The Manual-Work Inventory: Find What's Actually Worth Automating

The work nobody sees

Every business runs on a layer of manual work that never shows up in a job description. Someone re-types numbers from one system into another. Someone chases the same three people for approvals every Thursday. Someone routes paperwork by hand because "that's how we've always done it."

None of it is anyone's job. All of it is everyone's job. And it only works when people follow the SOP perfectly — which they do, right up until things get busy. Then steps get skipped, errors slip through, and you find out three weeks later when a customer calls.

This layer of work is bigger than you think, it costs more than you think, and most of it is invisible until you go looking for it.

This guide is a method for going looking. You don't need software to do it. You need two weeks, a spreadsheet, and a little honesty about how work actually gets done in your shop.

Why you inventory before you automate

Most companies do this backwards. Someone sees a demo, gets excited, and picks a tool. Then they hunt for a problem to point it at. That's how you end up with an AI chatbot nobody asked for while your service writers still re-key the same VIN into four systems.

The inventory flips the order. First you find out what manual work actually exists, what it costs, and which pieces are automatable. Then — and only then — you decide what to build. Half the value of the exercise is finding out what not to build. Some of the most expensive-feeling tasks turn out to be cheap. Some tasks that feel trivial turn out to be a six-figure leak.

You can't know which is which until you count.

The two-week log

For two weeks, have each person on the team (or start with one department) log every recurring task they touch. Not one-off projects — recurring work. The stuff that happens daily, weekly, or monthly whether anyone wants it to or not.

For each task, capture five things:

1. Frequency. How many times per day, week, or month does this happen?
2. Minutes per occurrence. Honest average, not best case. Include the waiting-around time — chasing an approval counts the time spent chasing, not just the time spent clicking.
3. Who does it. By role, and note who does it when that person is out. "Only Denise knows how" is a data point.
4. Error rate — and what breaks when it's skipped. Does anything downstream fail? Does a customer notice? Does money leak? A task that gets skipped with zero consequences is telling you something too.
5. Can the rules be written down completely? This is the big one. Could you write the SOP so completely that a new hire could do the task on day one with no judgment calls? Yes, mostly, or no.

Two weeks is the minimum. One week catches the weekly rhythm; two weeks catches the biweekly stuff and smooths out a weird week. If your business has a strong monthly cycle — closing the books, monthly ordering — add those tasks from memory and mark them as estimates.

A few tips that keep the log honest:

• Don't let people log what they think you want to hear. Tell them upfront: this is not a performance review, and nobody's job is getting cut because they logged their tasks. You're hunting for work to take off their plate.
• Watch for the "quick" tasks. People skip logging anything under five minutes. But a 4-minute task done 20 times a day is 80 minutes a day. Those are often the best candidates.
• Log the interruptions. "Answer the same question about inventory status" is a recurring task even though it doesn't feel like one.

What you're looking for

The classic candidates show up in nearly every business:

• Data re-entry. The same information typed into two or more systems. Order details from email into the ERP. Customer info from the web form into the CRM.
• Copy-paste between systems. Pulling a report from one tool, reformatting it, pasting it into another.
• Paperwork routing. Getting the right document to the right person in the right order, by hand.
• Chasing approvals. Reminding people to sign, approve, or respond. The task isn't the approval — it's the chasing.
• Status checking. Logging into a system just to see if something changed, then telling someone about it.
• Report assembly. Pulling numbers from three places into one spreadsheet every Monday morning.
• The steps that get skipped when it's busy. Ask your team directly: "When we're slammed, what do you stop doing?" Whatever they name is a process held together by spare time — which means it fails exactly when the business is doing well.

Scoring: three questions per task

Once the log is full, score every task 1–5 on three dimensions.

A. Rule-clarity

Could you write the SOP so completely that a new hire could execute it on day one, with no judgment calls?

• 5 — Yes. Every input, every decision, every exception has a written rule. "If the invoice matches the PO, approve it. If it's off by less than $50, flag it. Otherwise, escalate."
• 3 — Mostly. There's a clear path 80% of the time, with a known set of exceptions that go to a human.
• 1 — No. The task is judgment. Negotiating with a vendor. Deciding whether an angry customer gets a refund. Writing something that has to sound like you.

This score matters more than the other two combined. Automation runs on rules. If you can't write the rule, you can't automate the task — and no tool changes that, whatever the demo says.

B. Volume

Minutes per occurrence × frequency. A 5 is hours per week across the team; a 1 is a few minutes a month.

C. Cost of errors

What happens when this is done wrong or skipped? A 5 means real money or a customer relationship. A 1 means nobody would ever notice.

Reading the scores

• High rule-clarity + high volume = your automation candidates. This is the short list. Boring, rule-based, high-frequency work is where automation pays off — every time.
• High rule-clarity + low volume + high error cost — worth automating if the build is cheap, because you're buying reliability, not time.
• Low rule-clarity = not a candidate. And that's fine. Judgment-heavy work is what you pay people for. The goal is not to automate your team out of a job — it's to stop paying skilled people to do robot work. Automate the forecasting arithmetic and your buyers don't disappear; they stop doing arithmetic and start doing buying. The machine gets better at the rules, and the humans get better at the judgment — at the same time.

The math

Here's the only formula you need:

Minutes per occurrence × occurrences per year × loaded hourly rate ÷ 60 = annual cost of the task.

Loaded rate means wages plus taxes, benefits, and overhead — for most SMBs, roughly 1.3–1.4× the base wage. If someone makes $25/hour, use $33–35.

Example: a 12-minute data re-entry task, done 6 times a day, 250 working days a year, by someone with a $35 loaded rate:

12 × 6 × 250 = 18,000 minutes = 300 hours × $35 = $10,500 a year. For one task. That nobody thinks about.

Now compare that to the build cost. Most single-task automations are a one-time build somewhere in the low four figures to low five figures, plus a small amount of upkeep. If the annual cost of the task is 2–3× the build cost, it pays for itself in months. If the task costs $800 a year, leave it alone — the spreadsheet is fine.

Add a line for error cost where it's real: (errors per year) × (average cost to fix or lost revenue). Sometimes the error cost dwarfs the labor cost. A skipped step that occasionally ships the wrong unit to the wrong location is not a labor problem.

The warning: don't automate a broken process

The most expensive mistake in automation is pointing it at a process that shouldn't exist.

Before you automate anything, ask: why does this task exist at all? If people re-enter data because two systems don't talk, the fix might be connecting the systems — not building a robot to do the re-typing faster. If everyone chases approvals because the approval rule is fuzzy, fix the rule. You might delete the task instead of automating it, which is even cheaper.

Automating a broken process gets you a broken process at machine speed. You'll produce bad outputs faster, more consistently, and with more confidence. Fix first, then automate.

The copyable inventory table

Paste this into a spreadsheet and hand it to your team. One row per task.

Fill it in for two weeks. Sort by annual cost. Then look at the top ten rows and ask the three questions: can the rules be written down, is the volume real, and what do errors cost?

The rows that score high on all three are your build list. The rest of the list is just as valuable — it's your "don't build this" list, and it will save you from at least one expensive mistake. For the ten projects that waste the most small-business money, see the "don't build this" list.

Where Hoven fits

If you'd rather have a second set of eyes on the inventory — or you've done it and want to know what the top candidates would actually cost to build — most of our clients start with an audit: $999/day, and most take one to five days. You get a prioritized plan plus a "don't build this" list, and if it isn't worth more than you paid, you don't pay. Details on the pricing page, or get in touch.