AuthorSeptember 21, 2026Categories
- Accountex
- AI in Accounting
- Data Repository
The narrative surrounding Artificial Intelligence across the UK accountancy sector remains dominated by a single proposition: tool selection.
Firm leaders routinely debate whether to deploy ChatGPT Enterprise, Microsoft Copilot, or specialized native practice management tools. However, as highlights from Accountex Manchester2026make clear, focusing on software ahead of data architecture puts the cart firmly before the horse.
During their session “Winning with AI: Why Better Data Is the Real Competitive Advantage,”Ed Franklin (Head of Business Intelligence at CCH / Wolters Kluwer) and Yemi Wedderburn (CCH Central Product Specialist) outlined why software implementation is failing to yield measurable ROI for practices that neglect their underlying data foundations.
For UK practices navigating rising operational costs, regulatory shifts, and capacity constraints, the takeaway is unequivocal: AI will not repair a broken database.
The Hallucination Fallacy and the “Confidently Wrong” Risk
In a profession built on technical precision, the tolerance for error in practice management and financial data is effectively zero. Yet Large Language Models (LLMs) inherently prioritize conversational plausibility over mathematical accuracy.
Wedderburn highlighted a recent real-world financial review where a simple manual entry error. Entering €5,000 as half of €10,000 was processed within a large dataset. Had an LLM been instructed to pull key financial metrics without validation, the error would have scaled exponentially, floating roughly €500,000 of nonexistent capital into management reporting.
“If you start challenging AI… and see something that doesn’t look right, you’ll get ‘sorry, I made that up.’ But if you hadn’t challenged it, was it just going to carry on making decisions on that?” — Yemi Wedderburn, Wolters Kluwer
Franklin termed the most dangerous operational risk not as complete fabrication, but as the “confidently wrong” output, a plausible-sounding answer derived from incomplete or corrupted data.
The AI Value Chain for Modern Practices
Input Foundation
✓ Complete, Clean Data
+
✓ Human Context
▼
Resources & Whitepapers
APThe importance of UX in accounts payable: Often overlooked, always essential
Accounting SoftwareThe power of customisation in accounting systems
Accounting FirmsTurn Accounts Payable into a value-engine
AP8 Key metrics to measure to optimise accounts payable efficiency
Scalable AI Competitive Advantage
Working Backwards: The “Clean with Purpose” Strategy
Data cleansing across a legacy practice management system (such as CCH, IRIS, or Advance) is notoriously arduous. Many mid-tier firms stall because they attempt to scrub their entire client database simultaneously without a clear operational goal.
Wedderburn and Franklin advised a target-driven alternative: Reverse-engineer your data hygiene.
-
Define the Specific Business Question: Avoid broad inquiries like “Which clients need our attention most urgently?” Broader questions fail because terms like “urgent” are inherently subjective.
-
Identify Necessary Data Points: Determine exactly which fields dictate “urgency” (e.g., fee margin shifts, missed statutory deadlines, unbilled WIP, recent turnover).
-
Clean Only the Governing Fields: Isolate and cleanse data strictly relevant to that operational query.
-
Establish Governance Controls: Put input guardrails in place to prevent future data degradation before moving on to the next business question.
This deliberate approach turns a monolithic database project into an incremental series of quick wins. Over time, clean datasets compound, allowing subsequent AI queries to deliver higher accuracy with lower error rates.
Operationalizing Insights: The “So What?” Standard
A recurring topic at Accountex Summit Manchester was the abundance of unused practice reporting. Many practices generate automated monthly dashboards that fail to alter management behavior. Wasting administrative hours and cloud compute resources.
To ensure AI inputs and business intelligence drive real value, Franklin urged managing partners to apply two classic diagnostic techniques to every internal report:
1. The “So What?” Test
Every metric extracted by an AI tool or reporting dashboard must lead directly to a specific action. If an automated insight (e.g., “Client X email response times have dropped 20%”) does not alter a partner’s decision or workflow, the data point is a vanity metric and should be removed from the prompt queue.
2. The “Five Whys” Framework
Continuously challenge the root cause of practice inefficiencies before deploying automated tools to solve them.
-
Why? Clients take a long time to upload verification documents.
-
Why? The automated request email is ambiguous.
-
Why? The request form pulls legacy, unstandardized service codes from our database.
-
Root Cause: Inconsistent database tagging, not a software capability issue.
Practical Checklist for Practice Leaders
Before committing capital to firm-wide AI rollouts, audit your practice’s readiness against these foundational requirements:
-
Audit Practice Context: Ensure qualitative context (e.g., client relationship status, spoken fee agreements) is systematically logged into structured fields rather than trapped in individual partners’ heads.
-
Implement Usage Auditing: Turn on usage tracking for your existing BI and practice management reports. Permanently archive any report that has not been accessed in 90 days.
-
Conduct Input Benchmarking: Test your current database against simple queries before subjecting it to AI tools. If a senior partner would not trust a junior trainee to make a decision based on the existing raw record, do not feed that data into an LLM.
-
Institute “Human-in-the-Loop” Controls: Ensure every AI output is treated as a first draft requiring verification from a qualified accountant before reaching a client or board meeting.
AI capability is accelerating rapidly, but it cannot bridge the gap created by poor data hygiene. UK accounting firms that focus first on data governance and practical business questions will build a lasting competitive advantage. Those expecting software tools to solve underlying data issues risk scaling their errors instead.
- Accountex
- AI in Accounting
- Data Repository
