Most conversations about AI and software delivery are still hypothetical - a pilot, a proof of concept, a promising demo that never quite reaches production. We wanted a harder test than that, so we ran it on something we couldn't afford to get wrong: the CRM our own sales team lives in every day.
The problem with the obvious answer
The obvious answer was an off-the-shelf CRM: a recurring per-seat licence, a generic sales workflow, and yet another system to integrate with the Microsoft 365 tools our team already uses for calendar, Teams meetings and email. For a roughly 20-person sales team, that's a disproportionate amount of cost and integration debt for what is, at its core, a straightforward funnel: lead to prospect to closed deal.
What we built instead
- A purpose-built B2B CRM, designed and delivered by a small, senior AI-native engineering pod - not a multi-vendor delivery team
- A data model built and verified against a live database before a single screen existed
- Deep, delegated Microsoft 365 integration - calendar, Teams meetings, logged email - built directly on APIs we already pay for, with no new subscription cost
- Engineering governance enforced from day one, so speed never came at the cost of a system nobody else could maintain
What "AI-native" actually means here
It doesn't mean fewer standards. It means a small, senior team working with AI coding agents inside a disciplined process - phase-gated architecture, a verified data model, enforced invariants - shipping in weeks what a conventional programme would scope in quarters. The platform, erpost.com, is now live in production for two companies, including our own.
You can see it for yourself at erpost.com - and read the full story in our case study.
If you're weighing whether a small AI-native team could build something like this for your business, that's exactly what our AI-Native Delivery service is for - independent AI opportunity assessment and governance, plus hands-on delivery when building it yourselves is genuinely the right answer.