The short answer
Start with one repetitive, low-risk task your team does every week, not with a tool. Document it, time it, pilot AI on it for 30 days, measure the hours saved, then scale. Firms that start with 'let's get everyone ChatGPT' usually stall.
Key takeaways
- Pick the work first, then the tool.
- Your first pilot should be high volume, low risk and easy to check.
- Write a simple AI policy before anyone pastes client data anywhere.
- Measure hours before and after, or you'll never know if it worked.
Why most firms stall
I talk to a lot of firm owners who've 'tried AI'. Usually that means a few people signed up for a chatbot, used it for a couple of emails, and then went back to doing things the old way. Nothing was wrong with the tool. There just wasn't a job for it to do.
AI is a capability, not a strategy. If you don't point it at a specific piece of work, it becomes another login nobody uses.
Don't ask 'how do we use AI?' Ask 'what are we sick of doing by hand?'Hayden Walsh
Step 1: Find the boring, repeatable stuff
Get your team in a room for 45 minutes and list every task that is repetitive, text-heavy and annoying. Common ones in firms:
- Chasing clients for missing information
- Writing variance and review commentary
- Turning meeting notes into actions
- First drafts of client letters and emails
- Writing and updating SOPs
Next to each one, write a rough estimate of hours per month across the firm. You're not after precision. You're after the obvious big rocks.
Step 2: Choose your first pilot
Your first AI pilot should tick three boxes: it happens a lot, a mistake isn't catastrophic, and a human can easily check the output. Client email drafts and meeting summaries are great first pilots. Tax positions are not.
Step 3: Set the guardrails
Before you roll anything out, decide which tools are approved and what client information can go into them. This doesn't need to be a 40-page document. A two-page policy your team actually reads beats a perfect one nobody opens.
Step 4: Pilot, measure, decide
Run the pilot for 30 days with a small group. Time the task before and after. At the end, make a call: scale it to the whole team, fix what's not working, or drop it and try the next task on your list.
One workflow done properly will teach your firm more than ten half-tried tools.
Common mistakes
- Buying a tool before deciding what job it's for.
- Letting everyone use whatever AI tool they like with client data.
- Starting with the highest-risk work because it's the most impressive.
- Not measuring time before the pilot, so there's no way to show the win.
Questions accountants ask
How long before we see time savings?
With a well-chosen pilot, most firms see a noticeable saving on that task within 2 to 4 weeks.
Do we need a consultant to start?
No. You can absolutely start yourself with the steps above. Where outside help pays off is speed, avoiding the common traps, and implementing across the whole firm.
What if my team is sceptical?
Good. Sceptics make great testers. Involve them in the pilot and let the time savings do the convincing.

Chartered Accountant · Founder, Off Balance
Chartered Accountant with a background at Deloitte and Xero, plus years in sales, leadership and startups. Hayden helps accounting firms across New Zealand and Australia work out where AI fits, then actually put it to work.
More about Hayden →