Chasing client records is the most expensive work an accountancy practice does, and almost none of it is billable. It is also, unusually, work that can be automated end to end without touching a single professional judgement — which makes it the safest place for a practice to start.
The deadline is unforgiving in a way that shapes everything. HMRC requires an online Self Assessment return by 11:59pm on 31 January, with a late filing penalty if it misses. That same date is the payment deadline. Paper returns are due by 11:59pm on 31 October, and where payments on account apply there is a second date of 31 July. None of those move because a client was on holiday.
Why the chase costs more than it looks
The visible cost is time — someone drafting a fourth reminder in the third week of January. The real costs sit underneath it.
It concentrates. The work does not spread evenly across the year. It compresses into the weeks where the practice has least capacity, which is precisely when a senior's hour is worth most.
It is unevenly distributed. A small number of clients generate most of the chasing. They are rarely the largest fees.
It depends on someone remembering. In most practices the chase lives in a person's head, supported by a spreadsheet that is accurate on the day it was updated. When that person is ill in mid-January, the practice discovers how much was undocumented.
It degrades the relationship. By the fourth reminder the tone has changed. The client hears nagging; the practice feels ignored. Neither is a good position from which to discuss next year's fee.
What automating the chase actually means
It does not mean sending more email, faster. A practice that automates a bad process gets a worse one at scale.
It means the practice stops being the thing that remembers. Concretely, a system that:
- Knows what is outstanding per client. Not "has this client sent their records" but which specific documents are missing — the bank statements, the dividend voucher, the rental income schedule.
- Escalates on a schedule rather than on a mood. A defined sequence, with the tone and channel changing as the date approaches, agreed once rather than improvised each January.
- Stops the moment a document arrives. This is the part practices most often get wrong, and the one clients notice hardest. Being chased for something you sent last week destroys confidence in the whole system.
- Makes the state visible. Anyone in the practice should be able to see where every client is without asking the person who normally handles it.
Sequencing that does not annoy people
The sequence matters more than the tooling. A workable shape, adjusted to the practice:
- Early, specific request. Not "please send your records" but a named list of what is needed for this client. Generic requests get generic responses, which is to say none.
- A reminder that assumes good faith. Most non-response is not refusal; it is an email that arrived on a bad day. Restate the list.
- A reminder that introduces the consequence. Neutral and factual — the filing date, what happens if it is missed. It does not need to be dramatic to be effective.
- A change of channel. If email has not worked three times, a fourth email will not work either.
- Human escalation. At a defined point a person takes over, with the full history visible so the conversation starts from where the automation left off, not from the beginning.
The thing that makes this work is step five being a deliberate handover rather than a failure. Automation handles the sequence; a person handles the exception. That is the correct division, and it is the one that also keeps the relationship intact.
What stays with a person
Worth being explicit, because this is where automation projects overreach.
Anything requiring judgement about the client's affairs stays with a qualified person. Whether a receipt is allowable, whether an unusual transaction needs a conversation, whether a client's reluctance to provide something is disorganisation or a signal — none of that is a workflow problem. The automation collects and sequences; it does not interpret.
The same applies to the decision to disengage. A system can tell you a client has missed every deadline for two years. It should not tell them.
Start before the pressure arrives
The worst time to build this is January, which is when most practices decide they need it. By then the deadline is close enough that any new process is a risk, and the people who would set it up are the people with no time.
The work is better done in the quiet months. It also compounds: a practice that has automated the annual chase has already built most of what it needs for quarterly reporting under Making Tax Digital, where the same sequence simply runs more often.
We cover the wider picture — MTD, onboarding, where the website fits — on our AI for accountants page, and the intake side of it under services.
The short version
Record chasing is repetitive, sequenced and free of professional judgement, which makes it the highest-return and lowest-risk automation available to an accountancy practice. Automate the sequence, stop it dead when documents arrive, hand exceptions to a person with the history intact — and build it in June rather than January.