Taizen
Get started

Quickstart

Connect a call source and your CRM, then ship your first agent — about 15 minutes.

By the end of this you'll have a working agent that delivers something useful into Slack every weekday morning, built from your own calls and deals.

Connect a call source

Go to Settings → Integrations. Find your call provider and click Connect.

Most connect in one click over OAuth. A few ask you to create an API key in the provider first — their tile says which, and there are setup guides for the common ones, including Gong and Genesys Cloud.

Whichever it is, give it read access to calls and their transcripts. Transcripts are the part people miss.

Taizen validates the connection, then begins importing calls and transcripts. The first sync backfills history, so it can run for a while; you don't have to wait for it to finish before moving on.

Not sure whether your provider is supported? Check the integrations catalog. There are 90+ integrations available, and you can request one if yours isn't connected yet.

Connect your CRM

Same screen. Connect Salesforce or HubSpot.

This is what gives agents deal context — stage, amount, owner, close date, and the loss reason that a win/loss interview later gets to interrogate.

Connect Slack

Connect Slack too. It's the most common delivery destination: an agent that writes a brilliant brief nobody reads has not done its job.

Note the channel you want output to land in. A private channel with two or three people is a better place to start than #sales.

Start from a template

Go to Agents → New Agent. You can describe the agent in your own words, or pick from the template library underneath — grouped by team: Sales, Sales Enablement, RevOps, PMM, Competitive Intelligence (templates).

Good first agents, because they produce something visible within a day:

  • Meeting Prep — runs early on weekdays, scans the calendar for external calls, pulls email and call history plus company research, delivers a brief.
  • Auto CRM Update — runs after every call, writes the deal fields, logs next steps, updates the close date with the reason captured on the call.
  • Daily activity summary — a digest of what happened across the team yesterday.

Either way the agent builder opens: a chat on the left, the agent spec filling in on the right.

Tell the builder what you want changed

The builder is a chat. On the right, the Agent spec panel fills in as you talk — trigger, workflow, delivery — and you can edit any field directly.

Say what's different about your team in plain language:

Only include calls owned by the EMEA team, and deliver to #emea-deals at 7am UK time.

Test it before you ship it

Ask the builder to run it: "test this now". A test run appears in the spec panel with the output, how long it took, and what it cost. Test runs don't deliver anywhere, so you can iterate without spamming a channel.

Read the output. If it's wrong, say why — "it's including internal syncs, only external calls" — and test again.

Save and watch the first real run

Say "save it". The agent appears in Agents with its trigger and a status dot.

After the first scheduled run, open the agent and check the Timeline. Every run records its inputs, steps, output, duration and cost. If something looks off, runs and debugging explains what each status means.

What to do next

  • Deliver somewhere quiet first. Point a new agent at a private channel until you trust its output, then repoint it at the channel your reps read (validation and review).
  • Try a roleplay. Go to Roleplay, pick a bot, and run a five-minute voice session so you know what your reps will experience. See run a session.
  • Pick a scorecard. Scoring is what makes practice comparable to real calls. Start from MEDDPICC and adjust — see the scorecard builder.
  • Invite the team. Roles and invites.

On this page