Published by CXJ Labs, an automation company. Not independent. Read our disclosure.
The Automation Index

Can collecting onboarding documents be automated?

Mostly automatable: 74% of task time

The answer

Mostly. Sending the request list, chasing, renaming and filing, and pulling key fields into your systems are well suited to software. AI can now check whether a document is the right type and readable. In our model about 75% of the work is automatable. Spotting real problems in the documents still needs a trained eye.

The tasks, and how much of each is automatable today

Six tasks make up document collection. Sending requests and filing are about 95% automatable. Extracting key fields and chasing are 75 to 80% automatable. Checking completeness is about 70% automatable with AI. Reviewing for real problems, such as wrong years or mismatched names, stays about 70% human.

TaskActivity typeMin / weekAutomatable
Send the document request listChecklist chosen by client type and sent on signature.Processing data
base rate 69%		4095%
Chase missing documentsReminders listing exact missing items handle most clients. Persistent stragglers get a call.Interacting with people
base rate 20%		12075%
Check each document is the right one, complete and readableAI classifies document type, period and legibility well. Partial statements and odd formats need a look.Processing data
base rate 69%		8070%
Rename and file documentsA naming rule plus classification files everything consistently.Processing data
base rate 69%		6495%
Copy key fields into the client recordAI extraction handles printed documents well. Handwritten forms need checking.Collecting data
base rate 64%		8080%
Review documents for real problemsAI can flag a name mismatch or wrong year. Deciding whether it matters needs a trained person.Applying expertise
base rate 18%		6030%
Whole process44474%

Base rates are McKinsey Global Institute's technical automation potential for each activity type. The share for each task is our adjusted estimate. Method.

Hours saved per week

For an accounting practice onboarding eight new clients a week, document collection takes about 7.5 hours a week by hand. Automating the tasks above returns about 5.5 hours a week, around 260 hours a year. Clients also finish faster when reminders name exactly what is missing.

Modelled on: A small accounting practice onboarding about eight new clients a week, collecting IDs, bank statements, prior returns and signed engagement letters by email.

  • Eight new clients a week, each sending six to ten documents.
  • Most clients need at least one reminder; many send the wrong file first.
  • Files arrive by email with names like 'scan001.pdf' and must be renamed and filed by hand.
  • Key details, such as tax ID and year end, are typed into practice software.

7.5 hours a week by hand, 5.5back with automation. Formula: the sum of each task's minutes multiplied by its automatable share, divided by 60, rounded to the nearest half hour.

Tools that do the job

You need a portal or form that accepts uploads, storage with a naming rule, and a step that checks and extracts. Typeform or a client portal collects files. Google Drive or your practice software stores them. An AI agent classifies and extracts. n8n or Make chases what is missing and updates the CRM.

Typeform
Single upload form with a checklist tailored by client type.
AI agent
Classifies each file, checks the period and legibility, and extracts key fields.
Make
Renames and files documents, tracks what is missing and sends reminders.
Airtable
Tracks each client's checklist status for the team.
Gmail
Sends reminders that list the exact items still missing.
Xero
Receives client details once collected, so they are not typed twice.

A simple workflow

The workflow starts when a client signs the engagement letter. They receive one upload link with a checklist for their situation. Each file is classified, renamed and filed. Key fields go to the client record. Reminders list only what is still missing. A person reviews the complete set before work begins.

  1. TriggerEngagement letter signed. Choose the checklist by client type.
  2. RequestSend one upload link with the checklist and examples of what each document looks like.
  3. CheckClassify each upload. Wrong type or unreadable files trigger an immediate request to resend.
  4. FileRename to a standard pattern and file in the client folder. Extract key fields to the record.
  5. ChaseEvery three days, remind the client of the specific items still missing.
  6. ReviewWhen the set is complete, a person reviews it before work starts.

Keep these human

  • The final review of the full document set. Mistakes here become mistakes in the client's accounts.
  • Calls to clients who are stuck. Some do not know which document you mean; a two-minute call fixes it.

Where it goes wrong

  • Email attachments. If clients can still email files, half bypass the system. Reply to emailed documents with the upload link.
  • Trusting AI checks as approval. Classification is triage, not review. Keep a person signing off the complete set.
  • Sensitive data in the wrong place. IDs and bank details must go to secure storage, not a shared inbox. Decide storage before building.

What most guides miss

Most guides recommend a client portal and stop there. The time sink is not collecting files; it is the back-and-forth when the wrong file arrives. Checking each upload instantly and replying within a minute, while the client still has their files open, removes most chasing. Delayed checking is what turns one request into four.

The evidence