The 30-Minute Automation Audit: Find $10K of Busywork in an Afternoon

← Back to Blog

Every business has ten thousand dollars of busywork hiding in plain sight. It is not in one dramatic place — it is scattered across a hundred small routines that nobody thinks about because they have always been done that way. The report that takes three hours to assemble every Friday. The application data that gets retyped from one system into another. The email thread that needs a human to summarize and forward. None of it feels automatable on its own, and all of it is.

The problem is that most automation attempts never find this stuff, because they start with the tool. Someone reads about a shiny platform, buys a subscription, and then hunts for something to use it on. The audit in this article runs in the opposite direction: it starts with the work, finds the processes that are painful, repetitive, and rule-bound, and only then asks what should be automated. You can run the whole thing in an afternoon, with no consultants and no purchases. When you are done, you will have a scored list of candidates and a clear order of attack — the same output we produce before we build anything for a client.

Why Audits Fail When They Are Big Projects

The most common approach to finding automation opportunities is a full process audit: a spreadsheet of every workflow, interviews with every team, a consultant's report with forty recommendations. It sounds thorough, and it almost never gets acted on. By the time the report lands, the business has moved on, nobody owns the forty items, and the binder goes on a shelf. The audit fails not because the analysis was wrong but because the scope was too big to finish or to act on.

A good audit is short, cheap, and produces a short list. It does not need to be comprehensive — it needs to be actionable. If you find three candidates worth automating and fix one of them this quarter, you have already beaten ninety percent of the companies that hired consultants to find forty. The other reason small audits win is psychological: a thirty-minute exercise feels like something you can actually do, so you do it. This article is built on that principle — four passes, each one quick, each one producing a concrete artifact.

One more thing before you start: silence the inner critic that says 'we already know our processes.' You know the broad strokes. The audit exists to catch the details you have stopped seeing, the way you stop hearing the hum of your own refrigerator. The person doing the work every day will tell you exactly where the pain is — you just have to ask the right questions and write down the answers.

Pass 1 — Find the Sigh

Every process that needs automating has a tell: somebody sighs when they think about it. The sigh is not a joke — it is the most reliable signal in the building, because it comes from the person who does the work and knows exactly how much it sucks. Your job in pass one is to collect the sighs.

Walk the floor — or the Slack channels, or the Zoom rooms — and ask one question: 'What part of your week makes you sigh every single time?' Do not ask what should be automated, do not ask what is inefficient, do not ask what the company should invest in. Ask what makes them sigh. You will get honest answers, and they will cluster. The same three or four routines will come up from different people, and where the sighs cluster is where the money is.

Write down every answer on a whiteboard or a shared doc. No filtering yet, no 'we can't fix that.' Just capture. You are looking for five to ten recurring answers, and you want the language to stay in the words people actually used — 'the Friday report,' 'retyping applications,' 'chasing invoices.' Those names are the candidate list, and they are worth more than any generic category from a consultant's template.

• Ask every team member the sigh question, one-on-one or in a quick standup
• Capture answers in their own words — no editing, no judgment
• Cluster the answers and look for the same routine mentioned by two or more people
• Pick the five loudest sighs as your candidate list
• Assign each candidate a rough weekly time cost from whoever does it

Pass 2 — Follow the Paper

Now take one real artifact that flows through your business — an application, a purchase order, a lead, a timesheet — and physically follow it from arrival to archive. Where does it enter? Whose hands does it pass through? Where does it wait? Where does the same information get typed in a second time? This pass is where the invisible waste lives, because documents travel through systems that were bolted together over years, and every handoff is a chance to lose time or introduce an error.

The fastest way to do this is to sit with the person who processes the artifact and watch. Not ask — watch. People forget the steps they have internalized; they skip the awkward bits when they describe their job. In thirty minutes you will see the whole chain: the email inbox where the attachment lands, the manual data entry into the CRM, the spreadsheet that gets updated as a backup, the approval that waits on a reply, the filing that happens 'when there is time.'

Mark every step that is rule-bound — a step where the decision is always the same, the format is always the same, or the destination is always the same. Those are the steps that do not need a human. Mark every step that requires judgment — the exception, the relationship, the call — because those are the steps you must keep. The pattern to look for is a chain of rule-bound steps separated by human hands. That chain is one automation candidate, and you have just drawn its blueprint.

