Most Accounting Firms Don't Have an AI Problem. They Have a Workflow Problem.

AI can help teams review information, identify patterns and prepare drafts. But those benefits are difficult to realise when the underlying process is inconsistent, relies on disconnected systems or changes depending on who completes it.

A workflow that only works when one person runs it is unlikely to become easier when AI is added.

Before introducing AI into another process, put the workflow through these five questions.

1 - What workflow problem are you trying to solve?

A useful AI project should start with a workflow problem, not a tool.

Take a recurring process such as onboarding a client, gathering compliance information, preparing tax documentation or collecting signatures. Look at where work slows down, where information goes missing and where tasks are repeated unnecessarily.

For example, a firm may want AI to improve how it gathers information from clients. But what's actually causing the delay?

• Is information arriving through multiple channels?
• Is someone manually consolidating it?
• Are follow-ups needed because information is incomplete?

Once the problem is clear, it's easier to determine whether AI is the right solution.

If the use case can't be explained without mentioning a product, the firm may be starting with the solution rather than the problem.

2 - Is your process consistent enough to support technology?

AI needs a consistent process to support.

Before introducing new technology, ask:

• Do team members follow the same process?
• Are responsibilities and handovers clear?
• Are exceptions handled consistently?
• Does information arrive in a usable format?

As Melissa Voss, Co-Founder of FuseWorks, puts it:

Many firms are trying to solve process problems with AI before fixing them with consistency.

A simple test is to give the process to someone unfamiliar with it. If they can't work out what to do, when to do it or who owns the next step, the workflow probably isn't ready for more technology.

3 - Which parts need AI and which need automation?

AI and automation are suited to different types of work.

AI is useful when work involves interpretation, analysis or generating a first draft.

Automation is better suited to repeatable actions such as:

• Sending reminders
• Updating statuses
• Routing information
• Filing completed documents

For example, AI might help identify missing client information, while automation manages reminders, document routing and signature collection.

Document signing is well suited to standardisation because reminders, tracking and document handling can follow a repeatable process. Solutions such as FuseSign help firms connect signing with the wider workflow while reducing the manual effort involved in sending, chasing and tracking documents.

The goal isn't to choose between AI and automation. It's to use each where it adds the most value.

4 - Who owns the process?

Testing a tool is easy. Getting a team to use it consistently is much harder.

Every pilot needs someone responsible for:

• Testing the use case
• Documenting how it should be used
• Capturing issues and exceptions
• Helping the team follow an agreed process

Without ownership, people naturally develop their own shortcuts and workarounds. Valuable experimentation can quickly become another source of inconsistency.

The people doing the work should also be involved early. They're often the first to identify gaps, exceptions and practical issues that others miss.

5 - What evidence would justify continuing?

A pilot shouldn't just answer whether a tool works.

It should answer whether the workflow is actually better.

The pilot also needs clear boundaries. Firms should define what information can be entered, where human review is required, who remains accountable for the result and how inaccurate outputs will be handled.

Before testing begins, establish a baseline. Depending on the workflow, that might include:

• Completion time
• Delays between stages
• Missing information
• Rework
• Client turnaround times

At the end of the pilot, assess whether it reduced friction, improved consistency, saved time or introduced new risks.

If the answer isn't clear, refine the use case before expanding it further.

Start with one workflow

Accounting firms don't need to introduce AI across the entire practice overnight.

Start with one workflow and one clearly defined problem. Determine which tasks require judgement, which follow a rule and where technology can genuinely help, then test, measure and refine the approach.

The goal isn't to use AI wherever possible. It's to build a better workflow and apply the right technology at the right stage.

See How FuseSign Works See How FuseSign Works