Why trial for a term first

A tool's marketing page and how it actually performs in the classroom are often two different things. The demo runs on hand-picked examples, but your students' answers, your teachers' habits and your network conditions are all different. A term-long trial lets the tool run under real conditions for long enough to expose the problems the marketing page never shows.

In its clarification of February 2026, the EDB reminded schools to assess whether packaged solutions really suit their own needs. Trialling on a small scale, comparing the evidence, then deciding whether to buy for the whole school is what that reminder looks like as concrete steps. The funding runs until 31 August 2028, so there is time to try before you buy.

Define the scope: one subject, one term

The first step in a trial is to narrow the scope. Pick one subject panel and one term, rather than putting the whole school on it at once. A small scope means fewer variables, which makes the conclusions clear and easy to attribute. When you choose the subject, favour the use where the pain is most obvious, such as Chinese and English writing where the marking load is heaviest. That way, even if the results are only average, you can still see whether the tool helped where it was needed most.

At the same time, be clear about the aim of the trial: what is the question this term needs to answer? Is it whether marking really saves time? Whether students get feedback that is faster and more specific? Write the question down, so that at the end of term you know what standard to judge success by.

Set a baseline first

Without a baseline, there is no way to claim 'improvement'. Before the trial starts, record where things stand now: how long a teacher spends marking one piece of work on average, how much feedback a cycle produces, how long students wait to get their comments. These numbers do not need to be precise, but they must be recorded before the trial, or all you will have at the end of term is an impression with nothing to compare it against.

The baseline is also material for your later reports. The part of the interim and final reports that takes the least effort to write is the before-and-after data, and all of it starts with that first set of numbers taken before the trial.

Collect feedback and data every two weeks

During the trial, collect two kinds of information as you go rather than waiting until the end of term. The first is teacher feedback: a short chat or questionnaire every two weeks, asking what works smoothly, what gets stuck, and whether the AI's judgement can be trusted. The second is usage data: how many logins, how much work processed, and how often teachers change the AI's suggestions. The first is how people feel, the second is what actually happened, and you need both to get the full picture.

The advantage of collecting regularly is that problems surface mid-term and can be raised with the supplier, instead of piling up until everything goes wrong at once at the end. If a feature is consistently inaccurate, write it down, because that is a bargaining chip when you come to renew.

Let evidence decide: expand, adjust or stop

At the end of term, lay out the feedback and the data against the question you set at the start, and the conclusion usually falls into one of three. First, the evidence is positive and the tool is worth expanding to more subject panels. Second, the direction is right but there are clear weak spots, so you go back to the supplier to adjust, or run another term. Third, both the time saved and the results fall short of what you hoped, so you stop honestly and keep the funding for something more worthwhile.

Whichever it is, the decision rests on your own school's evidence, not on the supplier's marketing or another school's word of mouth. The trial record itself later becomes direct proof, in your report, that the school assessed the tool carefully.

Ask the supplier to support a real trial

A meaningful trial needs the supplier to match it with equal cooperation. Before you sign, it is worth putting the following requests directly, because a supplier who cannot answer or will not cooperate has told you something already:

  • Can one subject panel trial the tool free for a full term, rather than a demo account that lasts a week or two?
  • Does the trial version have the core features of the full version, or are some held back on purpose?
  • How is student data handled during the trial? How soon is it deleted afterwards? Can you provide proof of deletion?
  • Can you provide a usage report so the school can check the number of logins and the volume processed for itself?
  • If we decide to expand to more subjects after the trial, are the fees and terms set out in advance, so there is no last-minute price rise?
  • If we do not continue after the trial ends, can the data and records created during it be exported?