How Gleap uses Overwise.
In-app feedback software that helps product teams capture bug reports and feature requests. Gleap's buyer is a B2B SaaS founder — exactly Overwise's beachhead ICP. Dogfood case #2, run by the same founder.
The challenge
Gleap targets B2B SaaS founders — a segment where Apollo has decent coverage, so the bottleneck wasn't lead-finding. It was writing copy that didn't sound like every other AI-generated cold email landing in the same inboxes. The founder had been hand-writing each one and hitting a ceiling on volume he could sustain without the copy quality collapsing into generic AI-cold-email cadence.
The approach
Brand-voice extraction from the founder's sent folder gave the agent a working model of his tone — short, specific, anti-corporate. The MessageVerifier's cite-or-discard step kept the agent honest: every factual claim in a draft has to cite a span of the lead's actual context (recent funding, hiring page, tech stack, product changelog) or the draft is discarded and re-drafted. Cost-budget caps prevent runaway spend during scale-up pushes. Modes are mixed — autopilot for high-confidence leads where the signal is clear, one-click approval tasks for borderline ones the founder wants to read before they ship.
The result
The copy reads like the founder wrote it, because the brand-voice model was trained on emails he actually wrote. The cite-or-discard gate means the founder doesn't have to read every draft to trust it didn't fabricate something — verification fails are surfaced as AI Inbox tasks when they happen, not buried in a log. A second-order benefit: the per-lead reasoning surfaces signals the founder hadn't been looking for — prospects mid-switch from competing feedback tools, prospects who just shipped a public roadmap, prospects whose docs mention 'beta' in three different places. Signal the agent finds is signal the founder uses on the next campaign.
It reads like I wrote it. And I haven't touched outbound copy in weeks — the AI Inbox is the only outbound surface I open.