What small businesses are hearing right now is:
“You should be automating everything.”
It shows up in podcasts, LinkedIn posts, software demos, and advice from people whose businesses look nothing like the ones they’re advising. The message is vague but confident. Automation is framed as the obvious next step, the thing that will give you time back, scale your impact, or finally get you out of the weeds.
What’s missing from almost all of this advice is any explanation of what kind of automation is actually accessible to most SMEs, and what assumptions that automation relies on.
What you are not being told:
what automation is realistically accessible without R&D teams
what kinds of automation even exist in practice
which decisions are automatable vs which ones absolutely aren’t
why “high-precision automation” is not where SMEs should start
So people do the only thing that seems logical, aka they reach for the most impressive-sounding solutions first.
That’s where things start going wrong.
Not all automation works in the same way.
First off, it varies by how precise it needs to be. Some automation simply moves information around. Other automation tries to interpret, decide or act autonomously. The difference between those two categories matters far more than most advice out there acknowledges.
At one end, you have automation that applies simple rules. It triggers notifications, moves data between systems, creates tasks, or updates statuses. It doesn’t need to “understand” anything. It just follows instructions. If it fails, the failure is usually visible and easy to unwind. This is otherwise known as ‘low-precision automation.”
These automations:
move information
apply simple rules
don’t make judgement calls
are easy to reverse
fail quietly instead of catastrophically
Examples:
moving data between tools
triggering tasks or reminders
status updates
basic notifications
rule-based workflows with clear inputs
At the other end, you have automation that replaces judgement. It decides what matters, what comes next, or what action should be taken. This kind of automation assumes clean data, shared definitions, stable processes, and a high tolerance for mistakes while things get tuned. This is “high-precision automation.”
These automations:
replace judgement
make decisions
interpret ambiguous inputs
assume clean data and clear intent
fail in expensive, confusing ways
Examples:
AI agents acting autonomously
automated decision-making based on fuzzy signals
tools that “decide” what matters
anything that assumes your business logic is already well-defined
Most SMEs are being encouraged to start at the second end.
Why that’s backwards y’all
Small businesses are rarely low-complexity environments. Things are messy and full of one-off situations. Priorities are constantly changing and information is never complete. Not to mention context often lives in your people’s heads rather than in mapped out workflows or databases. That is the reality of small operations. You don’t have big budgets or R&D teams to design you the perfect solutions.
High-precision automation does not cope well with that reality. It needs inputs to be consistent and meanings to be agreed and stable. The business needs to already know exactly what good looks like, what counts and an exception, and what decisions are even safe to automate at all.
When those conditions aren’t in place, the automation just doesn’t work. You’re stuck overriding outputs and promising yourself you’ll fix it later. Assuming you can even get it that far along.
The result sure as heck isn’t efficiency.
Why low-precision automation is your only sensible starting point
Low-precision automation doesn’t try to think. This is crucial to your first attempts at any business automation. It it is least likely to break business critical things when it doesn’t work right. you might think this means it won’t be useful, but having a program that runs and moves information so that you don’t have to manually can be a huge time-saver. It might even help keep where all your info is and isn’t more tidy.
You don’t need perfect data. You won’t even need AI in order to reduce the number of times a human copies and pastes information. This kind of automation doesn’t make make decisions for you but it does make it easier for you to make them yourself.
Where AI and the ‘Agent’ hype falls apart
A lot of the current automation narrative is wrapped up in AI, agents, and autonomous systems. In theory, these tools promise to take action on your behalf. In practice, they assume conditions that most SMEs simply don’t have.
That technology isn’t useless, but it is being constantly positioned as a starting point when it is actually an end state. Hell, even the big kahunas are still trying to crack it (it is largely still an absolute dumpster-fire of issues.)
The advice to dive into the deep end with agentic AI assumes that the cost of getting it wrong will be acceptable while the system learns. We don’t know any non-series funded companies who can afford that cost. When something goes wrong for you, that ‘experiment’ costs you cashflow, reputation and trust.
We know it is very tempting to skip the unglamourous end of the scale here and jump straight to the exciting sounding tools that promise to think for you. But you are trying to outsource responsibility before you have your house in order. You have to start in the right place. Automating workflows first, (how info arrives, where it goes, who sees it) makes everything else much easier. Only then does taking a look at higher-precision automation make sense.
A clearer way to think about it
Automation is not a shortcut. It only amplifies whatever already exists.
When the foundations are in place, even simple automation can have an outsized impact.
