The Draft That Writes Itself: AI in Client Communication
Where AI genuinely saves time in an accounting practice is the everyday writing — status updates, chase-ups and query emails. Here's how to use it without losing your voice or your judgement.
Most conversations about AI in accounting jump straight to the dramatic stuff: reading financial statements, spotting anomalies, drafting tax positions. That's the wrong place to start. The biggest, safest time saving in a typical firm isn't in the numbers at all — it's in the writing that surrounds the work.
Think about how much of your week goes into words rather than figures. The email explaining why a BAS refund is smaller than expected. The polite nudge to a client who still hasn't sent through their motor vehicle logbook. The engagement summary that translates your work item board into plain English for a director who doesn't care about job codes. This is where AI earns its keep quietly, every single day.
Why communication is the right first job for AI
Client communication has three qualities that make it ideal for AI assistance. It's high volume, it's repetitive in structure but variable in detail, and it carries low risk when a human reviews the output before it goes out.
Compare that to lodgment. Your tax return software handles the return itself, and you would never let an AI file it unchecked — the stakes are too high and the margin for error too small. But the covering note that explains the outcome to the client? That's a paragraph you rewrite in slightly different form fifty times a year. Perfect candidate.
The pattern to look for is any writing task where the facts already exist in your accounting client management software and the job is simply to shape them into readable prose. The AI isn't inventing anything. It's assembling.
Four communication jobs worth automating first
1. Status updates that don't require you to write from scratch
Clients rarely ask "what's the WIP position on my file" — they ask "where's my tax return up to?" The answer already lives on your board: the job is at review stage, waiting on one outstanding query. AI can turn that structured status into a warm, human sentence in seconds. When your practice data and your drafting tool sit in the same place, this becomes a one-click reply rather than a context-switch into a blank email.
2. Chase-ups that stay polite on the tenth attempt
By the fourth reminder for the same missing document, tone tends to slip. AI is genuinely good at keeping a chase firm but courteous, and at varying the wording so it doesn't read like a robot copy-pasting itself. Feed it the outstanding items and the deadline, and let it produce a draft you can send as-is or sharpen in ten seconds.
3. Query emails during a job
Mid-job questions — a clarification on a large expense, a request for the missing bank statement — are formulaic but need to be specific. AI drafts the ask clearly; you add the judgement about what actually matters and hit send.
4. Onboarding and engagement explainers
New clients need the same core information: what you'll do, what you need from them, what happens next. Rather than pasting from a stale template that no longer matches your process, AI can generate a tailored summary from the engagement details already captured in your client accounting software, matched to the specific package they signed up for.
The rule that keeps this safe: draft, don't send
None of this works if the AI has the final word. The model should always produce a draft that a human reads before it reaches a client. That's not a temporary limitation to be engineered away — it's the whole point. The person retains responsibility for accuracy, tone and professional judgement; the machine removes the friction of the blank page.
A few guardrails make this reliable in practice:
- Feed it facts, not guesses. AI drafting works best when it's summarising real data from your system, not filling gaps. If the deadline, the outstanding items and the client's name are pulled from structured records, there's nothing for it to hallucinate.
- Never let it state a tax position or a figure it wasn't given. Keep AI on the wrapper — the covering note, the reminder, the explanation — and keep the substance coming from your work and your review.
- Review every client-facing message. A quick read is enough. The time saving is in not writing from scratch, not in skipping the check.
- Keep your firm's voice. Give the tool a sense of how your firm sounds — plain, warm, no jargon — so drafts land close to send-ready.
Why this belongs inside your practice system, not a separate tab
You can absolutely paste details into a general chatbot and get a decent email back. But the moment you're copying client names, deadlines and job details out of one system and into another, you've created friction, a privacy exposure, and a fresh chance for the wrong client's details to end up in the wrong message.
The value multiplies when the AI sits inside the same tool that already holds your client accounting data. In Finye, the AI can see the work item, the outstanding document requests, the compliance deadline and the client record — so a status update or a chase draft is built from live context, not retyped by you. That's the difference between a novelty and a genuine part of the workflow. It's the same reason strong account practice management software keeps records, boards and communication under one roof: every hop between systems is a hop where time and accuracy leak away.
Start small and measure the hour, not the hype
You don't need an AI strategy to begin. Pick the one message you write most often — for most firms it's the document chase or the "where's my return" reply — and let AI draft it for a fortnight. Track roughly how long each one used to take and how long it takes now.
If you save five minutes on twenty messages a week, that's over eighty hours a year returned to the practice, on a task nobody enjoyed anyway. No dramatic transformation, no replacing anyone's judgement — just the boring, valuable win of never starting from a blank page again.
The firms that get real value from AI this year won't be the ones chasing the flashiest use case. They'll be the ones who quietly automated the writing they were already doing, kept a human in the loop, and got their evenings back.