How to evaluate if an Edtech tool is ready for school rollout

When a school leader first hears about us, their first question is often “Do the teachers like it?” That's because we usually reach leaders after their teachers have already started using the product. Then, always, come the questions about data privacy, security and compliance, which is exactly the due diligence I'd expect.

Those two questions are the whole evaluation in miniature. The first asks whether a tool works for a teacher. The second asks whether a school can stand behind it. A tool is only ready for a school rollout when the answer to both is yes, and I think a lot of the rollouts that stall in secondary schools were only ever ready for the first.

There's no official test for this in England yet. The Education Policy Institute (EPI), a UCL research summary funded by the Economic and Social Research Council, and the DfE-funded EdTech Evidence Board have each said so since August 2025. This guide draws on what they, the DfE's AI product safety standards and the Education Endowment Foundation (EEF) agree on, and turns it into ten questions a headteacher, trust leader or head of department can put to any supplier before a pilot.

I've spent more than a decade building edtech and getting it adopted in schools, and I now build AI marking tools for secondary teachers, so I'm not a neutral observer. These are the questions my team holds itself to.

In brief

• An edtech tool is ready for school rollout when the people who carry the risk can approve it and teachers keep choosing to use it. They are two separate gates, and a pilot should test both.
• England has no single standard for judging whether edtech is effective. The EdTech Evidence Board, the DfE's AI product safety standards and the DfE's new certification consultation each cover part of the question.
• As proposed, the DfE's certification scheme would assess safety rather than effectiveness, and on the DfE's own timeline it is around two years away. Readiness stays the school's judgement.
• Get written answers on data, access, governance and how student work is used before any student data goes in. Then pilot in contrasting departments across at least one full assessment cycle, such as a mock series.

What makes an edtech tool ready for a school, not just a teacher?

A tool is ready for a teacher when one person chooses it and keeps using it. It's ready for a school when it also satisfies the people who buy it, carry the risk and need it used consistently across departments and, in a trust, across schools.

The EdTech Evidence Board's pilot report, published in September 2026, describes two levels of influence around edtech decision-making: leaders who often make the purchase, and teachers who then choose whether to use what's been bought. In a secondary school, the first gate is the headteacher, the trust's central team, the data protection officer and IT. The second is every head of department and class teacher who decides, one mock series at a time, whether the tool earns a place in their marking. The UCL summary found that teachers were seldom involved in purchases that affect their work, though it drew on 2022 to 2023 survey data, and the DfE's 2025 survey shows their involvement rising.

Why secondary rollouts stall at one gate or the other

The first gate fails when a tool can't be cleared to buy. In secondary schools, the share of leaders citing safeguarding and data concerns as a barrier to technology rose from 28% in 2023 to 52% in 2025, according to the DfE's Technology in Schools Survey as reported by Edurio. A tool that spreads through enthusiasts with no DPIA pack, no role-based access and no clear answer on how student work is used will stall at the data protection officer's desk.

The second gate fails when a tool is bought but not used. An assessment of the edtech market written for the DfE in June 2026 drew on usage records from one large trust, which it notes are not necessarily representative of the wider sector. It points out that a school can hold an active contract for a tool that sees little use. EdTech Innovation Hub's coverage of the report quotes a trust technology leader on a costly product teachers were assumed to be using: “when you look at the usage, they're not.” The Evidence Board found a related gap. Its fourth criterion, whether data from users shows a product can be used effectively in schools, was where its pilot cohort's performance diverged most sharply. Most of the evidence providers submitted was user feedback; few used analytics or product data.

There's a third failure that looks like success. One trust leader at an EPI roundtable said whole-school purchases are too often passed to an “enthusiastic middle leader”. In a secondary school, that might be one head of department whose department uses the tool well while the rest of the school never starts.

What England's frameworks cover, and what they leave to you

Three frameworks matter most for secondary leaders this year, and none of them covers the whole question.

