Before You Introduce AI: Five Questions Every Accounting Firm Should Ask

Accounting firms are looking to AI to improve efficiency and create capacity without simply adding more people. But before introducing AI to your firm, there is a more important question: are you ready for it?

AI can help firms review information, prepare first drafts, and complete parts of existing workflows faster. But if a process is unclear, inconsistent or spread across disconnected systems, it may simply make the same problems harder to manage.

These five questions can help firms determine whether AI is the right solution and what needs to be in place before proceeding.

1. What problem are we trying to solve?

“Using AI” is not a business objective.

A better starting point is a specific bottleneck within an existing workflow. Where is work slowing down? Which tasks are repeated unnecessarily? What is creating delays, frustration, or inconsistency?

For example, a firm may want to use AI to review information submitted by clients and identify missing or inconsistent details before work begins. But first, it needs to understand why those issues occur. Information may arrive through several channels, require manual consolidation, and only be reviewed once the job is underway. The clearer the problem, the easier it is to determine whether AI is the right solution.

Before discussing tools, firms should be able to answer three questions:

• What is going wrong today?

• Who is affected?

• What would a better outcome look like?

If the problem and workflow cannot be explained, the use case may not be clear enough yet.

2. Is the underlying process clear and consistent?

AI cannot compensate for a poorly understood process.

Before introducing it, map the workflow from beginning to end. Identify the required inputs, key steps, decisions, exceptions, and responsibilities.

This often reveals inconsistencies that were previously hidden. Team members may complete the same task differently; information may arrive through disconnected channels, or important handovers may depend on someone remembering what happens next.

As Melissa Voss, Co-Founder of FuseWorks, explains:

“You can’t automate a shambles. You need really defined systems and processes first, then you can get people and technology to support that.”

Firms do not need to document every process before experimenting with AI. Start with the workflow connected to the proposed use case. Remove unnecessary steps, agree on a consistent approach, and clarify where professional judgement is required.

A useful test is whether a new team member could follow the workflow using the information currently available. If not, the process may not be ready for AI.

3. Could existing automation solve it more reliably?

Not every repetitive task requires AI.

Some problems may be solved more reliably through automation, integrations or functionality already available within the firm’s technology stack.

AI can be useful when work requires interpretation, summarisation, pattern recognition, or preparing a first draft. Rules-based automation is generally better suited to actions that should happen the same way every time.

Consider a workflow that begins with collecting client information and ends with sending completed documents for signature. AI might review the incoming information and identify possible gaps. Automation could then send approved documents for signature, issue reminders, update statuses, and file completed copies.

FuseSign is one example of how firms can create this type of structured, repeatable signing process without adding more manual administration.

The question is not whether AI or automation is better, but which approach is suited to each part of the workflow. The most effective workflow may combine both, with each playing a clearly defined role.

Before introducing another tool, firms should ask:

• Does the task require interpretation, or could a clear rule achieve the same outcome?

• Does suitable functionality already exist in the firm’s technology stack?

• Will another application simplify the workflow or make it more fragmented?

The goal is not to justify using AI. It is to improve work in the most practical and reliable way.

4. Who will own implementation and adoption?

Even a well-chosen use case can struggle without clear ownership.

Someone needs to coordinate the trial, establish how the tool should be used, support the team, and improve the approach after launch. This does not need to be the most senior person in the firm, but they should understand the workflow and have the capacity to lead the implementation.

Turning an experiment into an agreed way of working requires the firm to:

• involve the people who complete the work;

• test the tool using real scenarios;

• establish one approach for the team to follow; and

• collect feedback and improve the process after launch.

Without that structure, individual experimentation can create inconsistent processes and additional risk.

The people doing the work should be involved early. They are often best placed to identify awkward handovers, missing steps and exceptions that may otherwise be overlooked.

5. How will we govern, review and measure its use?

Without clear guardrails, one team member may avoid AI completely while another relies on it for work it was never intended to support.

Accounting firms handle confidential financial and personal information. Before introducing AI, they need clear rules covering:

• what information can be entered into the tool;

• where human review is required;

• who remains accountable for the final output; and

• how inaccurate or incomplete results will be handled.

The firm should also agree on how success will be measured.

The purpose of a pilot is not to prove that AI works. It is to determine whether a particular use case improves the workflow it was designed to support.

Depending on the use case, firms might measure task completion time, delays and rework, client turnaround times, output consistency and team adoption.

These measures do not need to be complicated. They need to show whether the change is reducing friction, improving consistency, or creating useful capacity.

Clear governance is not a barrier to adoption. It gives teams confidence by clarifying the tool’s purpose, its limitations, and where human oversight is required.

Start with one controlled use case

Accounting firms do not need a firm-wide AI strategy before they begin.

As Melissa Voss advises:

“Don’t think you need to come up with a grand plan first. Start small, give it a go, and build from there.”

Start with one recurring workflow, define the problem, and establish clear ownership, controls and measures. If the use case produces a measurable improvement, refine the approach and consider where it could be applied next.

AI should improve the workflow, not sit beside it as another disconnected tool. The firms that gain the most value from AI will not necessarily be the first to adopt it. They will be the firms that understand their workflows, solve the right problems, and introduce AI with clear boundaries and human oversight.