AI vs Human: A Procurement Framework for Choosing Tutoring Platforms After the NTP

The end of the National Tutoring Programme changed the procurement conversation for schools. Before, many leaders asked, “Can we get tutoring funded?” Now the sharper question is, “Which model delivers the greatest cost-per-impact for our pupils, with safeguarding and curriculum quality intact?” That is exactly where the choice between AI-first platforms and human tutor providers begins to matter. Schools need a procurement framework, not a marketing brochure.

In practice, the market now includes AI-first tools such as Third Space Learning’s Skye, alongside human-led providers such as MyTutor and Fleet Tutors. Each model can work, but they solve different problems. AI tutoring tends to win on scale, consistency, and price stability; human tutoring tends to win on relationship, adaptability, and subject breadth. The wrong choice is often not “AI versus human” in the abstract, but a mismatch between intervention goal and delivery model.

This guide gives school leaders a practical way to compare platforms after the NTP, with attention to school budget, safeguarding, curriculum alignment, and impact measurement. It also borrows from decision-making frameworks used in other sectors, such as scenario analysis under uncertainty, hidden cost analysis, and measurement beyond vanity metrics. Those disciplines matter because tutoring procurement is no longer just an education decision; it is a resource allocation decision.

1. What changed after the National Tutoring Programme

From funded expansion to value-based procurement

The National Tutoring Programme normalized online tutoring at scale and gave schools a real-world test bed for intervention design. When the external subsidy disappeared, the purchasing lens shifted from access to efficiency. School leaders are now asking whether tutoring should be used for catch-up, targeted SEND support, core curriculum recovery, or exam preparation. That means every contract must justify not only price, but also measurable academic lift.

This is why the post-NTP market looks more selective. Schools are no longer buying hours; they are buying outcomes, reliability, and operational simplicity. In that context, platforms that can show progress reporting, secure workflows, and repeatable delivery have an edge. For a wider view of how schools are rethinking the market, see the best online tutoring websites for UK schools in 2026.

Why the “best” platform is now context-dependent

There is no universal winner because schools buy tutoring for different reasons. A primary school trying to close a persistent maths gap needs a different tool from a secondary school buying GCSE intervention for a cohort of borderline students. Likewise, a trust looking for consistency across many schools will value standardisation, while a small school may prioritise subject breadth and individual tutor rapport. Procurement frameworks should reflect those use cases rather than vendor claims.

Think of it the way a digital team chooses infrastructure: the cheapest server is not the best if it cannot scale, monitor, or protect data. In the same way, tutoring contracts should be evaluated on resilience and fit. If you are interested in operational decision-making under pressure, the logic in monitoring and observability is surprisingly relevant.

What school leaders must now defend to governors and finance teams

Post-NTP, leaders need to explain why a tutoring service is worth renewing, expanding, or replacing. Governors may ask why a higher hourly rate is justified when an AI system appears cheaper. Finance teams may ask why one provider offers unlimited sessions while another charges per tutor hour. The correct answer is not always about price alone. It should include attendance, completion, curriculum alignment, safeguarding controls, and whether the intervention can be sustained over multiple terms.

That is why procurement must move from anecdote to evidence. Schools need a framework that shows the path from spend to result, just as other high-stakes sectors use data to justify investment. For a useful analogy, consider data-led recognition campaigns where impact is measured, not assumed.

2. AI tutoring vs human tutoring: the real trade-offs

AI-first platforms excel at consistency and scale

AI tutoring platforms are compelling when schools need a standardised intervention that can run at scale without fluctuating tutor availability. Fixed pricing models can create budget certainty, especially for primary schools and multi-academy trusts. An AI system can deliver unlimited or high-volume practice, keep pacing consistent, and reduce scheduling friction. This matters when the challenge is not a lack of tutoring sessions, but the inability to deploy enough of them efficiently.

The strongest AI-first propositions, including Third Space Learning, also align well to curriculum sequencing in maths. They can be used to target common misconceptions, practise fluency, and reinforce prior learning without requiring the school to source tutors repeatedly. For schools that need predictable intervention loops, AI tutoring can behave like a dependable utility rather than a bespoke service.

