The 15 minute audit that finds your worst manual job

<h2>The job nobody put on the roadmap</h2><p>Every company has one process that runs on a person. It is not written down, it does not appear in anyone's job description, and it eats somewhere between twenty minutes and two hours a day. Ask an operations lead what it is and they answer in about four seconds, because they are the person.</p><p>The audit below takes fifteen minutes and produces a ranked list of those jobs. We built it for our own team, then started running it on the first call with new accounts, because guessing which flow to build first is a reliable way to build the wrong one.</p><h2>Step one: list the moves, not the tools</h2><p>Open a blank page. Write one line for every time information moves from one app to another because a person carried it. A move is somebody reading a value in one place and typing it into another. Ignore anything already automated. You want moves, since a list of tools tells you nothing about where the hours go.</p><p>A typical answer from a twelve person company looks like this:</p><ul><li>Support email into the task board, roughly forty a week</li><li>Website form submissions into the CRM, done in a Friday batch</li><li>New payments into the revenue sheet, twice a week</li><li>Task board status into the client update on Thursday</li><li>Refunds mentioned in chat into the finance tracker, when somebody remembers</li></ul><p>Five lines is normal. Ten lines usually means the company grew faster than its systems, which is also normal and not a moral failing.</p><h2>Step two: put four numbers on each line</h2><p>Each number is a guess you can make in under a minute. Do not go looking for data. The point is ranking, and a rough rank beats a precise study you never finish.</p><table><thead><tr><th>Move</th><th>Times per week</th><th>Minutes each</th><th>Cost of a miss</th><th>Score</th></tr></thead><tbody><tr><td>Support email to task board</td><td>40</td><td>3</td><td>High</td><td>360</td></tr><tr><td>Form fills to CRM</td><td>25</td><td>2</td><td>High</td><td>150</td></tr><tr><td>Payments to revenue sheet</td><td>2</td><td>25</td><td>Medium</td><td>100</td></tr><tr><td>Board status to client update</td><td>1</td><td>45</td><td>Medium</td><td>90</td></tr><tr><td>Refunds to finance tracker</td><td>3</td><td>4</td><td>High</td><td>36</td></tr></tbody></table><h3>How the score works</h3><p>Multiply times per week by minutes each, then apply a weight for the cost of a miss: one for low, two for medium, three for high. That is the whole formula. The weight is what separates a missed refund line, which produces an argument with a customer, from a missed planning row, which produces five idle minutes on Monday.</p><h2>Step three: check that the move is really a move</h2><p>Some lines turn out to be judgement work wearing a copy and paste costume. If the person doing it reads the message, decides which of four categories it belongs in, then types a summary in their own words, that is not a move. That is triage, and automating it produces confident nonsense.</p><p>The test we use is simple. Could a competent temp do this correctly on their first morning, given a one page rule sheet? If yes, it is a move. If they would need to ask a question, split the line in two: the mechanical half becomes a flow, and the judgement half stays with a person who now has more time for it.</p><blockquote><p>Our first flow replaced the job everyone complained about. The second replaced a job nobody had ever mentioned, and it saved twice as much time. We had been ranking by annoyance, not by hours.</p><p>Rosa Adeyemi, operations lead at Kestrel Freight</p></blockquote><h2>Step four: build the top line and stop</h2><p>Take the highest score and build only that. Resist the urge to build three at once, because three half-configured flows produce three sets of unfamiliar errors in the same week and nobody can tell which one is misbehaving.</p><ol><li>Write down the trigger in one sentence. If you cannot, the process is fuzzier than you thought and needs another pass.</li><li>List the destination fields and where each value comes from. Mark anything with no source, since those become defaults or stay empty.</li><li>Run it against ten real records with the destination pointed at a test board or a test base.</li><li>Compare all ten by hand. This step feels wasteful and catches nearly every mapping mistake.</li><li>Turn it on, then check the run history at the end of the first day and again at the end of the first week.</li></ol><h2>What the audit usually turns up</h2><p>Three patterns come up again and again. The first is a batch job someone does on Friday that could happen continuously, where the real cost is not the typing but the four days of stale data before it. The second is a low-frequency, high-consequence line near the bottom of the score column that deserves a flow anyway, usually anything involving money going back out. The third is a move that exists only because two teams disagreed about which app owns a piece of information, and no flow will fix that. Have the conversation instead.</p><p>When you rerun the audit a quarter later, most teams find the list has not shrunk. New moves appear as fast as old ones are automated, because that is what growth looks like from the inside. The difference is that the top of the list is now worth ninety minutes a week instead of six hours.</p>