Partner With Us

← All lessons / AI in practice / Lesson 09 of 12

Interactive lesson

Choosing AI tools: agree what matters before you shop.

Most organisations pick an AI tool because someone tried it and liked it. Learn a simple, defensible way to choose: set must-haves, weight what matters, then compare.

You’ll learn to
  • Separate essential requirements from weighted preferences.
  • Compare fictional tools and explain your assumptions.

Start with the problem, not the product

The first question isn't “which AI tool?” but “what problem are we solving, and do we need AI for it?” A good template, a better form or a simple automation is often cheaper and more reliable.

When AI is the right answer, choose in three steps. Agree your must-haves: anything that fails one is out, however good it looks. Then agree how much each criterion matters (its weight). Only then look at products, and score them.

Buy off the shelf

Quick and cheap to start. You accept the supplier's choices about data, what the tool can do and price.

Add to what you have

AI tools built into software you already use. Often the easiest to approve and support.

Build your own

Fits exactly, but needs skills to build, test and maintain. Rarely the first step.

Interactive

Compare three tools with a scorecard

Three fictional tools for drafting and summarising. Compare their ratings below, then choose how important each criterion is. The office add-on fits existing systems well; the specialist scores higher on the quality of case-note summaries. Try different priorities while keeping the must-have enabled, then see what happens when you switch it off.

How the score works: 100 means a tool rated 5 out of 5 on everything you've said matters. Scores are rounded for display; tools with the same displayed top score are shown as tied. These ratings are teaching examples. For real tools, gather evidence using your own tasks before assigning ratings.

Key ideas to take away

  • Check whether you need AI at all. A clearer form or a template can solve the problem more cheaply and reliably.
  • Must-haves come first. A tool that fails on data protection is out, however cheap or easy it is.
  • Agree the weights before looking at products. Otherwise the favourite tool decides the criteria, not the other way round.
  • Plan the exit. Check you can export your data and stop paying without losing your work.
  • Pilot small, with real tasks. Run a short trial, with a fixed end date, with a few users and measure the result before rolling out.

Check your understanding

3 quick questions. Try as often as you like. Your progress stays on this device; answers clear when you reload.

Want to go deeper?Methods, checklists and standards. Optional.

Questions for any AI supplier

  • Will you sign a data processing agreement? Where is data stored and processed?
  • Is our data used to train or improve your models? Can we switch it off, in writing?
  • How do you handle security incidents, and how quickly will you tell us?
  • What happens to our data, prompts and outputs if we leave?
  • How do you tell customers when the underlying model changes?

Total cost, not sticker price

Count licences, setup, training people, checking outputs and switching costs. A free tool that needs every output rewritten can cost more than a paid one.

A weighted scorecard in one line

score(tool) = 100 × Σ weight_c × rating_c  /  (5 × Σ weight_c)        for criteria c

Ratings and weights are judgements, so record who made them and why. That record is what makes the choice defensible later.

Continue exploring

We've picked resources that explain these ideas from different angles. They are independent of Code With Impact.

Sources for this lesson

The tools, ratings and scorecard are fictional and original to Code With Impact. The lesson draws on the frameworks listed under Continue exploring; no text or diagrams from them are reproduced.

How we reference: we link and credit, we never copy other educators' text, diagrams or code. See our full Sources & credits register.