TL;DR

A consultant's most valuable asset is spoken, and it scatters: findings live in call recordings, decks, and memory across a dozen engagements, and it walks out the door when a project ends. A knowledge system fixes that by giving every client conversation one structured, connected home.
You can build that system in Tana in an afternoon: client calls are captured without a bot, findings file as typed items you approve, one record per engagement stays current, and what you learn on one client is queryable across all of them.
After setup, it runs itself. Each call is processed automatically, every AI change arrives as a proposal you approve, and re-running extraction updates the existing record instead of stacking another summary.
The payoff compounds. The tenth engagement starts from everything the first nine taught you, and a new hire inherits the firm's expertise as a queryable record rather than a shelf of old decks.

Consultants sell expertise, and expertise is mostly spoken. It surfaces in discovery calls, stakeholder interviews, and steering meetings, then it scatters, into a recording nobody rewatches, a deck in a client folder, and the consultant's own head. Do that for two years and the firm's knowledge is real but unfindable. This guide is the fix: the concrete steps to build a knowledge system in Tana, so that every client conversation files itself into one connected, reusable record. If instead you want the ranked comparison of which AI tools consultants reach for, start with Best AI tools for consultant deliverables 2026; this is the setup companion. For the broader idea of a shared, current record that AI can work from, see What is context engineering for AI agents.

What makes it a knowledge system, not a pile of notes

Four properties separate a knowledge system from a folder of engagement files, and they are the checklist for everything below:

Uniform capture. Every client conversation is recorded and transcribed the same way, whichever tool the call happens in.
Structured findings. Insights come out in one shape, with the same fields, so they can be compared, counted, and filtered across engagements.
One connected record. New conversations update and extend what is already there, per client and across clients, instead of adding another standalone summary.
Queryable by you and your AI. You, your team, and your AI agents can ask the record questions and get answers grounded in specific client conversations.

Tana covers all four. Here is the setup.

How to build a consultant knowledge system in Tana, step by step

Six steps. The first three get a single client conversation flowing into the system; the last three make the conversations compound into reusable expertise.

Step 1: capture every client conversation without a bot

The system starts at the call, because anything you reconstruct afterward has already lost detail. In Tana there are two capture paths, and together they cover every conversation you run:

Calls you host: run them as a Tana meeting. Tana transcribes in real time with speaker identification, and when the client shares their screen, it captures screenshots of what they were looking at, each with an AI-generated description.
Calls on Zoom, Teams, or Meet: the desktop app captures the external call from your side, with no bot joining the meeting. You take part normally in the client's tool of choice; Tana transcribes the audio and captures shared-screen content in the background.

The no-bot path matters for consulting specifically. Clients are often wary of a recording bot appearing in their meeting, and many corporate policies forbid it. Capturing from your side removes that friction while still getting the conversation into the system. Capture is on by default, and you can pause it mid-call with off-the-record mode for anything the client does not want recorded.

Step 2: capture decisions and findings live, as typed items

Raw transcripts are storage, not knowledge. The turning point is when a stretch of conversation becomes a structured finding. In Tana you do not have to wait for the call to end: the Capture control takes a window of discussion and turns it into a typed item on the spot, a Decision, a Task, or a custom type you define. When a client states a constraint or a stakeholder reveals the real problem, you capture it as a finding while it is still in the room.

You do not have to catch everything live, though. When the call ends, Tana processes the meeting on its own: a summary, the decisions, and the action items, extracted as typed items and assigned to the right person. Live capture is for the moments you want pinned exactly; the post-call pass catches the rest.

Step 3: give findings a defined shape

This is the step that turns notes into a dataset. Ask chat to create a Finding type for you, in plain language: "a Finding type with fields for the client, the engagement, the affected area, severity, and status." Tana creates the type with fields and a workflow, so your findings get a board from day one, from fresh observation to validated recommendation.

Two configuration moves make the type pull its weight across every future engagement:

AI instructions on the type tell the AI how to fill it: "always link the client meeting the finding came from and quote the strongest supporting line." Every future extraction follows the house rules, so findings are consistent whether you or a colleague ran the call.
Skills attached to the type put actions where the finding lives. Attach a "draft recommendation" skill or a file-to-tracker skill, and any finding can trigger it from its own menu when it is ready to act on.

Because findings file into your team's space, they are shared the moment you accept the proposal. The consultant who ran the call and the partner who was not there see the same structured record, not a private write-up waiting to be circulated.

Step 4: keep one engagement record current

Here is where most consulting stacks fail. A notes tool gives you one more summary per client call, forever, and the tenth summary does not know about the first nine. A system needs the opposite: new conversations should update what you already know about the engagement.

Tana works that way by default. Re-running extraction updates existing outcomes rather than creating duplicates, and if you pin a document to the client meeting, say the engagement record or the open workstream the session was about, extraction treats it as the subject and prefers updating it over spawning a parallel one. The practical habit: before a session, pin the engagement doc the call concerns. After the call, that record is current, and it carries the evidence trail of every conversation that touched it. Week six revises the findings from week one instead of stacking a sixth summary beside them.

