Automating a Small Business: Where to Start

Automation usually starts not with a calculation but with irritation: "I am tired of holding everything in my head." Then a system gets bought, 200,000 gets spent, and six months later it turns out exactly the same things are still being done by hand, only now inside an interface as well. The problem is that what got automated was not what hurt, but what was easiest to buy.

Find where the time leaks out

Take a week and ask two or three employees to record what their day goes on, to the nearest half hour. Not for surveillance — for inventory. Almost always it turns out that 8–12 hours a week per person go on mechanics: retyping data from one place to another, assembling the same document, answering the same question, working out on a calculator what a formula could compute. On a salary of 80,000 roubles that is roughly 20,000 roubles a month per person thrown at copy-paste.

Criteria for the first process

• It repeats at least once a day, and preferably ten times.
• It can be described entirely by rules, with no "you have to think about this bit".
• A mistake in it costs money: a lost enquiry, a wrong total, an overdue document.
• It is not done by one irreplaceable person but by several, all of whom do it differently.
• It can be automated in two to four weeks, not six months.

What usually pays off first

In our experience three things deliver results in almost any business. First, automatic intake of enquiries: the website form, phone calls and messenger messages all land in one system without human involvement; we covered how this works in our piece on linking a website to a CRM. Second, templated documents: a contract, an invoice or an acceptance certificate assembled from a record in a second instead of twenty minutes, with no typos in the legal details. Third, customer reminders: a booking confirmation plus a reminder 24 hours ahead cut no-shows by 20–30 percent, and in a service business a no-show is a direct loss on an empty hour.

How to measure the effect

Before the rollout, record three numbers: how many hours a week the process takes, how many errors occur per month, and what one error costs. Two months afterwards, measure the same three. If the saving is less than the cost of owning the system, the automation did not pay off, and that is a normal experimental result, not a catastrophe. It is worse when nobody counted and everyone is happy with the feeling. We ask clients to record these figures before work starts — we discuss them at the first meeting, because once the system is live the original numbers can no longer be recovered.

What not to do

Do not automate a broken process: if enquiries get lost because nobody owns them, the system will simply start losing them faster. First describe the order of work in words and make sure it works manually. Do not buy a platform for future growth — modules opened once a year mean paying for air. Do not do everything at once: two new tools per quarter is the limit for a team under 20 people. And do not start with the hardest bottleneck, even if it is the most painful: the first automation has to end in a win, or you will not get approval for the second.

Where to go next

After a first successful step the logic is simple: pick the next process against the same list of criteria. Within a year most companies end up with a website, an accounting system and a customer portal — and that is when it makes sense to think about a single platform rather than a set of tools. For some industries such platforms already exist, for example Auto555 for rental and homade for salons; for everything else we build solutions around a specific process. If you are unsure about the order of steps, look at how similar tasks were solved in our projects — you can see there where each one started and how it ended.