FREE STARTER LESSON · THREE PRACTICAL EXERCISES

Plan it. Compare it. Follow through.

Turn one recurring task into a brief, a fair tool trial or a reviewed follow-up. Read the lesson here, then practice with our free tools.

OPTIONAL NARRATED COMPANION

Watch the starter lesson.

A short visual lesson with synthetic narration and fictional examples. The video opens on Scrimba and currently requires a free Scrimba account. The full text and exercises below are available here without an account.

Watch on Scrimba ↗

Choose one useful result.

Allow about 15 minutes for an exercise. These are planning and review tools. A downloaded plan does not activate an agent or send a message. Use the fictional examples below before entering your own work.

01 · AGENT LAUNCH PLANNER

Give an agent a clear job.

Leave with: a brief that an implementer can test, including an owner and a stopping point.

  1. Open the planner and choose “Try a fictional example.”
  2. Review its eight fields: outcome, trigger, inputs, reviewer, approval, limits, success check and destination.
  3. Build the test plan and download it. Keep every test marked “not tested” until you run it against a working implementation.
Worked example and review check

Task: prepare a weekly brief from three approved public sources. The marketing lead reviews each claim. Publishing requires approval. Set a sample budget of $5 per run and one retry; these are planning limits, not enforced controls.

Done for this exercise: all eight fields are specific, five tests are listed, and the reviewer can explain how a future workflow would stop. Actual implementation must enforce these limits.

Build my agent brief ↗
02 · TOOL TRIAL LAB

Test the same task twice.

Leave with: a reasoned choice based on results you checked.

  1. Choose two tools and give each the same non-confidential input, task and required format.
  2. Time generation, fact-checking and corrections. Record setup, failures and fees in your evidence notes; the tool’s total excludes setup and fees.
  3. Use the Tool Trial Lab to record what happened and download your decision. Choose neither if both fail an essential requirement.
Fictional example and review check

Tool A takes 4 minutes to generate and 10 to correct: 14 total. Tool B takes 5 minutes to generate and 2 to correct: 7 total. This is an arithmetic illustration, not a product benchmark.

Done for this exercise: each tool has the same input, a saved output, actual total effort and a recorded limitation. Recheck prices with each provider before purchasing.

Run my comparison ↗
03 · QUOTE FOLLOW-UP DESK

Decide the next useful step.

Leave with: a reviewed next action for one quote, with a date and an owner.

  1. Open the public pilot and load its fictional demo.
  2. Inspect one quote. Check the amount, status, last contact and next action before using a suggested draft.
  3. Choose a reasonable next step and export a backup. Send any approved message separately using your normal channel.
Practice prompt and review check

A fictional client has a quote but has not replied. Draft a short check-in that asks whether they need clarification. Avoid implying a deadline or discount that was never agreed.

Done for this exercise: the draft is accurate, an owner and follow-up date are clear, and you have exported a backup. The pilot keeps records in that browser unless exported; it does not send messages automatically.

Try the fictional quote demo ↗

Read the full narration.

The written lesson is free and available without a video account. Download the transcript ↓

Three practical tasks

Three practical tasks help you plan, compare, and follow up.

Choose your result

Choose your result before choosing more software. An agent brief tells an implementer what should happen and where to stop. A fair comparison helps you decide which tool earns a place in your process. A follow-up decision keeps a quote visible without sending an automatic message. These lessons use fictional examples. Open the matching tools from the AI Superpower Launch Lab when you are ready to practice.

Define the job

The agent's job should be small enough to check. In the Agent Launch Planner, the outcome is a weekly research brief. The trigger specifies the intended time and frequency. The inputs restrict the work to approved sources. The destination is a private draft. Entering a schedule in this planner does not activate a recurring job. It records the requirement for a later implementation.

Set boundaries

The boundaries make the brief usable. The reviewer owns the decision to accept or revise the output. Approval is required before publishing or sending anything. The example limits give the implementer a spending ceiling and a retry rule; the planner itself does not enforce either. Quality criteria explain how the reviewer will check the result. Use role names and sample inputs, and leave passwords and confidential customer records out of the plan.

Use the planner

Build, inspect, and keep is the practice sequence. Build my test plan creates the document after all eight fields are complete. Inspect the cases: a normal request, missing input, a duplicate trigger, an approval boundary, and a provider failure. Every case starts as not tested. Save or download keeps the work: a saved draft remains in that browser, while a download gives you a separate file. Pause here and build your own brief. Your deliverable is the plan, not a running agent.

Set up a fair trial

A fair trial starts in the Tool Trial Lab. Choose one brief, such as turning fictional meeting notes into an action summary. Define the reference before seeing the answers: which decisions must appear, who owns each action, and which dates are actually present? Then use two candidates with the same input and requirements. If one tool gets extra context or more attempts, record that difference. The comparison should describe the work you actually did.

Count review work

This fictional comparison shows why drafting speed alone can mislead. Candidate A produces a draft sooner, but an invented deadline adds checking and correction. Candidate B takes longer to draft and less time to review. The total is seven minutes instead of fourteen for this made-up case. No real product was benchmarked here. In your trial, record actual times, errors and corrections. For repeated use, also consider setup time, usage charges and how easily the output can be exported.

Decide from evidence

Your decision applies to this task and the cases you tested. Keep a tool when its checked output meets the requirements at an acceptable cost. Revise when a specific input or instruction change may address the problem, then run the same cases again. Stop when errors or correction work outweigh the benefit. Download the report with the limitations included. Pause here and write your success criteria before trying another tool.

Understand the next decision

An outstanding quote needs a decision before it needs another message. A customer may be waiting for an answer about access, timing or scope. The free Quote Follow-up Desk helps organize that review. It stores records in the browser and prepares drafts; it does not contact customers, synchronize a CRM or collect payment. Start with its fictional demo so you can learn the controls without using private records.

Use the quote desk

Review one estimate all the way through. Explore the fictional demo and choose a record from the queue. Check contact preference before deciding whether outreach is appropriate. A quote's age is not permission to contact someone. Choose the next action, such as asking the estimator to confirm access requirements, and set a manual review date. Save the update so the reason is visible later. A review date is a prompt for a person to decide; it does not activate a notification or send a message.

Know what the records mean

These records help with operations. Quoted value describes potential work and does not establish future income. Marked won records your update; confirm payments in the appropriate accounting or payment system. A full backup preserves the browser's records for later restoration. Browser storage can be cleared, and a local workspace is not a shared team database. Pause here: select one fictional quote, record a next action, and identify where you would keep an approved backup.

Put it into practice

Choose one small next task. Practice for free in the AI Superpower Launch Lab and keep the brief, comparison or quote-review record you complete. Review the result: what is usable, what needs correction, and what still depends on a person? Get help when useful. A three-hundred-fifty-dollar working session can help with one agreed task; workflow setup starts at fifteen hundred dollars and requires a confirmed scope. The free tools and existing self-paced courses remain free. This lesson uses synthetic narration and fictional examples; it does not claim client results.

OPTIONAL HELP WITH YOUR REAL TASK

Bring the draft. We’ll work through the next step.

A $350 AI Working Session focuses on one task, your current process and a practical next step. Scoped workflow setup starts at $1,500. We confirm scope, timing and final price before payment. This lesson and its exercises remain free.

Explore the Launch Lab ↗ · Continue with free learning paths ↗