Step 5: turn the record into client deliverables

A knowledge system earns its keep when the record produces the deliverable, not just when it stores information. Once the findings live in Tana, the output is a step away, not a retyping job:

A structured summary or findings document, generated from the transcript or from chat, that you edit inline and refine.
A slide deck as an artifact: a hero slide, takeaway cards, themed sections, action items, and a closing slide, drafted from the engagement record and shaped by you.
A before-and-after customer journey map when the deliverable is about how an experience changes.

Artifacts are editable inline and shared with the client via a link, so the deck stays connected to the evidence behind it rather than becoming a detached file. The ranked view of how this compares to drafting in ChatGPT, Claude, or Copilot is in Best AI tools for consultant deliverables 2026; the point here is that the deliverable is generated from the same record the rest of the system maintains.

Step 6: reuse knowledge across engagements, and connect it to your tools

The single-client record is useful. The cross-client record is the moat. Because every finding is a structured item in one space, you can ask chat questions that span your whole practice: "which manufacturing clients raised the same supply-chain risk this year?" or "what did we recommend the last three times a client had this org structure?" The AI answers from what was actually said, with the source sessions behind it, so a pattern you saw across clients becomes an asset you can point to in the next pitch.

Two extensions are worth setting up once the basics run:

A scheduled agent that briefs you before each client call by pulling what you decided last time and any open items, so you walk in current instead of re-reading the file. Describe it in a sentence and put it on a schedule; a weekly digest of new findings works the same way.
A connection to your other tools. Follow-up work files into the trackers and apps you already run on, including Linear, Jira, Slack, and HubSpot among others, as proposals. And through Tana's MCP server, a coding or research agent such as Claude Code can pull the relevant findings while it works and sync the result back, so the knowledge is available wherever you do the work.

Where a general chatbot fits

Plenty of consultants start by pasting a transcript into ChatGPT or Claude, and for a one-off analysis of a single call it does the job well. The limitation is structural, not a missing feature: a chatbot's output is a reply, and a knowledge system's output is a record. The analysis lands in a chat thread, in whatever shape the prompt produced, and connecting it to last month's calls, your other clients, and your deliverables is work that stays with you. That is fine for thinking through one conversation. A practice running many engagements needs the calls, the findings, and the follow-through in one connected place that stays current, which is the system the six steps above set up. For how the general assistants stack up on consulting deliverables specifically, see Best AI tools for consultant deliverables 2026.

What you have when it is running

After setup, the marginal cost of capturing knowledge drops to roughly zero. You run the client call; Tana captures it, extracts the findings with the evidence to support them, files them as shared structured items, and folds them into the record of everything the engagement, and the firm, has heard before. You query patterns across clients instead of trawling old decks. You generate the deliverable from the record instead of rebuilding it. And when a consultant leaves, the expertise stays, because it lives in a system rather than in their head. That is the difference between doing good consulting work and compounding it.

Frequently asked questions

What is a consultant knowledge system?

It is one connected, structured record of everything your client conversations produce: findings, decisions, and recommendations, captured the same way every time and kept current as engagements progress. The point is that expertise stops scattering across recordings, decks, and memory, and becomes a queryable asset the whole practice can reuse. In Tana you build it by capturing every client call without a bot, filing findings as typed items you approve, and keeping one record per engagement that updates itself.

How do I build a knowledge base for my consulting practice?

Six steps, doable in an afternoon in Tana: capture client calls natively or via bot-free external capture for Zoom, Teams, and Meet; use the Capture control to pin decisions and findings live; ask chat to create a Finding type so insights file in one structured shape; pin the engagement document to each recurring client meeting so extraction updates the existing record; generate deliverables from that record; and query findings across every client from chat. Each AI change arrives as a proposal you approve, so the system stays automated and reviewed at once.

Can Tana capture client calls that happen on Zoom, Teams, or Meet?

Yes. The Tana desktop app captures external Zoom, Teams, and Meet calls from your side, with no bot joining the meeting, which matters when a client is wary of recording bots or their policy forbids them. You take part in the call normally; Tana transcribes the audio, captures shared-screen content in the background, and processes the session into structured findings when it ends, so every conversation enters the knowledge system the same way regardless of the client's tool.

How is this different from just using ChatGPT or Claude for consulting?

ChatGPT and Claude draft well from what you paste in, and for a single call that is enough. What they do not give you is a record: the output is a reply in a chat thread, and connecting it to your other clients, past engagements, and deliverables stays your job. Tana is built as the system layer. It captures the call itself, files structured findings the whole practice shares, keeps one engagement record current, and lets you query patterns across clients. The tool comparison is in Best AI tools for consultant deliverables 2026.

How do consultants keep client knowledge from walking out the door?

Put the knowledge in a system, not in people. When every client conversation files into one shared, structured record, the expertise survives a consultant leaving or a project ending, because it lives in findings a colleague can query rather than in a recording nobody reopens. In Tana, capture and extraction are automatic, findings are typed and linked to the calls they came from, and the record stays current on its own, so the firm's knowledge compounds instead of resetting with each engagement.

Can AI generate consulting deliverables from my client meetings?

Yes. Once the conversations are in Tana, you can generate a structured findings document from a transcript or from chat, and produce a slide deck artifact with a hero slide, takeaway cards, action items, and a closing, drafted from the engagement record and shaped by you, then shared with the client via a link. The deliverable stays connected to the evidence behind it. See Best AI tools for consultant deliverables 2026 for how that compares to drafting in a general assistant.