Map a workflow: find the waiting before you automate

Describe up to eight stages of one workflow. Give each stage the minutes someone actually works on it and the minutes it sits waiting. Handoff adds them up, shows the share of elapsed time spent waiting, and writes a Markdown plan you can keep. Everything happens in this browser tab: your entries stay in this tab and are not stored, and no AI model is used to produce or interpret your numbers.

An illustrative before and after

Before: the handoff is invisible

An invoice arrives by email. Someone reads it, keys it into the ledger, and emails a PDF to a manager. The manager has no queue rule, so the PDF sits in an inbox until it is noticed. Once approved, the file is emailed again to accounts payable, where it waits for the next payment run. Nobody can say how long the invoice took end to end; each person only knows their own few minutes of work. The hypothetical numbers below are invented to illustrate the idea, not measured.

6 min work, 1,900 min waiting
No stated owner between stages
Elapsed time unmeasured, so the bottleneck is a guess

After: the same work, named honestly

The team writes down four stages with an owner role on each and a written rule for when a handoff triggers. They change nothing about the software yet. Now they can see that the two approval steps together hold almost all of the waiting, while the data entry step holds most of the actual work. That is a statement about their own notes, not a measured result, and it is not a return-on-investment calculation.

Waiting named per stage instead of hidden in inboxes
Five work stages, four explicit handoffs
A candidate for a small test chosen by evidence from their own numbers

How the mapper works

Start with three stages. The tool opens with three editable rows. Use the add and remove buttons to work with anywhere from two to eight stages.
Enter minutes, not impressions. For each stage give work minutes (active effort) and wait minutes (time the item sits before the next stage begins). Both are whole or decimal minutes between 0 and 100,000.
Name an owner role, optionally. A role such as “invoice clerk” or “approving manager” keeps the handoff explicit. Leave it blank if you do not want to say.
Read the totals. Work total, wait total and elapsed total are the sums of the stages. Wait share is wait total divided by elapsed total, shown as a percentage. If every stage is zero the tool says so rather than dividing.
Download a plan. The export is a Markdown file built locally in your browser. It lists your inputs and states plainly that the numbers are hypothetical and unmeasured.

JavaScript is required for the mapper. Handoff does all of its arithmetic in your browser and cannot calculate totals, draw the bar or build the export without JavaScript. The fieldset below stays disabled until the script runs. No data is sent to a server when scripting is enabled. You can still read the guides: Find the waiting before you automate, Make the handoff explicit, and Choose a small automation test.

The workflow desk

All time values are in minutes. Units are fixed; the tool does not convert hours or days, so convert before you type. Work and wait are kept as separate columns on purpose: work is active effort, wait is time the item sits between handoffs. Elapsed is their sum per stage.

Nothing calculated yet. Enter minutes for each stage, then choose “Calculate totals”.

What this tool does and does not tell you

The output is arithmetic on what you typed. It is not a measurement, a benchmark or an audit, and the numbers are hypothetical until you replace them with observations.
Wait share is a proportion of elapsed time in your own estimate. It is not an economic calculation, a return on investment, a cost model or a justification for a purchase.
“Largest wait” names the stage holding the most waiting minutes in your input. If two or more stages tie, all of them are named.
The tool does not decide what a person must review, approve or be told. Human oversight should be defined deliberately for each system; general framework guidance suggests roles and responsibilities in human–AI configurations be stated explicitly, and that oversight needs differ by context. Handoff only gives you a place to write owners down.
Elapsed time here sums your stages in sequence. It does not model parallel branches, rework loops, calendars, working hours or time-zone effects.

Where the ideas come from

The separation of work time from elapsed time follows standard lean usage: cycle time as the time to complete a process measured by actual timing; processing time as the time work is actually being done, typically a small fraction of lead time; and production lead time as the time to move all the way through a process from start to finish. The same lexicon treats nonvalue-creating time as time that adds cost but no value, including storage, inspection and rework. Applying these terms to office workflows is our extension, not a claim those sources make about software.

On oversight, the NIST AI Risk Management Framework notes that human roles and responsibilities in decision making and oversight of AI systems need to be clearly defined and differentiated, that human–AI configurations span a range from fully autonomous to fully manual, and that some systems may not require human oversight while others specifically may. The same framework lists trustworthiness characteristics—valid and reliable; safe; secure and resilient; accountable and transparent; explainable and interpretable; privacy-enhanced; and fair with harmful bias managed—and stresses that creating trustworthy AI requires balancing them based on context, with tradeoffs. Nothing on this page is a substitute for those documents or for advice about your own obligations.