Spend 90 minutes writing down everything you or your team does in a typical week. For each task, note: how long it takes, how often it happens, and whether it involves words, research, or data (the three things AI handles best). Don't edit or filter yet - just list.

Go back through your list and score each task on two dimensions: time cost (how many hours per week total?) and AI suitability (is it repetitive? does it involve writing or research? is the quality bar achievable?). Your top 5 scores become your target list for the month.

Pick the single highest-scoring task from your list and find the right tool for it. Don't research tools for all five tasks yet - that leads to tool overwhelm and paralysis. Create accounts for 2-3 candidate tools and set them up (5 minutes each). Most have free trials.

One sentence. 'By day 30, I want to be saving [X hours per week] on [specific task], with output quality that is [good enough to use after a light review / good enough to send directly / better than what I was producing manually].' Pin it somewhere you'll see it. This is your measurement benchmark.

Take your top-ranked friction task and do it with AI every time it comes up this week. Don't wait for the 'right' prompt - write a reasonable brief and see what comes back. Your first prompts will be rough. That's expected and useful: you'll learn faster from five imperfect attempts than from one overthought one.

If you're not already doing this: use ChatGPT or Claude to draft your next 10 emails. Not just the complex ones - all of them. The goal is to see whether AI drafts are faster to edit than writing from scratch. For most people, they are, even when the first draft needs significant editing. This becomes your second repeatable use case.

Pick a task you weren't sure AI could help with and try it. Good candidates: summarising a long document you received, generating a first draft of a proposal or case study, creating a standard response template for a common client question, or researching a topic before a client meeting. Low stakes, but potentially high reward.

Write 3-5 sentences: which tasks produced useful AI output? Which didn't? What made the difference? Jot down the 2 prompts that worked best. You'll formalise these into templates next week.

Take the prompt that worked best in week 2 and formalise it. Write it out in full with clear [placeholder] sections for the parts that change each time. Save it somewhere you'll find it: a pinned note, a Google Doc, your browser bookmarks. Test it three times with real tasks. Adjust the wording until it produces consistent output.

If you're using ChatGPT: go to Settings and set up Custom Instructions. Write a short description of your business, your role, and your communication style - this context applies to every conversation. If you're using Claude: create a Project and add an instruction note. This step alone saves 5-10 minutes of re-briefing per session.

If you have a team: pick one tool and one use case, write a single paragraph of guidance on how to use it, and share it in a team meeting or brief message. Don't overcomplicate this - 'Here's how I've been using ChatGPT to draft client emails. Try it this week and tell me if it's useful' is enough. If you're solo: add one automation or integration (e.g. connect AI to your email client, or try a purpose-built tool for your sector).

Whether or not anyone has raised it, now is the right time to set clear guidelines on what can and can't go into AI tools. Write three sentences covering: which tools are approved, what data should never be inputted, and what review standard applies to AI outputs. Share this alongside your team rollout note. This prevents problems later.

Go back to your week 1 friction list. For each task you've been using AI on: how long does it take now vs before? Be honest about this - count only real time saved, not theoretical efficiency. If a task now takes 20 minutes instead of 45, that's 25 minutes saved per occurrence. Multiply by weekly frequency to get a weekly hours figure.

For each AI use case: are you happy with output quality after your review process? How much editing is typically required? Have there been any errors or near-misses that reached clients? Are you using a consistent review process, or reviewing differently each time? The goal isn't zero editing - it's a reliable, repeatable standard you're comfortable with.

Write a short note (half a page is fine) covering: the tools you're using, the 2-3 use cases that produced the best results, your best prompt templates, and your data guidelines. This is the start of your AI workflow documentation. It's also what you'll show to new team members, clients who ask, or a consultant helping you go deeper.

Decide: which use cases will you continue and build on? Which tasks from your original friction list haven't you tackled yet? Are any of your current workflows ready to be automated? What training or support would help your team go further? Write 3 specific goals for month 2 and put them in your calendar.