For accountancy practices that have investigated AI, there’s only one question:
“We saved time. But did we make money?”
It’s blunt but it also cuts through the hype.
AI can draft faster, summarise faster, compare faster and prepare a first pass faster. But if the client still buys “hours” and the firm still prices by assumed manual effort, the commercial result can be disappointing.
Finding a way out of this trap is what this article is all about. Here’s what we cover:
AI trap #1: Speed isn’t the product
The danger isn’t that AI fails.
The danger is that it works so well that it exposes a fragile business model.
If a piece of work used to take four hours and now takes one, the client may expect the fee to fall. If the firm lowers the fee without changing the service proposition, it has transferred the value of AI to the client while keeping the implementation cost, review risk and governance burden.
Sadly, that’s not transformation. It’s margin erosion.
The efficiency trap appears when firms confuse time saved with value created.
A faster reconciliation is helpful, but clients don’t ultimately remain on your books because the reconciliation happened quickly. They stay because the figures are clean, the risks are understood, the story is explained and they feel in control of their business.
If AI simply accelerates the old hourly model, it can collapse the logic of that model. Time becomes a cost to be reduced, not a product to be sold.
That’s why practices must stop presenting AI as a cheaper way to do the same work.
The commercial opportunity is to create a better service: earlier insight, clearer client questions, proactive exception reporting, stronger working papers, documented controls and more regular conversations.
The output should feel more valuable, not merely faster.
AI trap #2: It’s not about repricing
Pricing should follow workflow redesign, not the other way round.
Start by mapping the service and asking where AI can reduce admin, where humans must review, where exceptions arise and where the client experiences value.
Then remove unnecessary steps: standardise templates, create human checkpoints and define the review evidence required before anything is issued.
The webinar set out a clear sequence: redesign workflows, reduce work, productise services, reprice outcomes, create human checkpoints and standardise templates and agent patterns.
Sequences like this matter. If you put AI into a messy workflow, you get faster mess. If you put AI into a controlled workflow, you get repeatable capacity.
A practical example is quarterly compliance and insights. The old model might be “VAT return preparation – £X”. The new model is stronger: “Quarterly Compliance and Insights Pack – £X”.
That package can include digital record checks, exception reporting, variance commentary, client questions, cashflow observations and a short review call.
The client is no longer buying a submission. They’re buying certainty, insight and reassurance.
AI trap #3: Clients value confidence, not time
The most valuable practices will translate AI efficiency into defined service tiers.
A basic tier might offer clean records, core compliance and exception lists.
A higher tier might add monthly commentary, board-pack narratives and proactive client questions.
A premium tier might add always-on monitoring, predictive cashflow conversations and regular management insight.
This isn’t artificial upselling. It’s a clearer expression of value that was previously hidden inside labour.
The pricing language should shift from activity to outcome: speed, certainty, unlimited questions within scope, proactive insights, predictive capabilities and tiered service packages.
Clients don’t wake up wanting a journal review; they want fewer surprises. They don’t value a beautifully reconciled control account in isolation; they value confidence that the numbers are reliable and that someone will tell them what needs attention.
AI trap #4: Governance makes it sellable
Some firms see AI governance as a brake on innovation. In reality, it’s what makes the service sellable.
The direction from professional and regulatory guidance is consistent: use AI with accountability, data protection, documentation, testing, explainability, and human oversight.
Accountants and bookkeepers already work this way in tax, payroll, bookkeeping, accounts preparation and assurance-adjacent processes. AI simply needs the same discipline.
An AI File should sit behind every material AI-supported service.
It records the purpose of the tool, the data used, the prompt structure, the version of the model or software, known risks, testing evidence, accuracy over time, reviewer notes and final sign-off.
That file protects the client, the firm, and the practitioner. It also helps insurers, professional bodies and clients understand that the firm has not delegated judgement to a black box.
AI trap #5: Old roles stay, but new ones arrive
The AI-enabled practice needs new responsibilities, even if the same person wears several hats.
The AI librarian manages approved prompts and templates. The model risk approver signs off tests and known failure points. The workflow owner keeps the human and AI sequence efficient. The data quality lead ensures clean inputs. The client communication lead turns output into plain English advice.
These roles protect both quality and profitability because they stop the team reinventing the same prompt, review, and client explanation every week.
Final thoughts
Don’t lead with “we use AI”.
Lead with what the client now receives: faster turnaround, clearer explanations, stronger controls, more proactive questions and fewer surprises.
The practices that learn this pricing shift will convert AI into margin, capacity, and client loyalty. The practices that don’t will still save time, but they may discover that the time saved has become someone else’s discount.
Frequently asked questions
Price AI-assisted services around client outcomes rather than hours saved. Build packages around certainty, proactive insight, speed, exception reporting, cashflow visibility and review conversations. The fee should reflect the value of the service, including human oversight, not just the reduced time spent producing it.
Not automatically. AI may reduce manual admin, but firms still carry responsibility for review, accuracy, controls, data protection and client advice. If AI enables a better service with clearer insight and faster turnaround, the value can increase even if the production time falls.
Value-based pricing links fees to the result the client values, such as reliable records, confidence, compliance, decision-ready information and fewer surprises. For bookkeepers, this can mean moving from hourly or task pricing to tiered packages that include reporting, commentary and proactive support.
At a minimum, a firm should define approved tools, data rules, prompt templates, review responsibilities, audit trails, testing evidence, known limitations and sign-off. An AI policy and AI file help show that humans remain responsible for judgement and client-facing output.
AI can draft commentary, identify anomalies, prepare client questions, summarise movements and turn financial data into plain English. The practitioner then adds context, prioritisation and commercial judgement. This can make advisory conversations more regular, clearer and easier to scale.
Download the free AI action workbook
No sales pitch. No jargon. Just practical steps you can actually use.