Hours to Seconds: How we automated our Finance Operations using AND Automate

A time draining, fragmented expense claim workflow

At AND Digital, our Finance team carries the huge weight of many business critical operations. Historically, our team used a time-draining process that required many different parts - fragmented systems, spreadsheets and manual workarounds.

As a business that’s taking the lead on driving AI through 2026, we needed a pragmatic solution. Previous attempts to fix the problem through our Luna system were paused due to cost, complexity and long delivery timelines. Could we automate an end-to-end invoice process across Luna, Power Bi, spreadsheets, Google Drive and Xero without heavy reliance on individual knowledge, and time?

With the launch of AND Automate, a five week initiative that tested if targeting automation could deliver, we were about to find out. Success would depend on immediate, measurable impact without too many changes to core systems.

The Challenge

The historical invoice process required around 100 clicks and eight hours of labour per cycle, across multiple people and systems. Finance team members had to check and manually apply business rules held mostly as historical knowledge by the ANDi. This knowledge silo was an operational risk during times of absence and peak periods.

Our challenges:

• Fragmented finance process

• Heavy reliance on single ANDi knowledge

• Previous attempts at a solution were slow and costly

Our solution: deliver a quick, low-risk process that reduced manual work and ANDi dependence and provided real value.

What we did

At AND, we provide a small cross-functional team with product, engineering and finance SMES, supported by AI and IT advisors. We wanted not only to build a tool, but a solution. There was no question, of course, that we’d use this collaborative solution-based approach for ourselves.

Our team of ANDis followed a structured framework:

• Rapid Discovery - Validated processes, discovered pain points, prioritised automation opportunities

• Proof of Visibility - Test platforms including generative AI using an assess framework that scored across governance, capability and readiness

• Build (staging) - Build, with a focus on the highest use-cases of data preparation and invoice validation

• Production preparation - Definition of governance and ownership, rollout sequencing both using a roadmap to ensure sustainability

• Final solution build (production) - Finish definition of governance, enable platform, build new automation, end-to-end validation and training of finance ANDis

Whilst this may seem like a regular approach, our ANDis delivered at speed due to a tight deadline. This speed depended on the success of this iterative process with daily check-ins that focused on a go/no go decision from every member of the specialised team to ensure value. We also used AI as a tool to help our ANDis work at this speed. During discovery, previous information was collected from finance - notes were taken by Miro AI that were then summarised into concrete areas and sorted by AI via priority. During feasibility, we used AI to help create a score card framework so we were able to evaluate how suitable the different platforms were.

In the end, we chose Power Automate over Uipath and GenAI tools. GenAI lost out due to its slow unpredictability where Power Automate offered a pragmatic approach to data governance, and immediate cost advantages.

The Outcome

In four weeks, the project delivered:

• ~90% reduction in manual work for invoice validation

  Validation of ~200 invoices reduced from a full morning (two people) to ~20 seconds by one person

• Manual clicks reduced from ~100 per invoice to ~10 per cycle

  Dependency on individual knowledge significantly reduced by codifying business rules into automation

• Estimated annual savings of over £8,000, with net benefits from year two onwards

What’s more…

• The solution unlocked a new operating model:

Invoice validation can now run continuously through the month, catching any issue earlier. This reduces end-of-month pressure.

Stakeholder confidence was high, and we rolled our model out for the next monthly invoice cycle.

Ultimately, AND Automate helped us prove our repeatable approach - small teams, pragmatic tooling and real business outcomes provide efficient solutions, without the need to wait months for system overhaul.