“Agentic-first” is the core differentiation lever against incumbents — but it needs to stay specific, not become a buzzword in the pitch. Two concrete places it changes the economics:

Configuration

Configuration is normally where small customers get burned — a Totara implementation partner spends billable hours mapping a garage’s specific compliance requirements into the platform. An agent that interviews the shop owner and self-configures qualification frameworks, complaint workflows, and improvement-plan templates removes the services layer that makes incumbents expensive and slow.

Ongoing maintenance

Chasing techs for expiring certs, drafting CAPA responses, prepping audit evidence packs — small operators have zero admin capacity for this today. An agent doing it for them is the product, not a feature bolted onto a system of record.

Why this matters for the business model

If agents replace the implementation-partner services model, the cost structure changes enough to profitably serve a segment (single-site garages) that’s currently unprofitable for incumbents to touch. That’s the actual wedge — not “AI features,” but a different cost structure that makes the underserved segment reachable.

Risk to hold onto

Agentic reliability in a compliance context is a double-edged sword — see opportunity/open-questions.md item 4. This needs a concrete answer before it’s a product spec: what’s agent-automated vs. agent-assisted with a human sign-off step, and where does that line sit for qualification records vs. complaint handling vs. audit evidence.

Second risk to hold onto, added 2026-08-10

The “why this matters for the business model” argument above assumes agentic automation is what makes single-site garages profitable to serve at all — that without it, the segment is structurally unreachable for incumbents. Desk research found a live counter-example: WoFMate reaches this exact segment profitably today via a plain self-serve SaaS product, no agentic architecture at all (see research/nz-specific-trade-compliance-check.md). That doesn’t kill the agentic-first thesis, but it does mean the argument needs to shift — agentic-first has to be argued as a deeper differentiator (the record-linking layer, ongoing-maintenance automation actually being done for the operator rather than just digitised) rather than as what makes this segment reachable in the first place. A non-agentic competitor already proved the segment reachable; the open question is whether agentic automation is a meaningfully better way to serve it, not the only way.