When a small business decides to automate something, the first instinct is almost always to fix what hurts the most. The thing that takes the longest. The thing the team complains about. The thing the owner stays late on.
That instinct is wrong, and it costs people about three months on average.
The thing that hurts the most is usually complex, multi-stakeholder, and politically charged. It hurts because it’s a hard problem, and hard problems make terrible first automations. The right first task is the one that scores well on six dimensions of feasibility, even if it’s not the loudest. Build that one first, prove the model works, and the political capital you earn buys you the right to tackle the painful one second.
This post is the rubric I use when a client sends me their automation wishlist. Same rubric, written so you can run it on your own tasks in under five minutes.
The 16-hour figure is from The Industry Leaders’ analysis of executive admin time. The 20% productivity gain comes from Salesforce’s small business framework research. The 2.5x cost overrun is from i3Solutions’ research on framework-based vs ad-hoc automation projects.
What goes wrong when you start with the worst-feeling task
The painful task is painful for a reason. Usually three reasons.
It’s high-stakes. If you mess up the payroll automation, people don’t get paid. If you mess up the invoicing automation, you don’t get paid. The political downside of a misfire is large enough that you’ll second-guess every step, slow the build, and build in so many manual checks that the automation ends up barely faster than doing it by hand.
It involves multiple people. Painful tasks usually touch sales, ops, and finance. Each owner has opinions about how it should work. The build turns into a six-week negotiation about edge cases instead of a two-week implementation.
It’s underspecified. Painful tasks are usually painful partly because nobody’s bothered to write them down. The first three weeks of the build are spent reverse-engineering “what we actually do” from interviews. By week four the team has lost interest and the project quietly stalls.
A small first win avoids all three traps. Pick a task with low stakes, one owner, and a process you could explain on a napkin. Ship it. Then the painful one becomes the obvious second build, and you have a track record that buys you the patience the longer build will need.
What I score, and why each one matters
| Dimension | Why it matters | Score 10 looks like |
|---|---|---|
| Frequency | Time saved compounds with how often the task runs. Daily beats monthly even when each run is shorter. | Multiple times a day |
| Time per run | The savings per run are bounded by how long it currently takes. 30 to 60 minutes is the sweet spot. | 30 minutes to 4 hours |
| Stability of steps | If the steps shift quarterly, you're rebuilding the workflow quarterly. Build on the parts that haven't changed in months. | Same for 6+ months |
| Number of apps | Each integration adds setup time, maintenance burden, and failure modes. Two to three is the sweet spot. | 1 or 2 apps |
| Judgment required | Rule-based automation can't make judgment calls. AI can handle some, but it complicates the build. Zero-judgment tasks ship faster. | No judgment needed |
| Documentation readiness | If you can't write the steps in five bullet points, you'd be automating today's improvisation. The first build needs a documented process. | Yes, easily |
Score each dimension 0 to 10 for the candidate task, sum the scores. Anything above 50 is a strong first-week candidate. Below 35, look for a different task. Anywhere in between, you have one or two specific dimensions to fix before building.
Three candidate tasks, scored side-by-side
A 15-person ecommerce business comes to me with three tasks they want to automate:
- Daily reconciliation of orders across Shopify, TikTok Shop, and Pancake POS into a master spreadsheet.
- Customer service responses to common product questions.
- Quarterly inventory reorder decisions based on sales velocity and supplier lead times.
Their gut ranking would be #2 (customer service is the loudest pain). Let’s score each properly.
| Dimension | #1 Reconciliation | #2 Customer service | #3 Inventory reorder |
|---|---|---|---|
| Frequency | 10 (daily) | 10 (multiple/day) | 1 (quarterly) |
| Time per run | 10 (45 min) | 3 (3 min each) | 10 (full day) |
| Stability | 10 (same for 2 yrs) | 5 (templates evolve) | 3 (logic shifts) |
| App count | 10 (3 apps) | 10 (1 app) | 3 (5 apps) |
| Judgment | 10 (none) | 5 (AI drafts, human approves) | 0 (real judgment every time) |
| Documentation | 10 (clear) | 5 (informal) | 0 (lives in owner's head) |
| Total | 60 | 38 | 17 |
Reconciliation wins decisively. The painful customer service task scores in the middle, with judgment as the weak spot. Inventory reorder is disqualified by both judgment and documentation gaps.
The recommendation is clear: build reconciliation first. It ships in two weeks, saves 15 hours a month, and builds the team’s confidence. Then build a customer-service AI assistant where the AI drafts and the human approves, which is a different shape of project and benefits from being second. Inventory reorder isn’t ready until the underlying decision logic gets written down.
The rule of thumb: if your top-scoring task is painful, automate the painful one. If your loudest pain is in the bottom half, automate something quieter first and come back to the loud one with momentum.
A five-minute version
We built a free tool that walks through these six questions for any task you have in mind: Automation Readiness Audit. Six questions, no email, gives you a score out of 60 and a verdict in under five minutes.
The audit is the same rubric this post describes, packaged as an interactive quiz. It also estimates how many hours you’d save annually if you automated the task, so you can put a real number against the build effort.
What to do with each verdict
Once you have a score, the next move depends on which band you landed in.
50 or higher (Strong fit). Build it. Either DIY with Zapier or n8n, or hire someone to build it once and document it. Two-to-four week build, runs for years. Don’t over-engineer the first version, get it shipping data into the right places and refine in week three based on what actually breaks.
35 to 49 (Promising, after one fix). Identify the lowest-scoring dimension. If it’s stability, document the current process and run it the same way for a quarter before automating. If it’s app count, simplify the chain. If it’s judgment, look at AI agent patterns instead of rule-based automation. If it’s documentation, write the five bullet points. Then re-run the audit and most tasks move into the 50+ band.
Below 35 (Not the right first one). Pick a different task. There’s almost always a higher-scoring candidate hiding somewhere in your operations. Daily admin tasks, inbox cleanup, recurring report generation, simple inter-app data sync. Run the audit on three or four candidates and pick the highest-scoring one.
The full method is in Workflow Automation for Small Business: Where to Start. The cost-and-effort comparison for build, buy, or hire is in How Much Does Business Automation Cost?. For specific examples of well-scoring first automations, see Outgrown Copy-Paste Workflows.
How I help with the first one
If you’ve run the audit, scored a task above 50, and want a second opinion before you build, send me an email. The first call is free. I’d rather tell you “this is the right one” or “actually, here’s a better candidate I see in your description” than have you spend three weeks building the wrong thing.
The reason I push back on the “automate the painful one” instinct so hard is that I’ve watched eight or nine clients try it. None of them shipped on schedule. The ones who shipped at all did it by quietly switching to a smaller first project halfway through. Save yourself the detour.