From pilot to profit: The real state of AI in Accounts Payable

• 1 in 3finance roles now explicitly ask for AI or machine learning skills
• 81%of AP teams have no AI in their workflow at all
• 18.4%average AP exception rate
• 4%of AP teams have exception management fully automated

The AI pilot trap

Most AP AI initiatives don't fail loudly. They stall quietly, stuck in permanent “pilot” mode. Without agreed success criteria, the pilot just keeps running quietly in the background.

1. 1

   An initiative starts

   The mandate is felt, and someone is asked to act on it

2. 2

   Tool before problem

   A vendor or in-house build is chosen before the problem is defined

3. 3

   Pilot greenlit anyway

   Momentum is easier to generate than clarity

4. 4

   It stalls

   Success was never defined tightly enough to know if it's working

5. 5

   Scaling can't be justified

   ROI is harder to prove than the business case assumed

Why AI projects in AP stall, and what to do about it

These five patterns show up repeatedly across the market, regardless of who's running the project.

1. 01

   Attempting to automate a broken process

   AI doesn't fix a bad process. It runs it faster and at greater scale, including all of its mistakes.

   The fixIn the full report
2. 02

   Undefined success criteria

   Most projects define success after the fact, once someone asks whether it worked.

   The fixIn the full report
3. 03

   Data quality as an afterthought

   The fixIn the full report
4. 04

   Governance gaps

   The fixIn the full report
5. 05

   Vendor selection before problem definition

   The fixIn the full report

From AI pilot to live programme: Mortenson’s blueprint

Ryan Ambrose leads finance process, data, and technology at Mortenson, one of the fastest-growing construction companies in the US. It jumped from No. 22 to No. 10 on ENR's 2026 Top 400 Contractors list as revenue climbed from $6.7bn to $10.8bn in a year, and now processes 300,000+ invoices a year on Oracle. But growth like that can punish a broken process fast.

What follows is Mortenson's framework to build a prioritised AI roadmap, in five steps, a blueprint that avoids the trap of AI pilots and instead creates a repeatable methodology for digital transformation.

1. Step 1Govern
2. Step 2Map
3. Step 3Score
4. Step 4Set KPIs
5. Step 5Activate, buy or build

How ready are you for AI?

The market has already answered whether AI belongs in AP. Mortenson’s blueprint shows what doing it in order looks like. The next step is knowing where you stand today, so we built two free tools to help.