AI notetakers produce transcripts nobody acts on. We build the system on top: AI reads the notes, extracts the actions and decisions, and proposes an owner — a person confirms, and decisions route into your decision log.
Common pain points
The notetaker captures everything that was said, so the real action items and decisions are buried in pages of transcript nobody re-reads.
An action item with no named owner and no due date is just a sentence in a doc — it doesn't move.
Decisions made in the meeting are recorded as talk, not as decisions, so weeks later the same thing is discussed again.
Follow-through depends on one person scanning the notes and chasing people — when they are busy, it doesn't happen.
Different teams use different notetakers, so there is no single rule for what counts as an action and where it goes.
The notes never connect to where the work actually lives, so the action items and the team's task list drift apart.
The system has three parts. A rule that picks the source-of-truth notetaker and the line between an action, a decision, and an FYI. AI that reads the output, extracts and classifies actions and decisions, and proposes an owner and a due date. And a routing path — actions to where the team works (Asana, Jira, Linear, or a shared channel), decisions to your decision log. A person confirms every owner and every decision; AI never assigns work or records a decision on its own.
How we work
We pick the notetaker that is the source of truth and define, with the team, what counts as an action, a decision, and an FYI.
We connect AI to read the notetaker output, extract and classify the actions and decisions, and propose an owner and a due date for each.
We route actions to where the team works and decisions into the decision log, with a person confirming, and hand over a runbook the team operates and edits.
What we build
Area 1
We define which notetaker the system reads from and the line between an action, a decision, and an FYI — so the system fills with work and decisions, not transcript. The rule is agreed with the team, with examples of the edge cases that confuse the most.
Deliverable
A source-of-truth rule: the chosen notetaker, plus the written definition of action vs decision vs FYI with examples.
Area 2
We connect AI to read the notetaker output and pull out the action items and decisions, proposing a named owner and a due date for each from the context of the meeting. The proposal is the starting point — a person confirms the owner before anything is treated as real work.
Deliverable
An extraction flow: each action item with a proposed owner and due date, ready for a person to confirm.
Area 3
The decisions the system finds don't live here — they route into your decision log (the one a Decision Capture system runs, if you have it) as a proper record: what was decided, who owns it, and why. This system is the bridge from notetaker output to that log, not a second place decisions live.
Deliverable
A decision hand-off: decisions routed into the decision log as records, with the boundary to the log written.
Area 4
We connect AI to read the notes, extract and classify actions and decisions, propose owners, and route them. What AI may never do — assign the work itself or record a decision a person hasn't confirmed — is written down. AI prepares; a person confirms.
Deliverable
An AI capability map with the limits written: what AI reads/extracts/classifies/routes, and what only a person confirms.
Area 5
Most teams accumulate several notetakers and govern none of them. We map which tools are in use, what each is actually relied on for, and recommend which to keep as the source of truth — so the output feeds one system instead of scattering across three. We name current tools as examples; we do not sell or integrate any of them.
Deliverable
A consolidation recommendation: which notetakers are in use, which to keep as source of truth, and why.
Area 6
We hand the system over as a runbook: the source-of-truth rule, the extraction and routing, the decision hand-off, and the AI limits. The goal is for the system to live without us — the team can adjust what counts as an action, change the routing, and bring a new owner up to speed without rebuilding it.
Deliverable
A runbook with the source-of-truth rule, extraction, routing, decision hand-off, and AI limits the team owns.
The action items and decisions come out of the notes as a short, owned list — not buried in a transcript nobody re-reads.
Every action item has a proposed owner and due date before a person confirms it, so it actually moves instead of sitting in a doc.
Decisions route into your decision log as records, so the same thing stops being discussed again weeks later.
Follow-through stops depending on one person scanning the notes, because the extraction and routing run every time.
AI reads, extracts, classifies, and routes — and a person confirms every owner and decision, with the limit written down.
One notetaker is the source of truth and one runbook holds the rules, so the system survives a tool change or a new owner.
No. A notetaker records the meeting; this is the system on top that turns its output into owned action items and routed decisions. We don't sell or build a notetaker — we work with whatever you already use (Fireflies, Otter, Granola, Zoom AI, whatever it is) and build the rule, the extraction, and the routing that the notetaker does not do.
Decision Capture is the decision log — the single place decisions are recorded. This system is the bridge: it reads your meeting notes, finds the decisions, and routes them into that log. If you don't have a decision log yet, we will set the hand-off up so it is ready for one. They sit next to each other on purpose.
Meeting Reduction decides which meetings should exist at all. This system assumes the meetings happen and are recorded, and turns the notes into action. One is about the calendar; this is about what comes out of the meetings you keep.
No. AI proposes an owner and a due date for each action item and drafts the decision records, but a person confirms every owner and every decision before it is treated as real. The line between what AI proposes and what a person confirms is written into the system.
We design the rule and the routing to fit the tools you already use, and name current tools as examples — but we do not make integration promises on this page, because tools change. In the diagnostic we confirm what is actually possible with your specific stack, and we will tell you honestly where a clean connection exists and where it doesn't.
It does not include building or selling an AI notetaker, a meeting-reduction engagement (that is a separate page), a decision log itself (we feed yours, or set up the hand-off), or a project-management tool. It also does not include integration guarantees with a specific vendor. If what you need is one of those, we will tell you and point you to the right work.
Meetings multiply because there is no rule for which conversations need to be live. We classify each recurring meeting and move status to a written async cad…
Read moreOperating cadenceDecisions get made in Slack and email, never written down, then reopened weeks later. We define what counts as a decision and where the log lives — and AI dr…
Read moreKnowledge ops, not a platformWe help your B2B team organise the knowledge you already have, make it findable, assign clear owners, and keep it updated — using the tools your team already…
Read moreBook a 30-minute call to look at what happens to your meeting notes today. We will talk through which notetaker is the source of truth, what counts as an action versus a decision, and how AI can extract and route them. If the real issue is too many meetings or no decision log, we will tell you and point you to the work that fits.
Book a 30-min call