← All insights
Process Architecture

Map before you automate

Michael Ahmed · 22 May 2026 · 2 min read

There's a seductive shortcut in every transformation programme: skip the mapping, jump to the automation. The mapping feels like overhead. The automation feels like progress.

It's a false economy. Automating a process you haven't understood doesn't remove the problem — it scales it. You end up delivering the wrong outcome faster, with a bot you now have to maintain.

The baseline is the asset

Before you can improve a process, you have to be able to see it. That means an honest As-Is map at a level of detail that actually drives decisions — not so high it's a marketing diagram, not so low it's unreadable. In practice that's BPMN 2.0 at Level 3 to Level 5, with a process taxonomy that lets the whole estate hang together.

Once that baseline exists, everything downstream gets easier:

  • Improvement becomes evidence-based. You can see the gaps, the duplication, the hand-off failures and the control weaknesses, and design a governed To-Be target state with clear ownership.
  • Automation becomes targeted. RPA and AI opportunities reveal themselves, and — crucially — you can build a business case with real effort and volume figures behind it.
  • Compliance becomes demonstrable. A mapped, owned process estate is exactly what a regulator wants to see, and it's most of the work of meeting frameworks like FCA Consumer Duty.

Rationalise while you map

Mapping nearly always surfaces a bloated document library underneath. This is the moment to rationalise it — in the work I've led, that typically means a 60–70% reduction in SOP volume while improving quality and consistency. Fewer documents, each one trusted, beats hundreds nobody reads.

The order matters

Our method is deliberately sequential: map it, improve it, automate it, make it queryable. Each step depends on the one before. Skip the first and the rest are built on sand.

If you're being pushed to "just automate something" to show momentum, the most valuable thing you can do is map one real end-to-end process first. It's slower for a fortnight and faster for a decade.

Book a discovery call