Choosing a tool How-tos Some effort180 min
Run a fair two-week AI tool trial in your department
A fortnight plan for heads of department: pick one job and two AI tools, use three real lessons, time your own edits, then decide at a meeting.
What a fair trial needs
A fair trial is one job, two tools, the same three lessons and a fortnight to do it in. Gather this before you start:
- Your scheme of work for the next two weeks.
- The report cards for both tools.
- One shared document holding the prompt, word for word.
- A phone timer.
- A log sheet with one row per draft: tool, lesson, minutes to edit, what you changed, usable yes or no.
Allow about 180 minutes per teacher across the fortnight: 20 reading the cards, 25 setting up, 120 on six drafts and 15 in the meeting. Two or three colleagues can each run all six drafts, so the log does not rest on one person’s habits.
Pick one job and two tools
Pick one job your department does every week, such as a 50-minute lesson with slides and a worksheet, and two tools that claim to do it. Read both report cards first: curriculum, what it makes, free plan, price per year, data and privacy, UK or US fit. Each card is filled from the tool’s own website and dated, and how we assess explains the method.
An example pair is Chalkie and Aila. Chalkie’s free plan gives up to 5 full-length lessons or activity sheets every 7 days, so three lessons fit. Aila is free for teachers and built on Oak’s curriculum content. Its site says the slides carry no images or diagrams. Chalkie publishes a data processing agreement. Aila’s privacy statement mentions none. Ask whoever owns your school’s data policy before anyone types.
Choose three real lessons
Choose three lessons you are due to teach in the fortnight, for one year group, on three different topics. A real lesson has a real objective in your scheme of work, so gaps show up. A made-up test lesson hides them.
Mix the three. Take one heavy on knowledge, one built on practical work or discussion, and one that follows on from a lesson already taught. Copy the objective and any required vocabulary from the scheme into the prompt. Write the prompt once and save it in the shared document. Paste the same wording into both tools.
Put no pupil names or personal data into either tool. Follow your school’s AI and data policy. This is not legal advice.
Run the trial in this order
Run the trial in this fixed order, and do not skip the checks.
- Read both report cards (20 minutes).
- Write the job in one sentence and pick the three lessons (10 minutes).
- Write the shared prompt and a five-line checklist (15 minutes). The lines are: matches the objective, covers the content in the scheme, suits the year group, facts are correct, nothing the department would not teach.
- Generate lesson 1 in both tools with the same prompt. Alternate which tool goes first.
- For each draft, start the timer, edit until you would teach it, stop the timer and log the minutes.
- Mark the draft against the checklist and the scheme of work, yes or no on each line.
- Repeat for lessons 2 and 3 over the fortnight. Teach the better version of each and log which.
- Hold the meeting in week two.
Log edit time yourself
Edit time is the minutes between opening a draft and being ready to teach it, and only your own timer can tell you what it was. Tools make their own claims. Chalkie’s site says lessons are “ready in minutes”. Aila’s says it builds a lesson “in minutes”. Those are claims from the tools, not measurements from us, and not yours either.
Timer off when the draft is ready to teach. Do not count the teaching. Log each change in a few words, such as “replaced two examples” or “cut the plenary”. Note any factual error you find. The teacher checks AI output before it reaches pupils, trial or not. For a method, see check an AI-made lesson resource in 30 minutes.
When it goes wrong, and mistakes to avoid
If a draft is unusable, log it as unusable and move on. You may re-prompt once, with the same wording in both tools, and log the extra minutes. If a week gets away from you, do the remaining drafts in 20-minute slots. Finish all six before the meeting, even if the teaching notes are thin.
Three mistakes spoil a trial:
- Changing the prompt for one tool, so the drafts are no longer comparable.
- Judging how the slides look instead of checking them against the scheme of work.
- Letting one enthusiast run it alone, so the log shows one person’s taste.
A tool that needs long edits on all three lessons has told you something. Write it down.
Decide at the department meeting and make it easier next time
Decide in a 15-minute meeting with the log on the table. Each person reads out total edit minutes per tool and the number of checklist lines failed. Three outcomes are open: adopt one tool, keep both for different jobs, or adopt neither. Write the decision down with the date and the date on each report card, because prices and privacy terms change.
For two cards side by side, see our Planit Teachers vs Chalkie comparison.
Your numbers cover three lessons and a few teachers. They answer the question for your department, not for every school. Keep the prompt document and log sheet as a template. Next time, change only the tools.
Frequently asked questions
How long should a trial of an AI tool last?
Two weeks is enough for three real lessons in two tools. It gives each teacher six drafts to edit and a week to fit the meeting in. About 180 minutes per teacher covers reading, setup, drafts and the meeting.
How many AI tools should we compare at once?
Two. Three tools means nine drafts per teacher, and the log gets harder to read. Run a second trial later if you want to test a third.
Can we use pupils' names or work in the trial?
Not unless your school has approved the tool for that use. Lesson planning needs no pupil data, so keep the trial to that. Follow your school's AI and data policy. This is not legal advice.
Do we need to pay for either tool?
Not necessarily. Chalkie's free plan gives up to 5 full-length lessons or activity sheets every 7 days, and Aila is free for teachers. Check the free plan line on each report card, because limits can change.
What if one tool is faster but the other fits our scheme better?
Go back to the log. Fewer edit minutes matter little if the draft failed the checklist lines on the objective and content. Many departments would take the slower draft that matches the scheme.