Every school improvement plan has two halves, and only one of them is writing.
The first half is the argument: what is actually wrong, which two or three things you will fix this year, and what you are deliberately choosing not to do. The second half is the paperwork that carries the argument. The self-evaluation. The plan itself. The report that goes to the board or the governing body. The policies the plan quietly commits you to. The evidence pack somebody asks for at the worst possible moment.
AI has become genuinely good at the second half. It cannot do the first, and the tools that imply otherwise are selling you the part you must not hand over.
This is for principals, headteachers, district and trust leaders, and anyone who has been handed improvement planning on top of a full job. I have put five tools against the five jobs a plan actually contains: getting the evidence, making sense of it, drafting the documents, writing the policies, and keeping the whole thing defensible.
How I built this list
I started from the school administration and safety category in our directory, which lists 26 tools, and widened to the full directory of 126 tools where a category boundary got in the way.
Four criteria. The tool has to serve a named strand of an improvement plan rather than a classroom task. It has to be in real use in schools now. Its data position has to be checkable. And its pricing has to be honest, which for this group mostly means published at all.
No vendor paid to appear here and no vendor saw this before publication. One tool in the list, the K12 AI Policy Generator, is built by my own team at AI Educator Tools. I have said so in its section rather than leaving you to notice.
Three things I left out, and why. The general classroom assistants, MagicSchool among them, are excellent teacher tools and are not planning tools, so they belong in a different list. GrantGettr looked like a good fit for the funding strand, but on 15 September 2026 its listed vendor site resolved to a consultancy homepage with no product page, so I could not verify it and have asked for the directory record to be checked. Alder Branch publishes no pricing I could confirm on the same date.
Two numbers from the directory are worth carrying into any purchase conversation. Of the 126 tools we list, 36 carry an independent 9ine security assessment, which is 29 per cent. And 48 of them do not state where they host data at all. That second number is the one that should slow you down, because an improvement plan is exactly the document into which people paste attendance data, safeguarding detail and named pupils.
All pricing and links verified on 15 September 2026.
The five jobs, compared
| Tool | Job in the plan | Pricing (15 Sep 2026) | Data hosting | Best for |
|---|---|---|---|---|
| Arbor | Holding the evidence | Quote-based, no public pricing | Not stated | Schools whose plan keeps citing numbers nobody can find |
| Edu Intelligence | Making sense of it | Quote-based, no public pricing | UK | Leaders who have the data and not the analysis time |
| SLT AI | Drafting the documents | £10 per user per month, £95 per year | UK | Writing the SEF, the plan and the governor report |
| K12 AI Policy Generator | The policy strand | Free | US | Turning a plan commitment into an actual policy |
| 9ine | Keeping it defensible | Quote-based, 14-day guest pass | Not stated | Schools whose plan includes adopting AI at all |
Best for holding the evidence: Arbor
A plan that cites numbers your team cannot reproduce is a plan that falls over in the first hard question, and this is a management information problem before it is an AI one.
Arbor is a cloud MIS: real-time attendance, behaviour recording, progress tracking, statutory reporting and staff cover, with analytics on top. It is the layer most improvement plans are actually arguing about, and the vendor now markets AI features within the suite rather than as a separate product.
What sets it apart in this list is verification. Arbor is one of the 36 tools in our directory carrying an independent 9ine security assessment, which matters more here than a feature list, because this is the system holding your most sensitive data.
Watch out for the obvious: replacing an MIS mid-cycle to improve your planning is a two-year project that will itself become the improvement plan. If you already have a capable MIS, the honest move is to use the reporting in it properly before buying anything new.
Best for schools whose planning stalls on finding the evidence. Not for schools that already have good data and no time to read it, which is the next job.
Pricing is quote-based with nothing published, as of 15 September 2026. On safety: our record does not state FERPA or COPPA status, hosting location or student data handling, which for an MIS is a gap worth raising with them directly. It carries the 9ine assessment.
Best for making sense of the data: Edu Intelligence
This is the tool that turns "we think attendance is driving the Year 9 problem" into something you can put in front of governors.
Edu Intelligence sits on top of the systems you already run, MIS, assessment, feedback and unstructured documents, and answers questions in plain language: which pupils are at risk, where the gaps are between cohorts, what relates to what. The vendor states a seven-day implementation with no data migration, and the product is built by Faria Education Group under the Welbee brand.
What sets it apart from Arbor's own analytics is that it reads the documents too, not only the structured fields, which is where a surprising amount of a school's actual knowledge sits.
Watch out for the pricing conversation. There is nothing public, so this is a procurement exercise rather than a card payment, and a small school should ask early whether it is sized for them.
Best for leaders drowning in dashboards. Not for a school that has not yet got its data into one place.
Pricing is quote-based, verified 15 September 2026. On safety: FERPA declared, COPPA not applicable, UK GDPR declared, hosted in the UK, collects student data including performance data and dates of birth, does not share with third parties, deletion available on request, minimum age 18. No independent 9ine assessment.
Best for drafting the documents: SLT AI
If the bottleneck is that nobody has three uninterrupted hours to write the self-evaluation, this is the tool in the list built for exactly that.
