When the AI lives inside the product.

Three personalities, one memory, your family threaded through every conversation.

Parents land in a chat with Jackie, Liam, or Nora: pragmatic, nurturing, or analytical. Each one has their own voice and opening style, but they share the same memory. The consultant carries the family's journey, their shortlist, and prior school comparisons across every conversation, so it remembers what you told it three sessions ago.

• Live quality checks on every response, so weak output is caught before the family sees it
• Smart model routing keeps responses fast and costs predictable, without dropping answer quality
• Strict data-handling controls. We don't allow third-party model providers to retain anything, because Canadian privacy law (Quebec Law 25, PIPEDA) matters when kids' data is involved

Matching that understands meaning, not just keywords.

Every school's profile gets turned into a searchable map of what it actually offers. When a parent says "we want a small school with strong arts, not too academic, walkable from the Plateau," the consultant matches against the real profile text. Specific facts (curriculum, tuition, programs) are pulled through a grounded lookup, so the model isn't free-styling.

• Smart caching across the system, so repeat questions don't run up costs
• Profile updates dedupe cleanly, so freshness doesn't break the index
• Every fact the consultant cites is grounded in the real school profile. No inventing details

Tools that act on the world, not just talk about it.

The consultant doesn't just chat. It does things. It shortlists schools, books visits, updates the family profile, and triggers post-visit debriefs. Each action runs through a small, well-defined set of operations, so nothing goes off the rails. The whole thing is wrapped in cost limits, fallbacks for bad upstream days, and per-user spending guards.

• Per-user, per-day cost limits, so a runaway loop can't burn through the budget
• Three layers of fallback. If a model vendor has a bad day, the product degrades gracefully instead of going down
• Smart caching on the parts of every conversation that don't change, so repeat turns stay cheap

A school-embedded AI guide, trained on the school's own website.

Partner schools get Emma: a per-school AI guide that learns from their own content. We read up to 500 pages from each school's website, break them into manageable chunks, and let Emma answer prospective-parent questions in that school's voice. She qualifies leads and routes the hot ones, all without parents leaving the school's site.

• Runs in an isolated environment per school, with its own permissions and data
• 10 quality checks covering FAQ, lead qualification, learning support, tuition, competitor questions, and defense against prompt injection
• Daily cost limit on every school, so the economics stay predictable

A consultant for schools, not just families.

Schools get Jarvis: a school-side AI consultant who helps them build a stronger profile, sharpen their marketing, and triage inbound inquiries. Same conversational feel as the parent-side consultant. Different voice, different scope. Separate login, separate audit trail, one product.

• Three scoped abilities: read the inquiry queue, respond to inquiries, and update the school's profile
• Marketing guidance baked into the chat. Jarvis critiques copy, suggests channels, and helps frame the audience
• Inquiries ranked with the same fit score the parent saw, so there are no surprises on either side

Structured insight from messy tour notes.

After a family tours a school, the consultant runs a debrief: pulls the family's notes and impressions, asks targeted follow-ups, and produces a clear synthesis. Fit score changes, curriculum alignment signals, social-emotional fit. The debrief feeds back into the family profile, so the next conversation already knows what changed.

• Stored as a structured record, not just a chat log. Searchable, comparable across visits
• Every model call logged with cost and a quality score attached
• Each debrief loops back into the family profile, so the consultant gets smarter with every visit