Human tutors still matter where nuance, motivation, and breadth are essential

Human-led platforms such as MyTutor and Fleet Tutors remain strong choices when schools need live explanation, flexible subject coverage, and personal rapport. Human tutors can reframe a concept in multiple ways, read hesitation in real time, and adjust the tone of a lesson based on a pupil’s confidence. That responsiveness is especially valuable in English, humanities, science problem solving, and exam coaching where a conversation can unlock understanding.

Human tutors also help when the intervention objective is wider than attainment alone. Some pupils need trust-building, academic mentoring, or confidence restoration after setbacks. In those settings, the emotional dimension of teaching matters. Schools should not assume an AI platform can replace every kind of human support, especially for older learners with complex needs.

The most important distinction is not “tech vs person,” but “task fit”

Procurement should begin by classifying the intervention task. If the task is repeated practice, standardised retrieval, or maths fluency, AI may deliver the best value. If the task is complex reasoning, high-stakes exam preparation, or motivation and confidence rebuilding, a human tutor may be better. Many schools will ultimately need a blended model, where AI handles the routine and humans handle the high-leverage exceptions.

This blended logic is similar to decisions in technology and operations where some functions are automated for scale and others remain manual for judgment. A clear frame for this kind of allocation can be seen in risk analysis for commercial AI and in practical discussions of hidden platform costs.

3. A procurement framework for choosing tutoring platforms

Step 1: Define the intervention job to be done

Before comparing vendors, write a one-sentence brief: what problem is this tutoring contract solving, for whom, and by when? Examples include “raise Year 11 maths borderline grades by one band,” “support KS2 fluency for pupils below expected standard,” or “provide remote GCSE science tuition across two time zones.” Without this definition, schools compare providers on features rather than outcomes. That leads to expensive, unfocused purchasing.

Once the job is clear, the platform choice becomes easier. AI-first tutoring is often strongest for broad, repeated maths intervention with a fixed curriculum map. Human tutoring works best when the need is multi-subject, pastoral, or deeply personalised. The procurement brief should also state whether the school values unlimited usage, bespoke match quality, or speed of deployment.

Step 2: Assign weights to cost, impact, safeguarding, and fit

A practical procurement framework should score each provider against four weighted criteria: cost-per-impact, safeguarding, curriculum alignment, and sustainability. Cost-per-impact is not simply the cheapest session price; it is the cost required to generate a measurable gain. Safeguarding includes DBS processes, live supervision rules, lesson controls, and data handling. Curriculum alignment asks whether the content follows the school’s scheme of work and exam board needs. Sustainability looks at whether the model can continue next year without blowing the budget.

A school might weight these factors differently. A trust focused on maths recovery may give 40% to cost-per-impact and 30% to alignment. A selective secondary school preparing candidates for exams may give more weight to tutor quality and responsiveness. One useful comparison mindset comes from scenario analysis, where you test options under best-case, expected, and worst-case conditions.

Step 3: Test the provider against operational reality

Procurement meetings can be deceivingly polished. Real-world delivery is where programmes succeed or fail. Ask how quickly the platform can onboard pupils, how attendance is managed, how missed sessions are handled, and how progress is reported back to staff. Check whether teachers can see lesson summaries, whether SLT can monitor engagement, and whether parents or carers can understand the intervention.

This is also where process quality matters. Good platforms make implementation simple; weak ones create admin drag that eats time from staff. Schools should take a page from fields that depend on reliable alerting and workflow, such as timely notifications and reliability-first decision making.

4. Cost-per-impact: how to think beyond hourly price

Why hourly rates can mislead

An hourly rate only tells you the sticker price. It does not tell you how many sessions a pupil actually completes, how much staff time is needed to coordinate the programme, or whether the tutoring translates into attainment gain. A lower hourly rate can still be poor value if attendance is inconsistent or if the curriculum fit is weak. Conversely, a fixed annual fee may be outstanding value if the platform can serve many pupils with high consistency.

