── ── Tax & compliance
Engagements Don't Stall on Tax Law. They Stall on Paperwork That Never Arrives.
August 16, 2026 · 7 min read · By Brad Ju
A flat list of twenty missing documents produces either paralysis or the five easiest items. The fix is to rank by what each item blocks rather than by what it is, name a substitute for every blocker, and escalate on a fixed clock instead of on how the preparer feels that week.
Ask a practitioner what went wrong on the engagement that ran three weeks past the date and the answer is almost never a technical question they couldn't resolve. It's paperwork that never showed up, followed by six weeks of polite, unstructured follow-up that everyone involved learned to ignore.
That's a process failure with a specific shape, which means it has a specific fix. We've published the one our practice uses as an open-source skill — MIT, no account, runs in any model — and this is the reasoning inside it.
A list of twenty items is the problem, not the record of it
The default artifact is an alphabetical list of everything outstanding. It produces one of two outcomes, both bad: the client looks at twenty items, reads it as a project rather than a task, and schedules it for never — or they do the five easiest and consider themselves done. Neither unblocks the engagement, because easy and blocking are unrelated properties.
So the first move is to re-sort the list by what each item blocks rather than by what it is.
| Class | Meaning |
|---|---|
| BLOCKER | Cannot start or cannot finish without it |
| COMPLETENESS | Can proceed; nothing is signed off until it lands |
| NICE-TO-HAVE | Improves the outcome, does not gate delivery |
There's a useful gate attached to this. If more than roughly ten items come back BLOCKER, the classification isn't the problem — the engagement is mis-scoped or the client isn't ready to be served yet. That calls for a scoping conversation, not a longer list.
The column most firms skip
For each item, record what it is in the client's language, one sentence of why it matters, its class, the institution it comes from, and a substitute. The substitute column is the one almost nobody keeps, and it resolves more stuck items than any amount of follow-up.
A client who genuinely cannot find a document can very often produce a statement, a portal export, or a letter that serves the same evidentiary purpose. Naming that alternative in the original request converts "I can't find it" from a dead end into a second option — before the client has spent two weeks not replying because they're embarrassed about the first one.
The "why" column decides whether anything arrives
"Required for the return" produces nothing; it's a category, not a reason. "Without this we can't support the deduction if it's questioned, so we'd have to leave it off" produces documents, because it names a consequence the client can feel. Stated once, neutrally, as a fact rather than a threat — and not repeated, because repetition converts a fact into pressure and pressure gets filtered.
The request itself has five properties, all required: short (blockers only, under ten items and ideally under five), specific ("your mortgage interest statement from the lender for that year," not "mortgage documents"), dated with a real date that has working room before the true deadline, consequence stated once, and one action — one channel, one reply, one destination. Then cut it by a third. Length reads as difficulty.
Escalate on a clock, not on a feeling
The part that actually changes outcomes is that escalation runs on fixed intervals with rising specificity and a changing channel — never the same message twice.
| Stage | Timing | What changes |
|---|---|---|
| 1 · Request | Day 0 | Blockers only, dated, single channel |
| 2 · Reminder | +5–7 days | Same list, shorter; lead with what's outstanding |
| 3 · Narrow | +5 days | Cut to the two or three that matter; offer substitutes |
| 4 · Channel change | +3 days | Email to phone, or reach the actual document holder |
| 5 · Decision notice | Deadline minus buffer | State the decision and the date you make it |
| 6 · Act | At that date | Extend, deliver documented-incomplete, or disengage |
Two gates keep the ladder honest. Never restart it silently: partial delivery followed by silence leaves you at stage 3 for the remainder, not back at stage 1. And stage 6 has to actually happen. A ladder with no terminal action teaches the client that your dates are decorative, and next season is measurably worse.
One more piece of the frame that's easy to miss: identify who actually holds the documents. It's frequently a bookkeeper or a spouse rather than the person who signs the return, and a chase aimed at the signer can run its full length without ever reaching the person who could have ended it on day one.
The log is the deliverable nobody asked for
Every disposition gets recorded: proceeded complete, proceeded with substitutes (what was accepted for what, and why it was adequate), proceeded incomplete (item, impact, client instruction, acknowledgement), extended, or disengaged. The chase log is the evidence — what was asked, when, how often, what the client said. If a delivered engagement is ever questioned, that record is the difference between a defensible position and a recollection.
What the reordering actually finds
A rental client, six weeks from the filing deadline, twenty-two open items, three rounds of "just checking in" that produced nothing. Reclassified: three blockers, eleven completeness, eight nice-to-have. The narrowed request went out with substitutes named, and the closing statement arrived within a day.
The finding wasn't a missing document, though. One of the three blockers was a K-1 that had not been issued at all — a fact the twenty-two-item list had been hiding for a month, and one that changed the plan from chase to extend. The stage-5 notice went out the day that was discovered instead of six weeks later. Ranking by blocker surfaced it; the flat list buried it.
Use it or fork it
Client Document Chase is one of four CPA/EA practice skills we open-sourced today under MIT at github.com/deciqAI/knowledge-skills — drop the folder into .claude/skills/, or paste the Markdown into any model as a system prompt. They're process shells by design: they carry the judgment structure and deliberately carry no thresholds, rates, or filing dates, because anything that expires is a liability inside a prompt file. Educational content — not tax advice.
FAQ
Why does a long list of missing documents produce nothing?
Because twenty items reads as a project and gets scheduled for never, while three reads as a task. A client working from a long list also tends to complete the easiest items, and easy has no relationship to what's blocking the engagement.
How should missing items be prioritized?
By what each one blocks, not by what it is. Classify every item as BLOCKER (can't start or finish), COMPLETENESS (can proceed, nothing signs off until it lands), or NICE-TO-HAVE, and put only blockers in the first request.
What is a document substitute and why does it matter?
An alternative that serves the same evidentiary purpose — a statement, a portal export, a letter from the institution. Naming one for every blocker up front converts "I can't find it" from a dead end into a second option, and it resolves more stuck items than additional follow-up does.
How often should you follow up on a document request?
On fixed intervals rather than on how the engagement feels — roughly a reminder within a week, a narrowed ask a few days later, then a channel change, then a decision notice before the deadline. What escalates is the specificity and the channel, never the volume.
What if the client still doesn't respond?
Take the terminal action on the date you stated: extend, deliver documented-incomplete with the client's acknowledgement on record, or disengage. Skipping it teaches the client that stated dates are optional, which makes every subsequent season harder.