What Is Kaizen? Inside Zenit's AI Discovery Engine

Kaizen is Zenit's AI discovery engine: an agent that joins a project's meetings, cross-references what it hears across different conversations, and builds an actionable brief —with scope, milestones, and the recommended squad— without anyone having to summarize a recording or fill out a form. Unlike a meeting notetaker, Kaizen asks follow-up questions when something's unclear and stays present through the whole execution of the project, not just the kickoff meeting.

Zenit organizes it into five acts —Listen, Understand, Document, Recommend, and Stay On— that run from the first meeting to the last milestone delivered. This article covers how it differs from a conventional AI notetaker and what each act actually does.

Why isn't an AI notetaker enough to do discovery?

Recording and transcribing meetings stopped being a niche decision a while ago: three out of four professionals now use an AI notetaker in their work meetings (Fellow.ai, The State of AI Meeting Notetakers, 2025). But the same study found something uncomfortable for that category of tool: 84% of people change how they speak when they know a notetaker is recording —they instinctively edit what they say for the transcript, not for the other team to understand. A perfect text file of a meeting where nobody quite said what they meant doesn't solve the underlying problem: someone still has to read that text, connect it to the other five meetings that week, and decide what's missing.

That's also the direction the broader AI tooling market is moving in: Gartner projects that less than 5% of enterprise applications included agents capable of acting independently in 2025, a number it expects to rise to 40% in 2026 (Gartner, August 2025) — the shift from assistants that summarize to agents that participate and decide. Kaizen was built on that same logic from the start: it isn't a transcription layer sitting on top of a project's meetings, it's a participant that understands the project.

How does Kaizen work? The five acts of discovery

Every project that goes through Kaizen follows the same arc, from the first conversation to the last milestone delivered:

Act 1 — Listen: participating, not transcribing

Kaizen joins the team's meetings as an active participant, not a silent bot recording in the background. It asks follow-up questions when something's unclear —whether one milestone depends on the previous one finishing, or whether they can run in parallel, for example— and spots in the moment what information is missing to move forward. Asking for that clarification live is exactly what a passive notetaker can't do: the ambiguity stays buried in a text nobody will re-read with the same judgment it was said with.

Act 2 — Understand: a map, not a stack of transcripts

No single meeting tells a project's full story. Kaizen cross-references what it heard across different conversations and different people —whoever leads the product, whoever will build it, whoever holds the budget— to build a single map of the process: what resources exist, what's missing, and above all, where two people on the same team described the same priority in contradictory ways. Cross-referencing meetings against each other is exactly what a transcript doesn't do on its own: each file stays isolated from the rest.

Act 3 — Document: the brief writes itself, live

What comes out of act 2 turns into a brief with scope, milestones, and team requirements —no manual meeting notes, no relying on someone remembering to write the summary after the call. Kaizen comes back to that same document later, once the project is underway, to check that what's being built still matches what was agreed at the start.

Act 4 — Recommend: why each squad fits, not just how it scores

With the brief closed, Kaizen cross-references it against delivery history, seniority, and real availability for each squad in the network, and explains the reasoning behind each option —not just a score. The final call always belongs to the company: Kaizen doesn't replace that judgment, it gives it the context a standalone profile never shows on its own.

Act 5 — Stay On: the act most discovery tools don't have

This is where the biggest difference from any meeting or matching tool shows up: Kaizen doesn't disappear once the match is made. It hands the full brief to the squad, watches progress milestone by milestone alongside SafePay, and flags risk signals before they turn into a real problem. If a genuine question about scope comes up, it goes back to the company to confirm before moving forward —the same brief from act 3, checked against how the project is actually going.

Why does staying on matter as much as the initial brief?

A well-built brief doesn't guarantee a successful project if nobody looks at it again. According to PMI (Requirements Management: A Core Competency for Project and Program Success, 2014), 47% of projects that fail to meet their goals do so because of inaccurate requirements management, and organizations waste an average of 5.1% of every dollar spent on projects for that same reason. Most of those failures don't happen in the kickoff meeting — they happen weeks later, once scope has quietly drifted and nobody went back to check it against the original brief. That's why Kaizen doesn't close its work at act 4 — act 5 exists specifically to return to that document during execution, not just to build it once.

Does Kaizen replace the company's own judgment?

No. Kaizen recommends with data and explains the reasoning behind each option, but the final call —which squad, what scope, what priority comes first— always belongs to the company. The goal isn't to take people out of the decision: it's for them to reach that decision with the full context of what came up across five different meetings, instead of a list of profiles that match by keyword.

Today Kaizen already listens, understands, documents, and recommends during squad pre-registration —it's the same mechanism it uses to interview the teams joining the network. Production support in act 5 and discovery on the company side are the next stage on the roadmap.