This is especially relevant for school budgets under pressure. Procurement teams often compare an AI product with an hourly human tutor and assume the cheapest-looking option wins. But if the human tutor only serves a narrow cohort effectively, while the AI platform can support much larger groups with comparable gain, the AI option may offer lower cost-per-impact overall. The question is not “What is cheaper?” but “What is cheaper for the learning gain we need?”

A simple cost-per-impact model schools can use

Start with the total annual cost, including staff coordination time, onboarding, reporting, and any platform fees. Then estimate the number of pupils served and the likely academic gain per pupil. This does not need to be perfect; it needs to be honest and consistent across providers. A school can compare options using the formula: total cost divided by expected improvement at cohort level. That gives a rough but useful procurement ratio.

For example, a fixed-price AI maths service might appear expensive in isolation, but if it supports 100 pupils across the year with low admin burden, the per-pupil cost can fall sharply. A human tutor service may produce excellent results for 15 targeted pupils but be less scalable. This is where schools should look at opportunity cost as well as direct cost.

Budget sustainability matters as much as short-term uplift

One of the most important post-NTP questions is whether the school can continue funding the intervention next year. A model that only works if a grant renews is fragile. AI tutoring often offers more predictable budget planning because costs are fixed or capped. Human tutoring can be less predictable if sessions vary by subject, tutor, or geography.

In financial planning terms, sustainability beats short-term discounts. Schools know from other sectors that hidden costs can distort a bargain, whether in software pipelines or procurement cycles. For a broader lesson in total-cost thinking, see how to spot real discount opportunities without chasing false deals and hidden cloud costs.

5. Safeguarding and compliance: non-negotiables in platform selection

What good safeguarding looks like in online tutoring

Safeguarding is not a tick-box. It includes verified identity, DBS checks where applicable, session monitoring, approved communication channels, clear escalation routes, and data protection controls. Schools should ask exactly who can speak to pupils, who records sessions, how safeguarding concerns are logged, and how moderators intervene. If a platform cannot explain this clearly, it should not be on the shortlist.

For schools, the risk is not limited to misconduct; it also includes weak visibility. A platform can be educationally strong but operationally opaque, which leaves staff exposed. The best providers make safeguarding easy to audit. That is one reason why procurement should assess platform governance with the same seriousness as academic performance.

AI platforms need special scrutiny on data and interaction controls

AI tutoring brings new questions about data privacy, content control, and output consistency. Schools should ask whether the system is trained or constrained to follow the intended curriculum, what data it stores, how student information is protected, and whether outputs are logged for review. AI systems should not be treated as black boxes. The school must know how the tool behaves when a pupil makes an unusual request, submits incorrect working, or needs escalation to a human.

Good AI design in education should borrow principles from secure system architecture. The logic is similar to zero-trust data handling, where sensitive information is protected by default rather than assumption. It also resembles the care needed in managing synthetic output responsibly in other fields.

Human tutor platforms must still prove consistency

Human does not automatically mean safe. Schools should still check tutor vetting, lesson moderation, escalation routes, and the consistency of tutor quality. A platform with many tutors and weak quality control can create uneven experiences, especially if the school uses multiple cohorts or subject strands. Safeguarding must be robust across all tutors, not just on paper.

That is why verification matters. If you want a wider lens on trust and identity in digital systems, the ideas in identity verification and recovery are useful analogies for school due diligence. Trust should be earned through process, not asserted in a marketing claim.

6. Curriculum alignment: the difference between generic support and true intervention

Curriculum fit is where schools win or lose time

The best tutoring platforms do not just teach “maths” or “English”; they teach the exact knowledge, methods, and misconceptions your pupils are facing. Curriculum alignment reduces wasted time because tutors and tools target the sequence the school is already following. This matters especially in intervention, where time is limited and every minute should move pupils closer to a defined outcome.

