, named HubSpot as their stack, said they wanted ad platforms beyond Google, Meta and LinkedIn, and the agent passed the conversation to a rep.

The program, which 6sense calls Customer One, runs on an AI Email agent after a form fill. The opening email asks about the prospect’s CRM and marketing tools, which 6sense features they want to explore, and when they are free for a call. Three questions, one handoff, no pretending to be the closer.

The configuration is where top sellers will find the lesson, because it is mostly guardrails. Setup needs three settings: sending hours, defaulting to 8 am to 5 pm though many US teams extend to 8 pm; CRM field mapping that includes the opt-out field; and the list of reps allowed to take over conversations. Replies the agent cannot match go to a person. Contacts with recent BDR or AE activity, such as in the last 30 days, are excluded so the agent never collides with a rep.

That last rule matters. In a companion piece, Hieb describes the failure mode of specific but wrong output, like congratulating a contact who left months ago or pitching a product the account already uses. Both pieces make the same point: agent quality depends on the account data behind it, and the exclusion rules keep it from stepping on your reps.

Teams can upload knowledge files up to 5 MB each, personalize paragraphs with up to four fields, set fallback text, and run sequences of four to five emails. Both pieces come from a vendor promoting its own product, so test the setup against your own conversion data. Sources: 6sense on Customer One and 6sense on agentic GTM output.