🧪 Practice

Identification exercise: given the profile of a fictional company, find three repetitive tasks that add up to ~10 hours per week.

Seeing opportunities is the most valuable skill in the cycle. Without raw material (the right task), there’s no point in automating anything.

Frequency × duration × pain; “10-hour” task; completion criteria; annotated answer key.

Measure the current process time and cost, and calculate the return on automating it using a spreadsheet template.

ROI is the number that justifies the project to you and sells it to the client. Without it, it’s just “guesswork”.

Cost/hour; hours/month; annual savings; payback; ROI formula.

Map the current process (as-is) step by step and the automated version (to-be), marking what the AI takes on.

The diagram reveals bottlenecks and shows exactly where automation fits—it becomes the project scope.

Manual × automatable steps; decision point; simple notation with text/arrows.

Starting from the to-be, specify the trigger, tools, AI model, logic, and actions in a one-page blueprint.

The blueprint is the bridge between idea and implementation. Those who design well build quickly and avoid rework.

Trigger; nodes; model; conditional logic; final action; failure points.

Actually build a 3-node workflow: a trigger, an AI call, and an output action.

The first end-to-end automation unlocks everything: after that, you just repeat the pattern with variations.

Trigger; prompt in the workflow; response parsing; action; test with real data.

Turn the automation you built into a 1-page proposal: problem, solution, ROI, price, and next steps.

Building without knowing how to sell leaves money on the table. The proposal completes the material's journey by turning it into a commercial deliverable.

Value-based pricing; proposal structure; ROI anchoring; clear CTA.

Prompt that takes a process description and returns steps, estimated times, and automation candidates.

Speeds up the audit phase from hours to minutes and standardizes the output.

Role + context; output format; variables {{processo}} and {{setor}}.

Prompt that lists and ranks automation opportunities by effort × impact.

Helps you choose where to start—the most common mistake is automating what doesn't matter.

Effort × impact matrix; quick wins; cut criteria.

Prompt that takes the objective and returns suggested triggers, tools, model, logic, and actions.

Gets the blank page out of the way and gives you an architecture draft to refine.

Explicit constraints; available stack; output in blocks.

Prompt that generates the conditional logic / pseudo-config for an automation node based on the business rule.

Translates a rule in text into implementable logic, reducing transcription errors.

If/else; field mapping; exception handling.

Prompt that extracts fields from free-form text and returns valid JSON with a fixed schema.

Reliable JSON is what connects AI to the rest of the workflow; without it, the automation breaks.

Schema; “respond with JSON only”; null for missing values; validation.

Prompt that builds a complete sales proposal from the problem, solution, and ROI.

Generates a sellable first version in seconds that you only need to adjust.

Tone; structure; value anchoring; {{cliente}} and {{preço}} variables.

The parts every well-made skill has: input, instructions, rules, output format, and examples.

Understanding the anatomy lets you create new skills in minutes using the same skeleton.

Input; role; rules; output; few-shot; versioning.

A skill that reads a message, classifies it into categories, and decides where to send or route it.

Triage is one of the highest-ROI automations and one of the most requested by clients.

Closed categories; confidence; "human" fallback; JSON output.

A skill that receives a long document and returns a summary + structured fields.

Contracts, notes, and PDFs are an endless source of repetitive manual work.

Chunking; target fields; source citation; output schema.

A skill that receives tabular data and generates a narrative report with highlights and recommendations.

Turns raw numbers into a communicable insight—a deliverable the client immediately sees value in.

Key metrics; period-over-period change; executive tone; fixed sections.

A skill that answers questions using only a provided knowledge base (RAG pattern).

Support grounded in the source prevents hallucinations and is the most marketable use case today.

Context recovered; "answer only from the knowledge base"; cite a passage; "I don't know".

How to take a template and adapt it to a new client without breaking it, while keeping version history.

Reuse is what scales an automation business: you never start from scratch.

Customization points; changelog; v1→v2; regression tests.