AI for EdTech Companies in Canada

When edtech companies call us

• A school board sent a privacy questionnaire about your AI feature. Sales is waiting while product and engineering work out where prompts go and which data elements the model sees.
• The product team wants a generative AI feature this year. A tutor, feedback on writing or lesson content generation, and nobody has chosen a model, host or review process.
• A Quebec customer asked for a Law 25 privacy impact assessment. Your team needs to explain where data is stored and processed, and who sees it.
• Support answers the same teacher and administrator tickets every week. Login, rostering and gradebook questions fill the queue.
• Developers use AI coding tools without shared rules. Leads want consistent review and secrets handling across the repository.

What we build and automate for edtech

One AI product feature, end to end. A custom AI feature such as writing feedback or content generation, with prompt design, guardrails for age-appropriate output, logging and an evaluation set. A teacher reviews or confirms anything that affects a grade or placement.

Model hosting and data path. We compare Azure OpenAI, the OpenAI API, AWS Bedrock and Anthropic on where data rests and where inference runs, then wire the chosen provider through your API layer. Your engineering lead approves the architecture.

School-board privacy pack. AI drafts answers to security and privacy questionnaires from your approved documentation, plus a one-page sheet per product with the data elements, purpose and operator that Ontario boards now need for their notices. Your privacy lead edits and signs off.

Support ticket triage and help-centre assistant. A support assistant answers routine teacher and admin questions from your help centre, in English and French, and passes account-specific cases to an agent with a summary.

AI coding practice for your developers. Claude Code training with review rules and repository conventions, so AI-assisted code meets the same bar as the rest.

Canadian rules that shape the work

PIPEDA and OPC guidance. PIPEDA applies to private-sector organizations that collect, use or disclose personal information in the course of commercial activity; Alberta, British Columbia and Quebec have their own substantially similar laws. The federal, provincial and territorial privacy commissioners' principles for generative AI ask for valid, meaningful consent where consent is the legal authority, and openness about privacy risks. Your product documentation needs to say what the AI feature collects and why.

Joint resolution on edtech (October 2025). Canada's privacy regulators called on schools, boards and edtech vendors to treat children's information as sensitive and to keep the best interests of the child paramount in development and procurement.

O. Reg. 52/26 (Ontario). From July 1, 2026, a school board must give written notice when a student's personal digital information goes to a third-party software application. The notice lists the data elements, purpose, app name and operator. A one-page sheet per product with these facts is what the board's privacy coordinator needs from you.

Quebec Law 25. Section 3.3 requires a privacy impact assessment for a new information system that handles personal information. Section 17 requires one before personal information is communicated outside Quebec, including to a service provider. Section 12.1 covers decisions based exclusively on automated processing. The feature design keeps a teacher in the decision and the privacy pack documents the data path.

Data location facts by provider. Azure OpenAI Standard deployments process prompts in Canada East for a few models only. For eligible customers, the OpenAI API can store data at rest in Canada, with processing outside Canada. AWS Bedrock in Canada (Central) keeps logs and data at rest in Canada while Claude inference runs in US or global regions. The full picture goes into the compliance review.

Which package fits

Edtech teams usually know the feature they want, so the AI Pilot Sprint is the common start: six weeks, one working feature or workflow. Teams with several competing ideas start with the Opportunity Audit.

Full scope of each package is on the pricing page.

How the work runs

The work follows the Remolda Cycle: Audit → Strategy → Implement → Empower → Evolve.

1. Audit. A working session with the product lead, an engineering lead and whoever answers customer privacy questions. We review the feature idea, the current data flow and recent board questionnaires.
2. Strategy. A one-page design: model, host, data path, human review points and the evaluation set.
3. Implement. The six-week build in your repository and cloud account, with weekly demos and a privacy pack drafted alongside.
4. Empower. Handover to your developers, plus AI coding practice for the team.
5. Evolve. A review after the first term of use: output quality, flagged cases and the next feature.

In our experience a pilot usually takes six weeks from kickoff. It depends on access to your environments, test data without real student records and your release calendar.