Start here · Lesson 1.5
Choosing an AI Tool for What to Automate With AI
Search for tools to help with what to automate with AI and you will find endless ranked lists. We have not run a head to head test of products, so we will not give you a ranking. A list of winners is also the wrong place to start. The job you choose decides which kind of tool you need, and your own test on your own task beats any list.
This guide gives you a scorecard for judging any tool on your real work. It sits alongside Lesson 1.5, Find the jobs to hand to AI, which helps you pick the job first.
Pick the job first, then choose the simplest tool that can take your brief, and score it on your own task, not on a demo.
Start with the shape of the job
Tools fall into rough kinds. What matters is what your job needs.
| Your job needs | Kind of tool that fits | Example of the job |
|---|---|---|
| Text in, text out, checked by you | A general chat assistant | Draft a weekly update from notes |
| Working with your own files | A tool that can read and edit files on your computer | Reformat a folder of documents |
| The same steps on a schedule | A workflow or automation tool | Send a standard message each Monday |
| Specialist output such as video or audio | A purpose built tool for that medium | Cut a clip from a recording |
Most small business first jobs are in the first row. A chat assistant with a good brief is often all you need, which means you may not need to buy anything new. Start with what you already have.
What we can say about named tools
We are careful here, because features change. Two points are documented by the vendors themselves.
- Anthropic's prompting documentation advises being clear and direct, giving context about the audience and goal, and using a few well chosen examples to steer format, tone and structure. OpenAI also publishes a prompt engineering guide for its models; read it for its own advice. A clear brief with context and an example is likely to carry over between tools, but test that yourself.
- Claude Code is described in its documentation as an agentic coding tool that reads your codebase, edits files, runs commands and integrates with development tools. If your job involves files on your own computer, that kind of tool is relevant. Check the current docs before relying on any feature.
Beyond that, do not take a blog post's word for what a tool does. Read the product's own documentation.
The scorecard
Run the same task through each tool you are considering. Score each line from 1 to 5, then add up. Use the same task, the same brief and the same input for each tool, so the comparison is fair.
- Time to first usable draftHow long from opening the tool to a result you would edit rather than discard?
- Quality against your examplePut the output beside your best past example. How close is it?
- Fits your inputsCan it take what you actually have, such as pasted text, a spreadsheet or a folder of files?
- Easy to checkCan you see what it did, so you can judge it quickly?
- RepeatableCan you save the brief and run it again next week without starting over?
- Privacy fitDo the tool's terms suit the data you will put in? Read them.
- Cost for your volumeLook at the current pricing page for your expected use.
A total near the top does not mean a tool is best in general. It means it is best for this job, for you.
How to run a fair test
- Choose a job you have already done so you know the right answer.
- Write one brief and use it unchanged in each tool. If you change the brief between tools, you are testing your writing, not the tools.
- Use the same real input each time.
- Time yourself from start to a draft you would send after light edits.
- Score the output the same day, side by side with your own version, ideally with the tool names hidden from yourself if you can.
- Repeat once on a second input. A single run can mislead.
If a tool needs a long setup before the first run, count that. It is part of the cost.
Red flags when reading tool pages
- Claims of specific time savings with no method shown.
- Testimonials with no names or sources.
- A feature list longer than the explanation of what the tool is for.
- No clear information on where your data goes.
- Promises that it works with no checking by you. AI tools can make mistakes, and the vendors state this on their own pages.
A short illustration
This is an example, not a test we ran. Say you run a small tutoring business and your first job is a weekly progress note to parents, built from your lesson notes. You try two tools you already have access to. Tool A needs no setup and gives a draft in a minute but sounds stiff. Tool B lets you save the brief, so next week is quicker, but takes longer the first time. On your scorecard A scores higher on time to first draft and B on repeatability. The scorecard does not pick for you, but it shows what you are trading, so you can decide with your eyes open. The numbers are invented, and your result will differ.
Keep it simple
Few jobs need a new tool. Before buying anything, try your current assistant with a better brief. For help with that, read How to Brief AI Like a Senior Hire. Add a tool only when the scorecard shows a gap, such as files, schedules or a specialist medium.
Questions people ask
Which AI tool is best?
There is no single answer. It depends on the job, your inputs, your data rules and your budget. Use the scorecard on your own task.
Do I need to pay for a tool to start?
Not necessarily. Check what you already have access to first, and look at each vendor's current pricing page for the facts.
Can I reuse the same brief in different tools?
Usually yes, since the advice on clear instructions, context and examples is similar across vendors. Expect small adjustments.
How often should I retest?
Tools change, so retest when you notice a problem or when a tool announces a change relevant to your job.
Where to go next
Choose the job, write the brief, then run the scorecard on two tools at most. Lesson 1.5 covers choosing the job and writing the brief: Find the jobs to hand to AI. For the weekly routine, see the Find framework.