Managing Three AI Products with One Operator

Running multiple software products with a micro team sounds romantic until you're staring at three separate Slack workspaces, four GitHub repos, and a support queue at 11pm on a Thursday.

MindFortress builds in three domains at once: Freya, a long-form writing assistant for novelists; AgentPik, a platform that deploys custom Slack bots for SMBs; and Reeve — the AI business operator at the centre of it all. Each product has its own customers, its own deploy pipeline, and its own weekly release cadence.

With two people on the team, every context switch has a cost. Before Reeve, the founder estimated spending 60% of each day on coordination — monitoring dashboards, triaging bugs, reviewing which deploy succeeded, chasing status across repos — and only 40% on actual building.

Reeve became the connective layer across all three products. Each product got its own named agent with a tailored memory profile — Freya's agent knows the writing partner's user personas and content roadmap; AgentPik's agent tracks Slack workspace health and bot deployment status; Reeve's own agent watches its pipeline workers and marketing metrics.

The Pipeline Manager handles cross-repo tasks that span multiple codebases — shipping a feature that requires a backend API change in one repo and a UI update in another is now a single task dispatched to two parallel workers, not a two-day context-juggling exercise.

Support for Freya is handled almost entirely by an agent trained on the product's documentation and common writing questions. For AgentPik, the Slack integration means customers get responses within minutes, any hour of the day.

Critically, every action every agent takes is logged to a shared Activity Feed. At a glance, the team knows exactly what Reeve did overnight — without reading a dozen status emails.

Within the first month of running all three products through Reeve, weekly deploys across repos climbed from roughly 4 to more than 20 — not because the team worked harder, but because so many of the coordination steps that used to require human attention were handled automatically.

Freya's user support response time dropped from an average of 6 hours to under 12 minutes. AgentPik onboarding went from a manual setup call to a fully automated flow. Reeve's own ad campaigns — managed by its own marketing agent — ran their first A/B test without anyone writing a brief.

Most importantly: the founding team went back to spending the majority of their time building, not managing.