AI-first maths platforms can be particularly strong here because they can be tuned to a curriculum pathway and deliver repeated practice at the right level. Human tutors are more flexible across subjects and can respond dynamically when a pupil’s misconception is not obvious. The key is to know whether the platform works from the curriculum outward or from a generic tutoring script inward.

How to audit alignment before you buy

Ask for sample lessons, topic maps, question types, and evidence of how the content matches exam boards or national curriculum strands. Check whether the platform reports by topic, not just by attendance. If the provider serves secondary pupils, ask how it handles tiering, specification changes, and exam-board differences. For primary, ask how it supports fluency, reasoning, and problem solving separately.

Schools should also ask teachers to test content against their own scheme of work. A one-hour internal review can prevent months of mismatch. When selecting learning systems, the discipline is similar to choosing the right landing page or content format for a local audience: fit matters more than generic reach. That principle is explored well in service-oriented landing pages and micro-market targeting.

Subject breadth versus depth is a strategic trade-off

If the school needs multiple subjects, human tutor marketplaces often look attractive because they can cover a wide academic span. If the problem is concentrated in maths, a specialist AI product may deliver deeper alignment and lower variability. Procurement leaders should beware of platforms that promise everything but excel at nothing. The best long-term value usually comes from a provider with a clear instructional niche and strong evidence in that niche.

This is a useful lens for multi-academy trusts. One provider may fit maths intervention trust-wide, while another provider fills a local GCSE English gap. A portfolio approach can outperform a one-size-fits-all purchase, provided reporting is consistent across suppliers.

7. Measuring impact: what to track and how to interpret it

Attendance is necessary, not sufficient

Attendance tells you whether pupils showed up. It does not tell you whether they learned. Strong impact measurement should track attendance, lesson completion, time-on-task, topic mastery, confidence, and eventual assessment movement. Leaders should be careful not to celebrate high utilisation without evidence of progress. Too many interventions look successful in administration and weak in attainment.

Impact measurement should also separate short-term and medium-term outcomes. A pupil may show immediate fluency gains but need more time to translate them into exam performance. For this reason, schools should set baseline, midpoint, and endpoint measures before any programme begins. This is how good procurement avoids the trap of impression management.

Use dashboards to make intervention decisions faster

Providers should make reporting clear enough for school leaders to act on. If a cohort is underperforming, staff should know why: attendance, content difficulty, motivation, or pace. If a group is succeeding, leaders should know whether to extend the programme, step down support, or redeploy resources. Data is only useful when it changes the next decision.

This thinking aligns closely with simple analytics stacks and with measuring impact beyond rankings. The same principle applies in education: don’t just collect data; use it to act.

What good impact evidence looks like at procurement review

At review stage, ask for a concise evidence pack: baseline data, cohort attendance, topic-level progress, teacher feedback, and a simple cost-per-impact summary. If the provider can only show testimonials, that is not enough. If they can show progress by subgroup, even better. The strongest vendors are comfortable being judged on outcomes because they know how to demonstrate them.

A rigorous review process also protects school leaders politically and financially. If a governor asks why a contract was renewed, the answer should be measurable. For a model of evidence-led communication, see creating impactful recognition campaigns using data, where structured proof matters more than slogans.

8. A practical comparison: AI-first vs human tutor platforms

The table below gives a simplified procurement view. It is not a substitute for due diligence, but it helps school leaders quickly frame where each model tends to be strongest.

For many schools, the answer is not one or the other. AI platforms may serve as the default layer for core intervention, while human tutors are reserved for subjects or pupils that need more interpretive support. That is a portfolio strategy, and it is often the most durable approach after the NTP.

9. A sample decision matrix schools can use today

How to score vendors fairly

Create a scorecard with 1-5 ratings for each provider across cost-per-impact, safeguarding, curriculum fit, reporting quality, and sustainability. Multiply each score by the weighting most relevant to your school. For example, a primary school with a tight maths gap might weight cost-per-impact at 35%, curriculum fit at 30%, safeguarding at 20%, reporting at 10%, and sustainability at 5%. A sixth form college may distribute weights differently.

