AI amplifies the operating model you already have. If it’s messy, well...

Most companies are now moving past individual AI gains, and are using AI to help teams deliver more, faster.

You might start using AI for stuff it's good at, like breaking work down, writing specs, code, tests... Then you realise specs are arriving faster than anyone can approve them, because approval still runs through a fortnightly forum.

The value lies in moving from automating how you work to reinventing your workflows.

Your operating model was designed for a time where building was expensive and time consuming (eg: collating customer research, writing code, or designing a UI). As we employ AI to do this work the cost and time to do this is no longer the bottleneck, so the operating model that was designed around it no longer makes sense. (This changes the economics of the entire operating model - but that's another topic to write about later.)

If you want to generate real value from AI adoption (ie: meaningfully contribute to EBIT; new products and new revenue; as well as cheaper build or operations) you need to reinvent your workflows. McKinsey’s 2026 survey found 80 percent of respondents report individual productivity gains but only 37 percent see ANY EBIT impact, and only about 6 percent are high performers (at least 5 percent of EBIT attributed to AI.)

Here's how we've been helping clients do it:

• Start with the outcome, not the AI task
  Design your workflow backwards from the outcome. Don't just drop agents into your legacy process.

• Pay off your workflow debt before you automate
  Think about your governance, approvals, meetings, handoffs, and exception management.

• Organise teams around the end-to-end outcomes
  Don't think of this as a big restructure, just get teams to align to the outcomes and involved domains. This is so important to help manage the cognitive load and to move fast.

• Make the human and agent split an explicit design decision
  After redesigning the workflows, explicitly define what stays human and what goes to the machine.

• Constraints move. Think up and downstream
  As one step speeds up, keep searching for what's now under pressure. These mobile bottlenecks may change over time, so keep searching for them and understand the end-to-end impacts.

• Stop pilots, and focus on a few critical end-to-end workflow rebuilds
  Get a named owner with authority to own this. Be clear on the outcomes. Focus your attention on this and think end to end. Don't let individuals redesign just their own AI solutions.

• Oh, and build in lots of rigorous product and engineering practices - the ones you should have had in place for decades, but have probably been sliced out piece by piece.
  eg: small batches; tight customer feedback loops; CI/CD.

If you want to dig into this in more detail, just send me a message.