The Post-Extension Window: How to Sequence Firm Automation Before Busy Season

By Emre Benian.

The October 15 extension deadline has passed, and for most firms the next few weeks are the quietest stretch they will see until spring. That makes them the last good window to change how work moves through the firm. Change it in late February and staff learn a new process while the phones ring and the review queue grows. Change it now and they learn it on lighter work, with time to fix the rules before volume arrives.

In his October 13 piece on tool stacking, Guilherme Petersen described automation workarounds that only one or two people understand. That is the risk of automating in a hurry: a connection someone built on a Saturday becomes something the firm depends on and nobody can explain. The answer is not to avoid automation. It is to sequence it, one workflow at a time, with an owner and a way to tell whether it worked.

Start with one workflow

Pick the workflow with the clearest rules and the most wasted hours. For most tax and bookkeeping practices that is document requests for individual returns. Staff send the organizer, then spend weeks sending reminders by hand, checking the portal and calling clients who are sure they uploaded everything. The missing item is often a single brokerage statement that holds up the whole return.

The rules here are easy to write down. A request list goes out per client, built from last year’s uploads. Each new file is checked against the list. Reminders go out on a schedule the firm sets, and each one names the specific documents still missing instead of asking the client to check the portal. That one change, naming the missing item, does more than any amount of reminder frequency.

Then set a stop rule. After a fixed number of reminders, the job moves to a person with a short summary: what is missing and how many times the client was asked. Some clients need a call from someone they know, and no reminder sequence replaces that.

Run the new workflow on a subset of clients first. A few dozen engagements will show where the rules are wrong long before they could cause trouble across the whole client list.

Check what your practice tool exposes before designing anything

Most firms already pay for practice management software that tracks jobs on a board. The software knows what stage a job is in, but someone still has to move it, and the triggers usually live elsewhere: in the tax software, the e-signature tool, the payment processor or the client’s inbox.

Before anyone designs an automation, find out what your product’s API and webhooks allow. Some tools let an outside system read and update jobs directly. Others limit that, so automation has to work through email, portal notifications or scheduled exports. That one fact decides much of the design and much of the cost, and it is far cheaper to learn it before a build than halfway through one.

Use AI where it reads, not where it decides

AI document processing is useful at the checking step. A model can read an uploaded file, decide whether it is a W-2, a 1099-B or a property tax bill, and pull the payer and tax year, so staff do not have to open every file to know what arrived.

It will also misread some documents, especially photos of crumpled paper and consolidated brokerage statements. Low-confidence results should go to a person, and nothing a model extracts should flow into a return without a preparer seeing it. The judgment stays with the firm.

Measure before you change anything

Pull a few numbers from your current tools now, while nothing has changed:

median days from organizer sent to documents complete;
the share of jobs marked waiting on client;
the age of items in the review queue;
inbound status calls in the busiest weeks;
days from a finished return to a signed authorization and a paid invoice.

Check them again after one full cycle. If waiting on client has not shrunk, change the reminder wording or schedule. If status calls have not dropped, the stages on the board are probably not being kept current. Without the before numbers, the firm is left arguing from impressions about whether the effort paid off.

Add status updates only when the board is current

Once document requests run well, the next candidate is often status messages. When a job changes stage, a short message goes out: documents received, in preparation, one question waiting on you, ready for your signature, filed. Clients get the answer before they call, and preparers stop being interrupted to read the board aloud.

This works only if stages are updated as work is finished. If preparers move jobs in batches on Friday afternoon, clients get stale updates and the calls continue. Part of the rollout is making the stage change part of finishing the task, which is a habit, not a technology.

Exceptions stay with people. A return going on extension, a notice from a taxing authority or a balance due much larger than last year should be explained by the preparer or partner. Automation can flag those cases and draft a note, but a person should send it.

Give every automation an owner

The fix for fragile workarounds is ownership. Each automated workflow needs a named owner in the firm, a short written description of what it does and what it never does, and an admin login the firm holds, not one that lives with a single employee or an outside builder. Build in accounts the firm owns, so the bills come to the firm and the workflows keep running if a person or a vendor leaves.

When not to automate yet

Some firms should wait. If there is no practice management tool yet, implement one first, because automating around spreadsheets and email folders gives the automation nothing reliable to read or update. With a small client list, the built-in reminders in most practice tools may be enough. And if partners do not agree on the stages and who owns each one, settle that in a short process session before anything is built. Automation makes an agreed process faster. It does not create agreement.

The weeks between now and the first organizers are short. One well-run workflow, measured and owned, is worth more by February than five half-finished ones.

===

Emre Benian is the founder of Benian Technologies, an AI implementation partner in Chicago that helps small and midsize firms find where they lose time, build the fix and train their teams to run it. benian.ai