# Workflow Handoff Audit ### Audit Reveal where repeat work stalls, loses context, or creates rework. --- ## Purpose Most AI redesign efforts fail not because the tool was wrong, but because no one mapped the workflow before automating it. This audit forces that mapping. Run it on one recurring workflow before adding AI tools, prompts, or automation to any part of it. --- ## Before You Start Select one workflow. Not a department, not a function, one specific recurring workflow with a clear start and end point (for example: incoming support ticket to resolution, new hire request to first day, draft content to publish). [ASSUMPTION] This audit assumes the workflow repeats often enough (at least monthly) that the time spent auditing it pays back. For a workflow that runs a few times a year, a lighter-weight conversation may be more proportionate than a full audit. Involve the people who actually do the work at each step, not only the manager who owns the workflow on paper. The manager's description of the workflow and the actual practice on the ground frequently differ, and the gap between them is often where the real problems live. --- ## The Seven Audit Questions Work through these in order. Write down the actual answer from the people doing the work, not the idealized version from the process documentation. ### 1. What starts the work? Name the specific trigger: a form submission, an email, a calendar date, a system alert, a verbal request. Vague triggers ("when someone needs it") are themselves a finding. If the start of the workflow is inconsistent or undocumented, that is the first gap to close, before anything else. ### 2. Who owns each step? List every step and the named person or role responsible for it. A step with no named owner, or with two people who each assume the other owns it, is a common source of stalled work. Flag every step where ownership is unclear. ### 3. What information must be available at each handoff? For each point where work passes from one person or system to another, list what the receiving party needs to proceed without asking follow-up questions. Compare this to what is actually provided today. Gaps here are where context gets lost and rework begins. ### 4. Where does rework occur most often? Ask the people doing the work directly: what gets sent back, redone, or corrected most often, and at which step. This is usually known precisely by the people closest to the work, even when it has never been written down or reported upward. ### 5. What quality check happens before work moves forward? Identify the actual check performed at each step, not the check described in a policy document. Distinguish a real check (someone verifies against a defined standard) from a nominal one (a checkbox clicked without review). Steps with no real check are the highest-risk points for introducing AI-assisted work without oversight. ### 6. Which step could AI support without removing accountable human judgment? For each step, ask whether AI assistance would speed up or improve the work while leaving a human accountable for the outcome. This is different from asking whether AI could complete the step alone. A step that requires relationship history, negotiation, legal judgment, or client-specific nuance is rarely a candidate for full automation, even if it looks procedural on the surface. ### 7. How will the team know whether the redesigned workflow improved? Define the specific measure before making any change: cycle time, rework rate, error rate, or a quality score. Without a defined before-and-after measure, no one can tell whether a redesign helped, and improvised improvements tend to survive on anecdote rather than evidence. --- ## Documenting the Findings For each step in the workflow, record: | Step | Owner | Info Needed at Handoff | Info Actually Provided | Rework Frequency | Real Quality Check? (Y/N) | AI-Suitable? | |---|---|---|---|---|---|---| | | | | | | | | Fill this table from direct observation and interviews with the people doing the work, not from the workflow's original design document. The design document tells you the intended process. This table tells you the actual one. --- ## Reading the Results - **Steps with no named owner** are handoff failures waiting to happen. Assign an owner before doing anything else with that step. - **Steps with high rework frequency and no real quality check** are the highest-priority candidates for redesign, with or without AI. Adding AI to a step that already lacks a quality check will scale the error, not fix it. - **Steps marked AI-suitable** should still pass through a human review checkpoint if the output affects a customer, client, employee, or carries legal, financial, or brand risk. AI-suitable does not mean unsupervised. - **A workflow with an unclear start (Question 1) or widespread ownership gaps (Question 2)** is not ready for AI redesign yet. Fix the structural gaps first. Automating an undefined process produces a faster undefined process. --- ## Before Automating or Adding AI Prompts Do not proceed to tool selection or prompt design until: - [ ] Every step has a named owner - [ ] Handoff information requirements are documented and gaps are closed - [ ] The step or steps with the highest rework rate are identified - [ ] A real quality check exists, or is planned, at every step retained after redesign - [ ] A specific before-and-after improvement measure is defined and a baseline is recorded --- ## Notes and Assumptions [UNVERIFIED] Rework frequency and cycle time figures should come from your own records or direct observation over a defined period, not estimation from memory. Recollection of "how often" something goes wrong is frequently inaccurate in either direction. --- ### Related WenceStudio Resources Eliminate handoff friction with tested multi-node automation blueprints: 1. **Self-Hosted n8n AI Agent Workflow Vault** (Automation Blueprint Vault) Four tested automation blueprints for self-hosted n8n instances covering inbound enrichment, support routing, content repurposing, and invoice parsing. - [Open n8n Blueprint Guide](https://docs.google.com/document/d/1zzEsEBgfjdP3qhIbFP1J0Yy_SnKpcDDikXQwQBhziSo/edit?usp=sharing) 2. **The AI Lead Router Automation Blueprint** (Blueprint & Make.com JSON) An importable Make.com automation that catches inbound webhook leads, scores intent, and routes qualified prospects directly to your CRM. - [Download Setup Guide (PDF)](https://drive.google.com/file/d/1Dh_nHjbf1_2RtIPfM_pbDzL5_cPluTYr/view?usp=sharing) - [Get Make.com Blueprint (.json)](https://drive.google.com/file/d/1WVn7Kbv2cg8AbTlXRLfST6QiRCOwbLjD/view?usp=sharing)