• The EdTech Evidence Board, delivered by the Chartered College of Teaching and funded by the DfE, sets prerequisites covering accessibility, data protection, safety, cyber security and integration with the systems UK schools use. It then judges six core criteria, from educational need through to positive effects on outcomes. Its September 2026 pilot covered 12 products.
• The DfE's generative AI product safety standards, updated in January 2026, are written mainly for suppliers, but schools can use them to assess products. Their assessment and analytics use case covers tools that automate the marking of student work.
• The DfE's proposed edtech safety certification scheme opened for consultation on 10 November 2026 and closes on 4 February 2027. It would be technically non-mandatory, but statutory guidance such as Keeping children safe in education would expect schools to get proportionate assurance for in-scope products, and certification would ordinarily provide it. It would not judge educational effectiveness.

Safety may get a certificate, but not soon: a government response is due in spring 2027, and the scheme needs at least 18 months to set up after that. Readiness won't get one at all, so it stays a judgement for the school.

There's also an MIS change to plan around. Since 1 October 2026, the Academy Trust Handbook has required trusts to award MIS contracts under a new DfE framework as soon as practicable after it launches, which is planned for September 2027. It also limits extensions to existing contracts in the meantime. The framework isn't live yet, and Al Kingsley, who chairs BESA's EdTech Special Interest Group, has pointed out that the DfE's planned data spine has no published specification either.

Ten questions to ask before a pilot

Five test the first gate and five test the second. Ask for written answers, and ask every supplier, including the ones your teachers already like.

Gate one: can the people who carry the risk approve it?

1. Which MIS products do you sync with, and what happens in a trust that runs more than one? A school-ready tool takes classes from the management information system (MIS) rather than relying on hand-typed class lists. Integration with the systems UK schools use is one of the Evidence Board's prerequisites. In its pilot survey, which leaned towards secondary respondents, the teachers and leaders surveyed stressed compatibility with their existing technology, while edtech companies did not rank it highly. With MIS contracts set to move onto the DfE's new framework, ask how the supplier will handle the switch.

2. Can a trust lead, a headteacher, a head of department and a class teacher each see and do different things? The DfE's AI standards expect administrators to be able to set different permission levels for different users, and the linked cyber security standards expect multi-factor authentication on accounts with access to personal data. Trust leaders told the EPI in its work on AI in trusts that they may need to define which decisions sit with the trust, such as approving AI tools, and which sit with schools or classrooms. If a tool's roles don't match that split, people will work around them.

3. Can you send your DPIA pack, accessibility statement and position on the DfE AI standards today? Trusts complete data protection impact assessments (DPIAs) for AI tools, but told the EPI they don't always understand the tools well enough to do it confidently. A school-ready supplier does the groundwork: a DPIA template, a data map, its sub-processors, and where data is stored and for how long. The Evidence Board found several providers believed they met baseline requirements but couldn't demonstrate it. Accessibility and online safety were the weakest areas.

4. Is any student or teacher work used to train AI models or improve the product? The DfE's standards say products must not store, collect or use work that students or teachers create, including to train models or to develop and improve the product, without permission. For students under 18, that permission comes from a parent or guardian. The supplier's answer should be one plain sentence.

5. What does onboarding look like across our schools, and what happens to our data if you stop trading? Rollout is a process, not a purchase. Falling staff confidence makes support more important: across all schools, the share of leaders who said more than three-quarters of their teaching staff were confident with technology fell from 54% in 2023 to 40% in 2025, according to the DfE's survey. One trust leader also told the EPI that their trust keeps small providers out of strategic roles altogether. Whatever your policy, know who you're dealing with.

Gate two: will teachers keep choosing it?

6. Which of our priorities does this serve, and how would we know by the end of a term? A teacher adopts a tool to clear this week's marking. A school adopts one to deliver a priority it has already committed to, such as faster feedback after mocks or more consistent marking across a department. The EEF's implementation guide starts with the need, not the product.

7. Can two teachers, or two schools, use the same assessment and compare results fairly? A personal workspace suits one teacher. A department needs shared assessments and consistent judgements across classes, and a trust can pool evidence across schools in a way one school can't. That depends on handling variety. One trust leader at an EPI roundtable, describing academies with different curriculums, timetables and exam boards, said “trying to make it bespoke for each of the academies is our greatest barrier”.

8. After one mock series, what will a head of department and a trust leader each see? One trust leader told the EPI that providers report engagement statistics when schools need to know whether students are making progress. Good reporting shows learning, not logins: which questions a year group found hardest, and where a misconception runs across classes. The UCL summary adds that dashboards should collect proportionate data and not add workload.

