Most Copilot rollouts in accountancy go wrong before anyone types a prompt. That’s the blunt assessment of Taimur Ijlal, Information Security Leader at the UK arm of discounts website Proxy Coupons. His message to adopters is to ensure, ahead of any other considerations, that the cyber environment they plan to bring Copilot into is hygienic.

“Copilot sees whatever users can already see across Microsoft 365,” he notes. “Years of loose SharePoint and Teams permissions suddenly become searchable. Payroll spreadsheets, board papers and clients’ working papers can all surface in summaries prepared for other clients.

“Artificial intelligence (AI) assistants are only as discreet as the permissions underneath them, so I’d start with an access and data-labelling review before any use-case workshop.”

Use priorities for business and practice

For Charles Radclyffe, CEO of EA Global AI, the biggest mistake that finance professionals make with Copilot adoption is starting with the technology rather than the work. “The question isn’t ‘where can we deploy Copilot?’” he says, “but ‘which parts of our workflow would benefit from being accelerated without removing human accountability?’”

Under that approach, he points out, use-case priorities vary between team types.

“In the finance department of a non-financial business, the focus is usually on productivity and operational efficiency,” he explains. “Teams typically see value in reducing time spent searching for information, producing reports and supporting month-end and management reporting activities. The most successful use cases tend to be those that sit around the financial process, rather than directly generating or posting financial data.”

In an accountancy firm, though, the emphasis is often different.

“Firms are ultimately selling expertise, judgement and trust,” Radclyffe says. As such, Copilot can help to accelerate the prep work behind client deliverables. For example, prior-year file reviews, client information summaries or technical guidance research. It can also contribute to draft communications.

However, Radclyffe stresses, the professional judgement delivered to the client remains entirely the adviser’s responsibility.

Here are some common starting points for finance teams and practices.

Finance teams: Excel analysis and reconciliations

According to New York-based Accountant and Tax Consultant Ashley Akin, month-end reporting, Excel analysis and reconciliations are “sensible starting points” for finance departments.

“Copilot can help identify unusual transactions, explain formulas, prepare variance commentary and summarise financial results for management,” she says, noting that accountants must still verify the figures, investigate discrepancies and approve the final reports. “Tax positions, revenue recognition and impairment decisions should remain with qualified professionals,” she says.

Practices: tax research and document reviews

Meanwhile, further to the areas that Radclyffe highlights, Akin says that Copilot can help firm-based professionals organise tax research, review documents for missing data and compile audit documentation.

Across finance departments and firms, she says: “Senior accountants should ask employees to track their work for two weeks, paying attention to repetitive tasks, time spent searching for data and work that requires extensive manual preparation. The best candidates are frequent tasks that consume considerable time but require relatively little judgement.”

Detecting use cases

Radcyffe takes a similar view, encouraging managers to map repetitive, time-consuming activities across the team, then assess them against business value and risk. He highlights tasks such as:

• summarising contracts,
• producing first drafts of management commentary,
• analysing large volumes of correspondence,
• preparing meeting briefings, and
• finding data across multiple documents.

These are often strong candidates, he says, “because they consume significant time while still sitting within established review processes.”

“Once your data’s in order,” Ijlal says, “plot candidate tasks by time consumed against the cost of an error. Drafting variance commentary, summarising long email chains and writing up meeting notes are good early wins. Anything feeding a tax return, audit file or published figure needs a reviewer who can trace each number back to source.”

He notes that what gets prioritised usually comes down to regulatory exposure, data quality and how comfortable staff are with checking AI outputs.

Prepare to pivot if a use case doesn’t work

In Radclyffe’s view, the factors that should guide Copilot adoption are fairly straightforward:

• the sensitivity of the data involved;
• the materiality of the output;
• the strength of existing review controls; and
• the time the candidate task currently consumes.

High-frequency, low-risk activities should be prioritised before tackling more complex processes.

However, he notes, it is important to treat adoption as an ongoing programme, rather than a one-off rollout. “If a particular use case isn’t delivering measurable improvements in quality, efficiency or user adoption,” he says, “organisations should be prepared to pivot. The most successful deployments are rarely those that identified every use case correctly on Day One. They’re usually in organisations that continuously test, measure and refine where Copilot can genuinely complement their people’s expertise.”

Akin recommends that after 60 to 90 days, accountants should compare how long tasks took before Copilot with how long they take now, including time spent checking and correcting the tool’s work. “If employees rarely use it, find out why. They may need more training, or the tool may simply not be useful for the chosen task.”

She adds: “The real cost isn’t the Copilot subscription. It’s the additional checking, corrections and potential compliance problems if the team starts trusting its output too much. If reviewing the work takes longer than preparing it manually, that particular use case needs reconsideration.”

If results disappoint, Ijlal suggests that teams run a short pilot with a small group and log every correction. “Compare hours saved against review time added. When review is swallowing the gains, narrow the scope instead of scrapping the tool and use Microsoft Purview audit logs to see what Copilot is actually pulling into its responses.”

“As a Microsoft partner,” Radclyffe says, “we typically find that the greatest returns don’t come from simply switching on Copilot. They come from helping organisations identify where human expertise adds the most value and then using Copilot to remove the administrative and information-management burden around it.

“The goal isn’t to replace professional judgement – it’s to give professionals more time to apply it.”