When scoring, keep evidence in writing. Ask vendors for references, sample reports, and implementation timelines. Avoid scoring based on the personality of the demo. Good procurement is repeatable, transparent, and defensible. That is especially important when purchasing pressure is high and budgets are constrained.

What to ask before signing

Ask whether the platform can support your intervention timetable, whether it will integrate with your safeguarding expectations, and whether staff training is included. Ask what happens if uptake is low, if a tutor is unavailable, or if pupils fall behind schedule. Ask how data is shared, how often reporting is updated, and how quickly you can pause or extend usage. These questions reveal whether the provider is operationally mature.

It is also worth asking about exit routes. A school should know how to end the contract cleanly if outcomes disappoint. Vendor lock-in is a real risk in education procurement, just as it is in software. Responsible decision making means planning for change, not just launch.

When the right answer is a pilot

Not every decision needs a full-year commitment. If the school is unsure whether an AI or human model fits best, run a small pilot with clear success criteria. Measure attendance, engagement, teacher satisfaction, and topic gains over a term. A good pilot should reduce uncertainty, not create more of it.

Pilots are especially useful for trusts. They let leaders compare delivery models in similar contexts before scaling. Think of it like testing a product in a micro-market before national rollout; the logic behind micro-market targeting is highly relevant to education procurement.

10. The long-term sustainability question: what should schools build next?

Build a tutoring strategy, not a one-off purchase

Schools that get the best value from tutoring usually treat it as part of a wider intervention system, not a standalone fix. That means defining which pupils qualify, what progress looks like, how teachers hand over information, and how success affects the next term’s provision. When tutoring is embedded in a strategy, not a reaction, outcomes improve and budgets are easier to justify.

AI-first platforms can help schools create an always-on core intervention layer that is easier to scale and sustain. Human tutor platforms can supply the flexibility to respond to spikes in need, exam years, or specialist subjects. The long-term sustainable school is likely to use both, but with clear rules for when each is deployed.

Protect your budget from intervention drift

Intervention drift happens when a tutoring programme expands beyond its original purpose. A scheme that began as GCSE maths support becomes a general catch-up tool with no clear exit criteria. Budgets leak, focus weakens, and impact becomes hard to prove. Strong procurement avoids drift by setting start points, review points, and stopping rules.

Schools should review tutoring contracts just as carefully as they review teaching and learning initiatives. If a programme cannot show improvement, it should be reshaped or retired. The discipline here resembles how organisations assess performance under change, much like reliability-first operations and turnaround thinking.

Final recommendation: decide by job, not by brand

The strongest procurement framework after the NTP is simple: define the learning job, score the options honestly, and choose the platform that gives the highest measurable return for your context. AI tutoring is often the best choice when scale, predictability, and curriculum consistency matter most. Human tutoring is often the best choice when breadth, adaptability, and relationship-based support are critical. The real skill is not choosing one ideology; it is choosing the right delivery model for the specific intervention.

That decision should be evidence-led, budget-aware, and easy to defend. If your school can explain why a platform was selected, how safeguarding was checked, how curriculum fit was verified, and what impact will be measured, you have a procurement framework worth trusting.

Pro Tip: If two providers look similar, choose the one that gives you the clearest evidence trail. In post-NTP procurement, visibility often matters more than a polished demo.

Frequently Asked Questions

Related Reading

• 7 Best Online Tutoring Websites For UK Schools: 2026 - A practical shortlist of providers, pricing signals, and safeguarding basics.
• Why Teachers Leave: The Real Workplace Frustrations Schools Need to Fix - Useful context on staffing pressure and why tutoring often fills gaps.
• Rapid Creative Testing for Education Marketing - A smart model for testing ideas before committing budget at scale.
• How to Use Scenario Analysis to Choose the Best Lab Design Under Uncertainty - A strong decision-making analogy for comparing vendors under uncertainty.
• How to Use Branded Links to Measure SEO Impact Beyond Rankings - A useful reminder that measurement should focus on outcomes, not surface metrics.