SLT AI offers a large set of structured tools aimed at school leadership writing: self-evaluation, the school development plan, deep dive and inspection preparation, governor reports, appraisals, EHCP plans and attendance letters, with outputs shaped by your school's context. It is grounded in a maintained knowledge base of current DfE, Ofsted, EEF and KCSIE documents, which is why its drafts read like sector documents rather than generic prose.
What sets it apart from asking a general assistant the same question is that grounding. A general model will write you a confident SEF from its training data and will not know which framework you are inspected against this year.
Watch out for two things. The first is the temptation the whole article is about: a fluent draft of your priorities is not a decision about your priorities. The second is a records mismatch worth knowing before you plan a budget. Our directory lists SLT AI as freemium and its feature list mentions free AI chat, but the vendor's public pricing page showed no free tier on 15 September 2026. Ask them directly.
Best for UK leaders writing to an inspection framework. Not for US districts, where the knowledge base is aimed at the wrong jurisdiction, and not for anyone wanting analysis of their own data.
Pricing on 15 September 2026: £10 per user per month, £95 per user per year, falling to £85 per user annually for two to five leaders and £75 for six to ten, with custom quotes above ten. The vendor prices in pounds only. On safety: FERPA and COPPA not applicable, UK GDPR declared, hosted in the UK, does not collect student data, deletion on request, minimum age 18, Cyber Essentials certified per the vendor. No independent 9ine assessment.
Best for the policy strand: K12 AI Policy Generator
Every improvement plan that mentions AI creates a policy obligation, and most of them leave it unwritten for a year.
The K12 AI Policy Generator walks you through a structured question-and-answer process, either reviewing a policy you already have or building one, against a thirteen-step framework and linked to your local regulation. It is free.
Disclosure before anything else: this one is built by AI Educator Tools, which is my own team. That is a reason to weigh it against alternatives rather than to trust it, and if you already have a policy process you rely on, use that instead. What it does offer that a blank document does not is the structure, which is usually what stalls the job.
Watch out for treating the output as finished. A generated policy that has not been argued over by the people who will enforce it is a document, not a policy. There is more on that in the guide to writing an AI policy for schools.
Best for schools that have committed to AI in a plan and have nothing written. Not for a trust or district needing a legally reviewed policy across many schools.
Free, verified 15 September 2026. On safety: FERPA and COPPA not applicable, GDPR declared, hosted in the US, does not collect student data, deletion on request, minimum age 18. No independent 9ine assessment.
Best for keeping it defensible: 9ine
If your plan says the school will adopt AI this year, this is the strand most plans forget until something goes wrong.
9ine provides AI and EdTech governance, data privacy work including DPIAs and vendor assessments, and cyber security testing, delivered as a platform with specialist support behind it. The vendor points specifically at the 2026 KCSIE requirements, which now bring AI into safeguarding systems.
What sets it apart from doing this in-house is the vendor due diligence, which is the part that quietly eats a data protection lead's year. It is also the organisation whose independent assessments appear against 36 tools in our directory, which is the standard the rest of this list is measured against.
Watch out for scale. This is a service purchase with a procurement process attached, and a single primary school may find the fit awkward. There is a 14-day guest pass with no payment, which is the sensible way to find out.
Best for trusts, districts and schools adopting AI at any scale. Not for a school wanting a tool it can use this afternoon.
Pricing is quote-based, verified 15 September 2026. On safety: our record does not state hosting or student data position, which for a governance vendor is worth asking about directly. If safety is your main question, start with the compliance guide to safe AI tools for schools.
How to actually choose
Work backwards from the strand that is stuck, not forwards from a demo.
If you cannot evidence your claims, the problem is your MIS and its reporting, and no drafting tool will rescue it. If you have the evidence and no time to interrogate it, buy analysis. If you know exactly what is wrong and the document is the bottleneck, buy drafting, which is also the cheapest thing in this list by an order of magnitude. If the plan commits you to something you have not written a policy for, do that before the term starts. And if the plan involves adopting AI across the school, the governance strand is not optional, whoever provides it.
Then apply one test to whatever you buy. Ask where the data goes, and refuse to accept silence as an answer. Nearly two in five of the tools we list do not say, and an improvement plan is not the document to find that out with.
One more thing worth putting in the plan rather than in a purchase order. An IBM study published on 2 September 2026, surveying 1,019 K-12 education professionals, found that only 24 per cent of educators say the education system is adapting very well to AI, that only 20 per cent have had extensive AI training, and that lack of training is the top barrier they name, at 42 per cent. That is a staffing and development line, not a software line, and it is the strand most likely to decide whether anything in your plan actually happens.
What I would do this week
Take last year's improvement plan, and mark every sentence that is a decision. Most leaders find fewer than they expect, and the rest is description.
Then hand the description to a tool and keep the decisions. That is the whole method, and it is why the cheap tool in this list is often the one that changes the most.
The bottom line
AI can now produce every artefact a school improvement plan contains, and not one of the judgements the plan exists to record. Buy for the strand that is stuck, check where your data lives before you upload anything, and browse the full directory when you know which job you are solving.
The plan is the thinking. The paperwork was never the point.
Dan Fitzpatrick is a Forbes contributor, bestselling author and founder of The AI Educator. More from Dan.
