Software engineering. Applied AI.

Teacharo

Selected work / Education

From a spoken note
to a school workflow.

Form From developed Teacharo’s voice-first assistant and student-support evidence workflows. We connected spoken notes, AI-generated drafts and teacher review in a working platform, with pilot-school use by August 2026.

By
The product
An assistant for teacher administration and NCCD evidence workflows.
Our contribution
Application development, backend services, AI processing and cloud infrastructure.
The result
A live production platform in pilot-school use by August 2026.

The task

Turn captured information into something a teacher can use.

A spoken note can be the starting point for a parent email, incident report, student note or lesson plan. Teacharo brings that preparation into one platform, alongside a separate workflow for Nationally Consistent Collection of Data (NCCD) evidence for student adjustments.

The product handles sensitive information and supports professional judgement. Form From’s task was to connect the user experience, processing and review steps across the application.

01 / The relationship

The draft is not the decision.

The draft is not the decision.Capture: A teacher records a note. Prepare: Transcribe, then de-identify text. Draft: Generate a proposed output. Review: The teacher decides what is used.01 / VOICE02 / PREPARATIONProposed output03 / DRAFT04 / TEACHER REVIEW
  1. 01

    A teacher records a note.

  2. 02

    Transcribe, then de-identify text.

  3. 03

    Generate a proposed output.

  4. 04

    The teacher decides what is used.

Illustrative workflow. No student data or client interface is shown.Select a stage to follow the path.

Our contribution

Build the product around preparation and review.

We developed the early TeacherPA product and carried substantial engineering responsibility into Teacharo, spanning the application, backend, cloud infrastructure and delivery process.

The assistant connects voice capture and transcription with text de-identification, AI drafting and teacher approval. De-identification happens after transcription. Teacharo’s text privacy service uses Microsoft Presidio and a local spaCy model to detect and substitute identifying entities; a separate residual-identity audit uses managed inference. These are distinct processing steps with their own data paths.

We connected the teaching assistant and NCCD workflow on the platform, supported by organisational permissions, linked task records, automated checks and staged releases. Teachers review drafts before using them. The architecture distinguishes preparation, privacy checks and the professional decision about what to do with the result.

The result

A working platform for learning with schools.

By August 2026, Teacharo had a live service and pilot-school use. The platform brought together note capture, draft preparation, teacher review and NCCD evidence workflows.

That is the delivery milestone established here. Measured changes in teachers’ administration time are not yet part of this case study.

What this means for another project

Make the workflow fit the responsibility.

For teams handling sensitive records, data preparation, access and review are part of the product experience. We design those steps alongside the AI feature.

Discuss a similar project
All selected workResetData