Every time the same information gets typed twice, a human is doing the work of a bridge.

Pass 3 — Count the Clicks

Pass three is a gut check on volume. An automatable process needs to happen often enough that the time savings add up. The question is simple: how many times a week does this routine happen, and how many minutes does each occurrence take? Do not estimate from memory — count it. Look at the logs, the ticket counts, the spreadsheet timestamps. The difference between 'a few times' and 'forty times a week' is the difference between a nice idea and a real project.

The math is brutally simple and it is the whole point of the pass. Forty occurrences times fifteen minutes each is ten hours a week, which is roughly a quarter of a full-time employee. Fifty-two weeks of that is more than five hundred hours a year — and at a modest fully-loaded hourly cost, that is a five-figure number sitting in a routine nobody ever noticed. That is where the title of this article comes from. Not from one dramatic process, but from the quiet arithmetic of repetitive work.

Write the arithmetic next to each candidate from pass one. You now have the two numbers that matter: the sigh score (how much it hurts) and the volume score (how often it happens). Both matter. A painful process that happens twice a year is not worth automating; a mildly annoying one that happens two hundred times a week very much is.

• Occurrences per week, from logs or counts — never from memory
• Minutes per occurrence, from the person who does it
• Weekly hours = occurrences × minutes ÷ 60
• Annual hours = weekly hours × 48 (leave room for reality)
• Annual cost = annual hours × fully-loaded hourly rate

Pass 4 — Price the Minutes

Now put a number on the list. Use a fully-loaded hourly rate — salary, benefits, overhead, the cost of the desk and the software — not the raw wage. For most roles that lands somewhere between $35 and $75 an hour. Multiply the annual hours from pass three by that rate, and you have the annual cost of each routine. This is the number that turns a conversation about 'maybe we should automate' into a decision you can take to anyone with a budget.

Be honest about the friction too. The best candidate is not the most expensive process — it is the process that is expensive AND simple to automate. A routine that costs $40,000 a year but involves judgment and exceptions is a harder project than a $15,000 routine that is pure rules. The scoring matrix in the next section exists precisely to keep you from falling in love with the big number and ignoring the difficulty.

One caution: do not automate a broken process. If the Friday report is painful because the data is messy, automating the mess just produces the mess faster. The audit assumes the process is worth keeping and simply needs its hands removed. If pass two revealed a process that is fundamentally confused, the first project is to fix the process — and that is a people conversation, not an automation one.

The Fully-Loaded Rate

The single biggest mistake in pricing busywork is using the hourly wage instead of the fully-loaded cost. The wage is what the employee takes home; the loaded cost is what the business actually pays to have them in the seat — payroll taxes, benefits, software, equipment, supervision, and the overhead that follows a headcount. In most organizations that lands between 1.4 and 2 times the wage. If you are not sure, ask whoever does the accounting; the loaded rate is a number they know.

Use the loaded rate in the pass-four arithmetic and the candidates will reorder themselves. A routine done by a $22-an-hour assistant at a loaded rate of $40 suddenly looks like a much better target than the same routine done by a manager whose loaded rate is $85 — and that reordering is exactly what you want, because it points your first automation at the work that is both repetitive and cheap enough to hand to a machine.

The Friction Test

Before any candidate makes the top three, run the friction test: how hard would it actually be to automate? A process that already lives inside one system and needs one rule is a weekend project. A process that spans three systems, needs approvals, and touches customer data is a real build. Score friction honestly on the same one-to-five scale as the other axes, and subtract it from the enthusiasm. The winning first project is not the one that saves the most money — it is the one that saves real money at low friction, because it will actually ship.

The Scoring Matrix: Volume × Rules × Cost

By now you have a candidate list with three attributes: pain (from the sighs), volume (from the clicks), and annual cost (from the minutes). The scoring matrix turns those attributes into an order of attack. Give each candidate a score from one to five on three axes: volume, rule-boundedness, and human cost. Multiply them. The candidates with the highest product are your top three, and the top three are all you need.

Volume is how often it happens. Rules is how completely the process follows a logic you could write down — a process with a clear if-then structure scores five, one with judgment calls scores one or two. Human cost is the annual number from pass four, scaled so the worst offender is a five. A candidate that scores high on all three — happens constantly, follows rules, and eats real money — is the textbook automation target. A candidate that scores high on cost but low on rules is a process-improvement project in disguise, and you should treat it that way.

