The question I get asked more than any other is not “how do I automate this.” It is “where do I start.”
And that question matters more than people think. Get the answer wrong and you spend three weeks automating something that saves you 20 minutes a month. Get it right and you free up half a day every week before the end of the month.
I have audited the operations of a lot of small businesses. What I have found, consistently, is that the bottleneck is almost never the automation itself. It is the decision about what to automate first. Most owners either pick the thing that is annoying them right now (which is not always the thing that costs the most) or they try to automate everything at once and stall out on complexity.
There is a better way. It takes about 30 minutes and it starts with a scoring framework.
Why you should not automate everything
First, the obvious thing that does not get said enough: not everything should be automated.
Automation works when a process is consistent. When the steps are the same every time, when the inputs are predictable, when the output can be defined clearly enough that you would know immediately if something went wrong.
When those conditions are not met, automation does not save you time. It produces wrong output that you then have to fix, which costs you more time than the manual version.
The rule is simple: automate the reliable, keep the judgement calls human.
The four criteria for a good automation candidate
Score every process against these four things. Each one gets a rating of low, medium, or high.
1. How much time does it take?
Add up the total time this process consumes in a week, including the time it takes you personally and the time it takes any team member.
- Low: under 30 minutes a week
- Medium: 30 minutes to 2 hours a week
- High: over 2 hours a week
A process that takes 5 minutes a week is not worth automating first, even if it is annoying. A process that takes 3 hours every Monday morning absolutely is.
2. How often does it run?
Daily tasks compound differently from monthly ones. A process that runs every day has 20 times the impact of one that runs once a month, even if each individual instance takes the same amount of time.
- Low: monthly or less
- Medium: weekly
- High: daily or multiple times a day
3. How complex is it?
Complexity here means: how many different decisions does the process involve? How many possible paths are there depending on the input?
A process where step 3 changes based on 15 different client scenarios is complex. A process where every client gets the same sequence is simple.
- Low: highly variable, judgement-heavy, exceptions everywhere
- Medium: mostly consistent with a handful of known exceptions
- High: same steps every time, predictable inputs, clear output
4. What is the risk if something goes wrong?
This is the one most people skip and then regret.
Client-facing errors are expensive. Financial errors are expensive. Reputation errors are expensive. Internal process errors are much cheaper to recover from.
- Low risk: purely internal, no client impact, easily reversible
- Medium risk: client-adjacent, visible but recoverable
- High risk: financial consequences, contract implications, direct client impact on delivery
The scoring matrix
Take any process in your business and score it:
| Criterion | Score |
|—|—|
| Time consumed (Low=1, Medium=2, High=3) | |
| Frequency (Low=1, Medium=2, High=3) | |
| Simplicity (Low=1, Medium=2, High=3) | |
| Low risk (Low=1, Medium=2, High=3) | |
| Total out of 12 | |
Processes scoring 9 or above: automate first. These are your highest-return, lowest-risk opportunities.
Processes scoring 6 to 8: second wave. Worth automating, but do the 9+ group first.
Processes scoring 5 or below: leave alone for now. Either the impact is too low, the complexity is too high, or the risk is too great.
One thing to note: a process that scores high on time and frequency but low on simplicity is not a lost cause. It might need to be broken into stages first, with only the consistent parts automated and the variable parts staying manual. Score each stage separately if needed.
What tends to score highest
Across the businesses I have audited, the same types of processes show up in the top tier repeatedly.
Reporting. Weekly or monthly reports where the data comes from the same sources every time. Same structure, same format, same recipients. This is one of the most consistently high-scoring automation candidates. The manual version takes 2-3 hours. The automated version takes minutes.
Email drafting for recurring scenarios. Not all email. Specifically the emails that follow a predictable pattern: proposal follow-ups, onboarding sequences, check-in messages, invoice reminders. The content varies slightly but the structure does not.
Client onboarding documents. Contracts, welcome packs, briefing documents, access credentials. Once you have a client signed up, the sequence of things you send them is almost always identical. This is highly automatable and usually saves 90 minutes per new client.
Internal summaries and briefings. Pulling together what happened this week, what is due next week, what needs attention. If you are doing this manually, pulling from three or four different tools, you can automate the whole thing into a daily briefing that runs before you open your laptop.
FAQ responses. If you find yourself writing the same email three times a week answering the same question, that is not a communication task. It is a routing problem.
What tends to score lowest
These categories consistently come in under 5 across different business types. Leave them alone, or at least leave them until you have the high scorers running smoothly.
Strategy and positioning decisions. What to charge, who to target, how to position an offer. This is not a process, it is judgement. AI can give you analysis and options, but the decision is yours.
Initial client calls. The first conversation with a prospective client is relationship-building work. It reads people, responds to tone, adjusts in real time. No automation does this well and none should.
