Agent Skills · Team adoption & AI practices

A practical example

Prepare a recurring report from approved exports, flag missing data and route the draft to a reviewer before distribution.

Reusable instructions, examples and scripts teach an agent how your team does a task.

Illustrative scenario — not a client result.

SKILL.md

---
name: campaign-review
description: "Analyse your Google Ads campaigns with GPT Astra"
---

Analyse the attached Google Ads exports. First validate columns, periods, currency, duplicates and conversion definitions. Show totals and formulas. Calculate ratios from sums, never by averaging ratios. Flag zero denominators and insufficient data. Produce: five numerical observations linked to source rows; separate hypotheses; missing checks; and three proposed tests with goal, scope and stopping criteria. Do not invent causes or data. Do not modify campaigns.

How to assess progress

• The example does not establish higher profitability: conversion value and margin are unknown.
• Do not mix click-through and view-through conversions without documented definitions.
• Keep the file, calculation and context behind every recommendation.

Our method

1. Prioritise

   Map tasks, time spent and the business KPI to improve.

2. Build

   Connect your data and tools into a tested workflow.

3. Enable

   Train your team with reusable prompts, Skills and clear review steps.

4. Improve

   Measure time saved, approved output and business outcomes against a baseline.

How to assess progress

Compare total handling time, failed runs, correction effort and data accuracy over equivalent cycles.

Observe repeat use, review quality, completion time and the ability to explain when AI should not be used.

What you receive

• Role-specific use cases and practical exercises
• A working guide covering review and escalation
• An adoption plan with owners and follow-up checks

Workshop format and delivery language are agreed beforehand. Tool certification and legal compliance certification are not included.