Embedding AI into a 20-Year Recruitment Platform

We've been evolving Indus Recruitment's platform for more than twenty years — from the original IRDS migration off Access through to today's IndusApp. Embedding AI into that stack taught us a few things that apply to almost any NZ business sitting on systems that already work.

The short version: don't bolt on a shiny chatbot and hope. Put AI where consultants and clients already spend their day — search, reports, profiles, and document intake — and keep people on the decisions that matter.

Start with the busy, repetitive jobs

For Indus, the high-volume wins were obvious once you watched how people worked: hunting for the right temp with keyword search, turning dense timesheet and absence data into something a client could act on, drafting employer-facing candidate blurbs, and re-keying CVs from the inbox.

Those became Smart Search, Indus Insights on employer reports, employee AI summaries, and CV / inbox document processing — sitting inside IndusApp, not beside it. No invented ROI percentages here; the point is the pattern: pick the grunt work first, prove it, then expand.

Keep consultants in control

AI drafts and ranks; people still place, approve, and talk to clients. Summaries are there for quick context while browsing skills and docs. Insights turn raw report data into short titles and detail — a human still owns the relationship. That human-in-the-loop discipline is the same advice we give every AI project: automate the volume, keep judgement where it costs money or trust if you get it wrong.

Build on data people already trust

IndusApp already held temps, placements, timesheets, documents, portals, and messaging. AI on top of that is useful. AI on shaky, disconnected data isn't. Platform modernisation (portals, SMS/WhatsApp, reporting UX) and the AI layer reinforce each other — the intelligence only earns its keep when the operational foundation is solid.

One shared foundation, not five one-offs

Features share a production-minded AWS Bedrock setup — config, caching, and error handling done once, reused carefully. That's less drama to maintain, and clearer boundaries for what the AI is allowed to do.

IndusApp is bespoke for Indus — we're not selling their platform as a product. What we are showing is the approach: embed AI into the systems you already run, for the processes that eat the week, with people still in charge.

Want the capability list? See the IndusApp project, or book a free chat about your own stack.