Creative briefs. The brief for a piece of content or a campaign is where the thinking happens. You can automate the brief template. You cannot automate the brief itself.
Final financial sign-off. Any process where the last step is approving a payment, a contract, or a financial decision stays human. You can automate the preparation. You sign off.
High-stakes client communications. When a project goes wrong, when a client is frustrated, when you need to reset expectations. These conversations need a human in the room. Even a draft reviewed by AI is high risk if the context is sensitive.
A worked example
Here is how a real scoring session might look for a small consultancy.
Process: writing and sending the weekly project update to clients.
- Time consumed: 3 hours per week across 8 clients. Score = 3.
- Frequency: weekly. Score = 2.
- Simplicity: same structure every time, data comes from the same project tool, format is consistent. Score = 3.
- Risk: client-facing, but low stakes. A delayed or slightly wrong update is easy to correct. Score = 2.
Total: 10 out of 12. Automate this first.
Process: scoping new project proposals.
- Time consumed: 4-5 hours per proposal, roughly twice a month. Score = 2.
- Frequency: twice a month. Score = 1.
- Simplicity: every project is different, scope varies, pricing model has exceptions. Score = 1.
- Risk: gets the price wrong and it comes out of your margin. Score = 1.
Total: 5 out of 12. Leave this alone for now.
The scoping process feels harder and more painful. But automating it would be a mistake. The weekly update, which feels routine and almost invisible, is where the real time is bleeding.
The order of operations
Once you have scored your processes, the sequence is straightforward.
Start with whatever scores highest in your 9-12 band. Run the automation for 2 weeks before moving to the next one. This gives you time to catch anything that does not work the way you expected.
When the first one is running cleanly, move to the next. Do not try to run three automations in parallel during the first month. The compounding happens when each one is solid before you build the next, not when you have five half-finished ones competing for your attention.
I made this mistake in my own business. I started five automations in the same week because I was excited about the potential. Six weeks later, three of them were not working properly, one had sent a client a malformed email, and I had spent more time fixing problems than the automations had saved me.
Start with one. Make it work. Then add the next.
The audit question that changes everything
Most of the scoring exercise above requires you to know two things you probably have not measured: how long each process actually takes, and how often it actually runs.
Most business owners estimate both figures. And the estimates are almost always wrong, usually by a factor of two in the wrong direction. You think the weekly report takes an hour. It takes two and a half.
The most useful thing you can do before scoring is spend one week tracking your actual time by process. Not forever, just one week. Write it down when you switch tasks. Note what you were doing and how long it took. By the end of the week you will have a much more accurate picture than any estimate gives you.
That data is what makes the scoring matrix worth doing properly.
If you want this done for you
The scoring exercise above is something you can run yourself with a spreadsheet and a week of data. It works.
If you want it done faster, and you want the output to feed directly into an automation build, the brain build includes an audit skill that does this across your entire business in one session. It covers seven areas: lead generation, sales, onboarding, delivery, content, communications, and reporting. Every opportunity is scored out of 30 across time saved, revenue impact, and implementation effort. It produces a Phase 1, Phase 2, and Phase 3 roadmap before you leave the session.
The quality gate is set at 80 out of 100. The audit does not finish until it clears that threshold.
That is the brain-audit skill, step 3 of the brain build. More on how it works: Clever Operators Brain Build.
FAQ
What if I only have a few hours a week to spend on this?
Then the time criterion becomes your primary filter. Look only at processes that consume 2+ hours a week. Everything else waits. High-frequency, low-time tasks are worth automating eventually, but they should not be your first priority when time is tight.
What counts as “high risk” for a small business?
Anything that touches money, contracts, or a client relationship that is not going smoothly. If a mistake would cost you real money to fix or would damage trust with a paying client, classify it as high risk and keep a human in the loop even if you automate the preparation steps.
Can I automate client-facing processes at all?
Yes, but with specific boundaries. You can automate the preparation: the draft, the summary, the report. You review and send. The automation handles the labour, the human handles the relationship. That split works well in most client-facing situations.
What if my business does not have repeatable processes yet?
Then documentation comes before automation. You cannot automate a process that does not have consistent steps. Build the SOP first, run it manually 3-5 times to confirm the steps are right, then automate. Trying to automate an undocumented process produces undocumented chaos.
How long does it take to see results from an automation?
A well-chosen first automation, running in the 9-12 band of the matrix, should save you measurable time within the first week. If it is not saving time after two weeks, either the process was not the right choice or the automation was not built correctly. Fix one or the other.
Does this framework work for solo operators as well as businesses with teams?
Yes. The scoring is based on time and risk, not headcount. Solo operators often find the time savings from automation are more significant because there is no one to delegate to. Every hour an automation saves is an hour that comes directly back to you.