9. Can we see usage by school, department and teacher without asking you for a report? A tool that only works for enthusiasts depends on one champion, and the EPI's roundtable report warns that when a lone digital lead leaves, the expertise goes too. Leaders need to see adoption early enough to act, well before renewal.

10. What evidence do you have from schools like ours, and what exactly did it measure? Trust in supplier claims is low. The Evidence Board report cites a BESA (2023) finding that just 8% of educators trust the claims made by edtech companies, and Teacher Tapp's 2026 research puts it near 6%. The DfE's AI standards also say suppliers should not exaggerate the impact or capabilities of their tools. Treat workload claims with care: trust leaders told the EPI that freed-up time tends to fill with more work.

How to run the pilot

Implementation is part of what makes edtech work. The EEF's 2025 review lists planned implementation among nine mechanisms of effective use. It also found a lower average impact for disadvantaged students, so edtech can widen gaps unless the rollout plans for it. I'd run a pilot in seven steps: need, evidence, safeguards, pilot, measure, decide, scale.

1. Need. Name the priority, and give the decision to a senior leader rather than an enthusiastic volunteer.
2. Evidence. Ask question 10 before you shortlist, not after the pilot.
3. Safeguards. Get written answers to questions 1 to 5 before any student data goes in.
4. Pilot. Choose two or three contrasting departments, ideally in more than one school. One trust at an EPI roundtable found its staff weren't using its AI strategy because it didn't fit their subjects.
5. Measure. Agree success criteria with the supplier upfront, and run for at least one full assessment cycle, such as a mock series. In the one trust's data analysed for the DfE's market assessment, assessment products spiked around testing periods.
6. Decide. Scale, adjust or stop, and check whether workload actually fell or was replaced by different work.
7. Scale. Train every teacher, not one champion, as the EPI recommends.

What to ask if the tool marks student work

AI marking falls under the assessment and analytics use case in the DfE's AI standards, so it deserves a closer look. Start with accuracy, measured properly. Ask how closely the tool's marks agree with qualified examiners, compared with how closely those examiners agree with each other, in your subjects and on your exam boards' mark schemes. Then ask whether a teacher can see the evidence behind each mark, how accuracy varies by question type and with handwriting, and how teachers keep the final judgement before results reach students. I've written separately about how to evaluate AI marking accuracy.

Two gates, one decision

Teachers decide whether a tool is good. Leaders decide whether it's safe to scale. I think the rollouts that last are the ones where top-down meets bottom-up on purpose, and both sides would choose the tool again at renewal. If a supplier can't help you answer both questions before the pilot, the pilot isn't ready either.

Sources

• Chartered College of Teaching for the DfE - Piloting the EdTech Evidence Board: report on phases one and two, September 2026.
• DfE - Generative AI: product safety standards, updated 19 January 2026.
• DfE - Introducing a certification scheme for assuring edtech and AI product safety, consultation, 10 November 2026 to 4 February 2027.
• PUBLIC for the DfE - Assessment of the education technology market in England, June 2026, and EdTech Innovation Hub - DfE puts England's EdTech market at £5.9bn and calls for clearer usage data, 29 June 2026.
• DfE - Technology in schools survey report: 2024 to 2025, 27 November 2025, and Edurio - DfE Tech Survey: staff confidence is the missing piece in your digital strategy, January 2026.
• DfE - Academy Trust Handbook 2026, effective from 1 October 2026.
• Education Policy Institute - Edtech decision-making and inclusive practice: insights from multi-academy trusts, December 2025, produced in partnership with Sync.
• Education Policy Institute - Exploring multi-academy trust approaches to artificial intelligence, January 2026.
• ESRC Education Research Programme, UCL - Why aren't schools able to procure the technology they need to support high-quality teaching?, August 2025.
• Education Endowment Foundation - How edtech interventions can support learning, July 2025, and A school's guide to implementation, 2024.
• BESA - Al Kingsley (EdTech SIG chair) letter to the Permanent Secretary, August 2026.
• Information Commissioner's Office - When do we need to do a DPIA?.