From the first week of January to the fifteenth of April, a tax practice runs two operations at once. One prepares returns. The other one — the one that actually sets the pace — chases documents, because a return cannot move until the client's K-1 shows up, and the client is confident it is coming any day now.
The bottleneck is the client, and it peaks in February
The rhythm repeats every year. W-2s and most 1099s land in late January. Brokerage 1099s arrive in mid-February and then get corrected in March, which is why a return prepared too eagerly gets amended. K-1s come from partnerships that are themselves on extension, so a client with two partnership interests may not be complete until September no matter what anyone does. Between those dates, staff time goes into an expensive activity: emailing grown adults to ask whether they have checked their mail.
The same pattern runs outside tax season in the bookkeeping and advisory side of the house — the month-end close that stalls waiting on bank statements, the missing receipt that holds a reconciliation open into the second week. Document collection is not incidental to this work. For most small firms it is the largest controllable variable in when work ships and how much overtime the season burns.
Automating the chase without automating the relationship
Last year's return defines this year's checklist. A client who had a brokerage account, two rental properties, and one partnership interest last April has a knowable document list this January, before anything arrives. The mechanism is plain: generate the per-client checklist from the prior year, mark items off as they land in the portal, and send reminders only for what is still missing.
Two design rules decide whether this helps or hurts. First, reminders must stop the moment a document arrives — nothing burns goodwill faster than being nagged for something you already sent. Second, replies must land where a person actually looks. A client who answers "my broker says March 3rd" has given the firm real information, and a system cannot act on it, but a human has to see it the same day. Automated collection with an unwatched reply channel is worse than no automation, because the client believes they have told you something.
Engagement letters before any work moves
The engagement letter renews annually, and it is supposed to be signed before work begins — professional-liability carriers and malpractice case law agree on why. Firms start work on unsigned engagements every season anyway, because tracking who has signed is a spreadsheet chore that loses to everything more urgent. This is exactly the kind of dull, mechanical work automation is good at: send the letter, track the signature, remind, and flag any client whose return has entered preparation without a signed letter on file. The flag is the valuable part. It converts a quiet liability into a visible one.
The four hundred questions
Inside the firm, the season is a stream of small lookups. How did we treat the vehicle depreciation for this client last year. What is our standard footnote for a late K-1. Which clients have a June 15 estimate and have not been reminded. The answers exist — in prior-year workpapers, in firm memos, in the tax software — and getting them usually means interrupting the one person who knows.
Retrieval over the firm's own documents turns those interruptions into lookups. Staff ask in plain language; the system finds the relevant passage in the firm's own files and shows it, with a citation to the source document, and the staff member who asked reviews it before relying on it. That review step is a design requirement, not a courtesy. The output is an internal draft for a professional, every time.
The line that does not move: nothing goes out as advice
There is a professional reason and a legal one. The professional reason: the firm signs the return and the firm answers for it, so a generated answer is raw material for a person, never a deliverable to a client. The legal reason is sharper. Section 7216 of the Internal Revenue Code makes it a crime for a preparer to knowingly or recklessly disclose tax return information outside the preparation of the return, and Section 6713 adds civil penalties that do not require intent.
That is why the architecture matters more than the model. Client records belong in an index running on infrastructure the firm controls — not pasted into a consumer chatbot whose terms permit retention, and whose logs, as one 2025 copyright case demonstrated, can be frozen and pulled into discovery in litigation the firm has nothing to do with. Keeping the records on infrastructure you control is the starting point for a compliance conversation with your own counsel. It is not a substitute for that conversation, and anyone who tells you it is has something to sell.
The calendar is a client deliverable
Quarterly estimates fall on April 15, June 15, September 15, and January 15, with state dates that do not always match. Entity deadlines stack on top: partnership and S-corp returns in mid-March, C-corps in April, payroll filings quarterly, 1099s that your business clients must issue every January and forget every January. A deadline calendar generated from each client's entity type, with reminders that go to staff and to the client, is unglamorous automation. It also prevents the expensive kind of apology, and clients experience it as the firm being on top of things — which is the product.
What this does not solve
It will not review a return, take a tax position, or shorten the preparation itself. If a prior-year workpaper was wrong, retrieval will surface the wrong answer with a confident citation, which is why the human review step is not optional. It cannot make a partnership issue K-1s sooner; September is still September. A firm with sixty returns and one partner may find the checklist inside its tax software is enough. And no system fixes a client roster that ignores every reminder — automation lowers the cost of the chase, but the chase remains.
Where to start
Pick ten returns from last season and count the emails and calls it took to complete each client's document set. Multiply the average by your client count. That is the season's chasing budget in touches, it takes an afternoon to compute, and it is the baseline any system has to beat.