AI coaching teachers won’t work without real trust

AI coaching – A new UK-built AI teaching coach, Starlight, is being pitched as a way to support private, whole-lesson reflection. But its core test is not technical capability—it’s whether teachers believe the audio-based feedback can’t quietly turn into surveillance, add w

When teachers talk about AI, the hesitation usually comes first. Not because they doubt what the technology can do, but because they worry they won’t be able to trust what it’s doing—especially when it touches something as personal as classroom practice.

That tension sits at the center of a pitch for Starlight. a UK-built AI teaching coach designed to support private. whole-lesson reflection. The idea is simple enough to sound appealing to any over-stretched teacher: record a lesson audio. upload it. and receive a structured coaching report within minutes. Starlight’s reports draw on the transcript. identify strengths. suggest a development focus. and offer practical actions teachers can keep. adapt. or ignore.

For educator Ross Morrison McGill, the trust question is not theoretical. McGill—who founded TeacherToolkit in 2007 and is widely recognised as one of the leading influencers in education in the UK and across the world—described how his own best professional development came from a very different kind of coaching. In the 1990s. he trained to teach and had a mentor who placed a video camera at the back of the classroom. The habit of reviewing footage, week after week, became the development he valued most. In that model. he said he didn’t have a weekly coach. a video platform. or an AI assistant analysing what was happening.

He also points to the pattern teachers have learned to fear: every new platform promises to save time, improve feedback, and transform professional development. Too often, the day-to-day reality becomes “another login, another dashboard, another thing teachers are expected to use.”

That’s why the first question he urges is not “what can the technology do?” but whether teachers would actually use it, trust it, and return to it.

The most immediate objection he raises is the risk of generic feedback. Teachers, he says, do not need bland comments that could apply to any classroom, in any school, on any day. He notes that teachers sometimes even receive generic feedback from people who know them well and observe their teaching regularly.

Starlight’s response is whole-lesson coaching analysis. Instead of relying on a short clip. a selected performance moment. or a generic prompt. the platform works from the actual lesson transcript—so the reflection can dig into details that disappear after a busy lesson. The kinds of teacher questions it aims to support are specific and practical: how much did the teacher talk; how long did they wait after asking a question; where did pupils contribute; where did the teacher’s explanation drift; and what could be changed in the next lesson.

The second objection cuts deeper: trust again, but this time tied directly to accountability fears. McGill frames it as the question of whether lesson audio could be used for performance management, accountability, or inspection. In his view, the concern is legitimate because “developmental” systems can drift toward monitoring.

He argues Starlight is built to sit outside that shift. The report is private to the teacher, while senior leaders only see anonymised, aggregated trends. The leadership view does not name individual teachers. He also says teachers can delete their data at any time. For him, that distinction is not a technical footnote—it’s the line between coaching and compliance.

The third objection is workload. Another platform that asks teachers to do more admin is unlikely to survive in schools that already carry too much. Starlight’s promise is that uploading takes under a minute and reports arrive within minutes. It also says the Lesson Toolkit turns the transcript into useful resources. including recap prompts. retrieval questions. key vocabulary. EAL support. and follow-up materials teachers would otherwise create from scratch.

That is where the argument for sustainability moves beyond reflection and into the daily grind of planning and preparation.

McGill also situates the pitch within a wider research and practice debate. He says instructional coaching has a strong research base. but scaling it is difficult because coaching works best when it is specific. regular. and rooted in evidence—yet schools struggle to find enough time and expertise for every teacher. He adds that there is growing interest in AI-assisted coaching. citing Harvard’s Centre for Education Policy Research and its M-Powering Teachers project. which explores AI-supported teacher coaching in mathematics.

At the heart of his message is a boundary: Starlight is not presented as a replacement for human coaching. Instead, it’s positioned as a way to make coaching more frequent, more specific, and more sustainable. He connects that to the principles he says he advocates in teaching: coaching. not compliance; insight. not surveillance; growth. not grading.

The ethics debate is treated as unavoidable rather than optional. He says any school considering AI feedback for teachers should ask serious questions before a pilot begins. because with lesson audio. trust cannot be vague—it has to be designed into the system. He adds that the school remains the Data Controller, while Starlight hosts data in the EU. Teachers can delete their data at any time. individual coaching reports remain private to the teacher. and school leaders only see anonymised. aggregated trends. He also points to a published DPIA.

The emphasis is practical: teacher trust isn’t something schools can bolt on after a tool is introduced. If teachers suspect that private reflection could drift into performance management, he warns, the culture will collapse before the software can help.

So before adopting any teacher coaching platform. he urges school leaders to ask four straightforward questions: will teachers choose to use it; is individual feedback genuinely private; does it reduce workload or add to it; and does it support professional judgement—along with what evidence will show that practice changes over time.

McGill ends with a clear statement of intent: he argues the promise of AI in schools won’t be found in shiny dashboards. but in tools that help teachers think more clearly. work more efficiently. and retain control over their professional growth. He says that if he were still in the classroom, he would be using it without hesitation. He also notes “You can try Starlight for free this term.”.

AI coaching for teachers Starlight teacher CPD lesson audio GDPR consent instructional coaching teacher workload education policy UK education