AI in Finance

AI won't fix a broken finance function. I will.

Where AI creates real value in finance. Where it fails. What needs to be true before you deploy it. Written by an FCA who has done the transformation work and can see clearly where AI fits and where it does not.

AI is being sold as a transformation tool. It is not. It is an amplifier. It amplifies what already works in a finance function. It also amplifies what does not.

Finance functions with clean data, coherent processes, and genuine system adoption are ready for AI. The tooling now available to them is impressive and getting better quickly. Finance functions without those foundations will spend significant budget discovering that automation is not transformation.

I am a Fellow Chartered Accountant with Big 4 training and twenty years of operational experience as a founder and senior finance leader. I have built finance functions from scratch, delivered ERP implementations, designed controls frameworks, and automated processes across commercial and public sector environments. I now apply that operational lens to AI adoption in finance.

Not a vendor. Not selling AI tools. Writing and advising from the position of someone who has done the transformation work and understands exactly what needs to be in place before AI adds value rather than cost.

Where AI creates real value

Reconciliation and Matching

High-volume transaction matching is where AI creates some of its most defensible value in finance. LLMs handle unstructured text matching across ledger entries. Pattern recognition handles exception identification. Finance functions with clean data and documented processes can reach 80-90% automation rates on high-volume reconciliation.

Document Processing

Intelligent document processing extracts, classifies, and validates structured data from invoices, receipts, contracts, and statements. On structured document types from established suppliers, 85-95% straight-through processing is achievable. The exception handling design matters as much as the extraction capability.

Anomaly Detection

AI that monitors full transaction populations rather than samples changes the controls picture. Duplicate payment detection, unusual journal patterns, expense anomalies. Organisations with well-implemented AP anomaly detection typically recover 0.5-1.5% of invoice value in year one from duplicates alone.

Forecasting Support

Machine learning improves financial forecasting in specific conditions: high-volume data, complex non-linear relationships, patterns too intricate for manual model building. In the right context, ML demand forecasting improves accuracy by 20-35% against statistical methods. In the wrong context, a well-built spreadsheet is better.

Audit Preparation

AI-assisted audit preparation from the finance team's side: ensuring every balance is supported before fieldwork starts, identifying your own anomalies before auditors do, and organising documentation at scale. AI-assisted audits from the auditor's side mean full-population testing is now standard. Finance teams need to prepare accordingly.

AI Readiness Assessment

Before any of the above. A structured assessment of where your finance function actually is: process documentation, data quality, system adoption, team capability, governance readiness. Scored across five dimensions. The result tells you what to do next, not what to buy next.

Where AI fails in finance

AI does not fix bad process design. It automates it, faster and at scale.

AI does not fix data quality problems. It learns from inconsistent data and produces consistent nonsense.

AI does not replace human judgment on genuinely complex decisions. Accounting treatment, provision adequacy, contract interpretation, stakeholder management: these remain with the finance professional.

AI does not fix change management failure. A team that did not adopt the ERP will not adopt the AI layer on top of it.

The organisations that have done the transformation work are in an excellent position. The ones that skipped it and bought AI instead are going to find out what that costs.

The framework

01

Assess

Process maturity, data quality, system adoption, team capability, governance readiness. Five dimensions, honest scoring. The assessment tells you where you are, not where you wish you were. Free self-assessment framework →

02

Fix the Foundation

Document processes at task level. Remediate data quality. Improve system adoption. Build the governance structure. This is not a delay before AI. It is the work that determines whether AI delivers. The roadmap →

03

Prioritise by Value

Identify the five highest-volume repetitive decision processes in your finance function. Assess each against your readiness findings. The right candidates combine high volume, documented process, clean data, and a team ready to change. The full strategy framework →

04

Evaluate Vendors Honestly

Integration reality, model explainability, exception handling, implementation track record, total cost of ownership. Ask to see a customer case study where the implementation was difficult. The evaluation framework →

05

Pilot Narrow, Measure Hard

One process. Defined inputs and outputs. Agreed success criteria before you start. 60-90 day measurement period. Automation rate, error rate, cycle time. Numbers, not impressions. First 90 days guide →

06

Scale Deliberately

What worked, why it worked, and what needs to change before the next deployment. Scale from evidence. Agentic AI within three to five years. The functions that are ready will benefit significantly. Agentic AI: what to know now →

What I do

I work with finance leaders and their teams on AI readiness and AI-informed finance transformation. Engagements typically cover one or more of the following.

AI Readiness Assessment

A structured assessment of your finance function across the five readiness dimensions. Process documentation review, data quality audit, system adoption analysis, team capability evaluation, governance gap analysis. You leave knowing exactly where you are and what to do next.

Finance Function Preparation

The foundation work that readiness requires. Process redesign and documentation. Data quality remediation. System adoption programmes. Governance framework build. Delivered alongside your team, not handed over and left.

AI Implementation Oversight

Independent oversight through vendor selection, pilot design, implementation, and the first 90 days in production. Someone who has been through enough technology implementations to know where projects go wrong and can apply that pattern recognition before the problems become expensive.

CFO and FD Advisory

Senior finance leadership on a fractional or interim basis for businesses navigating AI adoption alongside broader finance transformation. The combination of operational finance experience and AI strategy is not common. It is what this moment requires.

Get in touch

If your finance function is not ready for AI and you want to understand what it would take to get there, or if you are already in an AI project and want independent oversight, get in touch.

I respond to every serious enquiry personally.

hello@maebhcollins.com

From the Blog