PRACTICAL PLAYBOOKS

Make Bots part
of your workday.

Clear briefs, useful results, and routines you can trust enough to review.

Getting started

Your first useful result

Start with a bounded job and evidence you can inspect.

  1. Choose an outcome

    Name the output, its audience, and what done looks like. A research note with five sources is easier to review than “do my research”.

  2. Give it sources and limits

    Attach supported text, CSV, JSON or Markdown files, or name authorized public pages. Say what must remain a draft.

  3. Review the work

    Check sources, calculations and incomplete steps. Use the work log to inspect tools and approve requested actions. Save the result before expanding the task.

Try this brief
Research this topic and return a concise, dated brief with five primary sources. Separate confirmed facts from uncertainty. Do not send or publish anything.
Review the bot recipe ↗

Sales

Build an outreach review list

Research first. Leave the final send with a human.

  1. Define the customer

    Provide industry, company size, geography, and a reason your offer helps.

  2. Verify fit

    Ask for sources and evidence for each account. Missing contact information should stay missing.

  3. Review drafts

    Read personalization and claims. The bot should leave messages unsent until you approve exact recipients and content.

Try this brief
Find five accounts that fit the customer profile I provide. Cite the evidence, mark unverified contacts, and prepare a short unsent outreach draft for each.
Review the bot recipe ↗

Product

Turn a CSV into a decision brief

Verify the data before drawing a conclusion.

  1. Describe the export

    State the date range, timezone, currency, units and meaning of each column.

  2. Check the calculation

    Ask the bot to use Python, flag duplicate rows and missing values, and show denominators.

  3. Separate facts and recommendations

    Keep observed changes distinct from their possible causes. Ask for a reproducible report saved to shared files.

Try this brief
Analyze the CSV I attach with Python. Check duplicates and missing data, calculate the metrics with explicit denominators, and save a short report with limitations.
Review the bot recipe ↗

Operations

Schedule a repeatable workflow

Turn a successful one-off job into a reviewed routine.

  1. Prove the job once

    Run the task with the needed tools and sign-ins. Confirm that the output is useful.

  2. Draft the routine

    Name the schedule, local timezone and time. A bot-created routine stays paused until you enable it.

  3. Enable and monitor

    Review Routines, turn it on, and check Activity for failures. An active plan, credits and valid tool sessions are required. Ask the bot to stop when inputs are missing.

Try this brief
Draft a weekday 09:00 Asia/Kolkata routine for this proven workflow. Keep it paused for my review and state its required inputs and tools.
Review the bot recipe ↗

Getting started

Teach a browser workflow

Record your steps, then review the routine they produce.

  1. Open the bot’s computer

    Sign in directly in the browser. Use Take over so the bot pauses while you work.

  2. Start teaching

    Choose Teach a workflow, name it, and carry out the actual navigation, clicks and field entries. Password values are excluded from recorded steps.

  3. Save and review

    Finish the teaching session, inspect the schedule and instructions, and enable the saved routine only when ready. New sign-ins and site changes can still need you.

Try this brief
Help me record this browser workflow as a routine. Explain the steps I should demonstrate, and keep the saved routine paused until I review it.
Review the bot recipe ↗

Operations

Put specialists in one thread

Give the team a shared outcome without duplicating the same job.

  1. Choose participants

    Create the bots you need. Add them to one conversation with Participants.

  2. Define handoffs

    Ask Chief to coordinate a specific specialist task, wait for the result, and produce one final brief.

  3. Check capacity

    Your plan sets how many jobs can run at once. Shared files, skills and workspace memories provide common context; private bot memories remain scoped.

Try this brief
Coordinate the specialists in this conversation. Define distinct responsibilities, delegate only the work needed, wait for their results, and return one consolidated brief with sources.
Review the bot recipe ↗

Getting started

Share a recipe, not your data

Reuse instructions while keeping sign-ins and history out of the export.

  1. Review the instructions

    Make sure the template contains no customer secrets or private information. Instructions themselves can be sensitive.

  2. Export from bot settings

    Export template downloads the job recipe. Sign-ins, messages, files and memories are excluded.

  3. Import and review

    Use Import template in the workspace, review the instructions, then create the bot. Imports start with supervised browser actions.

Try this brief
Help me turn this successful process into reusable instructions. Use placeholders for customer-specific details and keep credentials and private information out.
Review the bot recipe ↗

Getting started

Choose how bots share a browser

Keep existing per-bot sign-ins, or explicitly opt into a shared workspace browser.

  1. Review who has access

    Workspace members can view shared browser context. Add only people authorized for the connected accounts.

  2. Finish running work

    Finish or stop jobs and teaching sessions before changing scope. Existing sign-ins are not copied or deleted.

  3. Select a scope

    Open Workspace settings and choose per-bot or shared browsing. Sign in separately in the newly selected browser. Human takeover and browser action locks still apply.

Try this brief
Explain the difference between per-bot and shared-workspace browsing, including sign-ins, membership access, and when a human takeover pauses bot actions.
Review the bot recipe ↗