Write the top three on a fresh page and put the rest away. The temptation now is to start automating everything at once, and that is exactly how automation projects die — too many balls, no owner, no momentum. Three candidates, one owner each, and a clear order is the entire plan. If you genuinely only found one strong candidate in the whole audit, that is a perfectly good result. One process, done well, with real numbers behind it, is worth more than a roadmap of ten.

• Score each candidate 1-5 on volume, rules, and human cost
• Multiply the three scores — the top product is your first project
• Assign one owner per candidate — a person, not a committee
• Set a 30-day window for the first project, with a written verdict at the end
• Ignore everything below the top three until one of them ships

The Fix Queue: What to Do with the List

The audit is only worth something if it becomes a queue, so turn the top three into a simple fix queue with three columns: process, owner, and verdict date. The owner is the person accountable for making it work — the same ownership rule we stress in our breakdown of automation that pays for itself. The verdict date is when you decide, in writing, whether the automation is expanding, being adjusted, or being killed.

For the first project, pick the one that scores highest on rules, not the one that costs the most. A smaller, cleaner win builds the muscle and the credibility; the expensive, gnarly process can wait until the team has seen automation work once. And run the project the boring way: baseline the numbers first, automate with a human in the loop, compare, and write the verdict. The pattern is the same one we use everywhere — it is boring on purpose, because boring processes get followed.

The final step is a calendar reminder for ninety days out: run the audit again. Processes change, volume changes, and new sighs appear. The audit is not a one-time event; it is a habit that keeps the candidate list fresh. If you run it quarterly, you will always know where your next automation win is, and you will never again be the business that buys tools and hunts for problems. The tools are easy — the list is the hard part, and now you have it.

What Makes a Good First Project

The ideal first automation is small, rule-bound, high-volume, and low-risk. It happens daily, follows a logic you could write on a napkin, and its failure mode is an annoyance rather than a disaster. Think a report that assembles itself, an ingestion step that stops being manual, a notification that routes itself. If the project touches customers or candidates, keep a human in the loop until the numbers prove the machine — that guardrail, plus the others we keep returning to, lives in our guide to building a practical AI stack.

When to Call In a Builder

Some of what the audit surfaces will be simple enough to build with off-the-shelf tools and a patient afternoon. Some of it will need real integration work — connecting systems that were never meant to talk, handling the exceptions, building the checks. The line is not about size; it is about risk and plumbing. If the process touches money, compliance, or customer data, or if it needs three systems to behave like one, that is a builder's job. The audit has done its part by defining the target clearly; our piece on taming tool sprawl covers how to think about the plumbing without buying a whole new stack.

Why Quarterly Beats Annual

A once-a-year audit catches the processes that grew over twelve months, but the best candidates are the ones that appear in a quarter — a new tool the team adopted, a product line that doubled volume, a regulation that added a step. Quarterly runs keep the list current, which matters more than the list being complete. Thirty minutes every ninety days is a rounding error on the calendar and a permanent early-warning system for busywork. Set the reminder today; future you will thank you in March.

The Thirty-Minute Recap

Here is the whole audit on one screen, for the afternoon you actually run it. Pass one collects the sighs and gives you the candidate names. Pass two follows one document and shows you where the hands touch the work. Pass three counts real occurrences so the list has volume. Pass four prices the minutes at a fully-loaded rate so the list has money. The scoring matrix multiplies volume, rules, and cost to produce the top three, and the fix queue assigns an owner and a verdict date to each.

The entire exercise takes a focused afternoon the first time, and about thirty minutes once you have done it before. The output is not a strategy document — it is three projects with names, numbers, owners, and dates. Run one, measure it, write the verdict, and then run the audit again with the confidence that comes from having actually shipped. That is the whole system, and it will keep finding your next ten thousand dollars as long as you keep running it.

• Pass 1 — Find the sigh: five candidate names from the team's own words
• Pass 2 — Follow the paper: one document traced end to end, rule-bound handoffs marked
• Pass 3 — Count the clicks: real occurrences per week, never memory
• Pass 4 — Price the minutes: annual cost at the fully-loaded rate
• Score volume × rules × cost, pick the top three, assign owners and verdict dates
• Put the ninety-day reminder in the calendar before you close the doc