AI Operations (FDE) insight

How Can I Automate My Business? The Method, One Handoff at a Time

Hayat Amin · Updated 2026-09-12

How can I automate my business: the five step method for a small or mid sized company, with the hours priced from Bureau of Labor Statistics figures and the platform prices read from the vendors' own pages on 12 September 2026.

You automate a business one handoff at a time. Find the job that repeats every week, runs on a rule anyone could follow, and moves information between two systems by hand. Wire that single job, count the hours it gives back, then take the next one.

That sentence is the whole method. The rest of this piece is how to run it: how to find the job, what an hour of it actually costs you, which tool to buy, what the published prices were when this was written, and how to tell when a process has outgrown a subscription and needs an engineer.

Where automation actually lands in American companies

Before the method, the map. The United States Census Bureau ran an artificial intelligence supplement to its Business Trends and Outlook Survey, and six researchers, Kathryn Bonney, Cory Breaux, Emin Dinlersoz, Lucia Foster, John Haltiwanger and Aditya Pande, published the analysis as working paper CES-26-25 in April 2026. Over the reference period of November 2025 to January 2026, 18 percent of firms used artificial intelligence in a business function, rising to 32 percent when weighted by employment, and adoption was expected to reach 22 percent within six months.

The useful part is where it lands. Among firms that had adopted, the most common functions were Sales and Marketing at 52 percent, Strategy and Business Development at 45 percent, and information technology at 41 percent. Fifty seven percent of those firms used it in three or fewer functions at all. At the level of the individual worker, the leading uses were writing, document analysis and information search, and 65 percent of firms kept it to three or fewer tasks.

Read that as permission. The companies getting value are not running twenty automations. They are running two or three, in the places where paperwork and repetition already live.

Step one. Write down handoffs, not tasks

Most owners begin by listing what they do. That list is too long and it never converges. List handoffs instead: the moments when work changes hands or changes systems. The order that gets retyped from an email into the accounting package. The enquiry that sits in a shared inbox until somebody notices it. The spreadsheet somebody rebuilds every Friday afternoon from the same three reports. The invoice that gets chased by memory.

Automation lives at the seams, not in the middle of anybody's craft. So keep a log for one working week, and write a line every time you or a member of staff types something that already exists somewhere else. Record the two systems involved and the minutes it took. Five days of that produces a better automation plan than a month of software demonstrations.

Step two. Price the hour before you price the tool

A handoff is worth automating when the labour it eats costs more than the software that replaces it, and you cannot know that until you price the hour.

The Bureau of Labor Statistics puts the median pay for secretaries and administrative assistants at 48,310 dollars a year, or 23.23 dollars an hour, as of May 2025. About 3.5 million people hold those jobs in the United States, and the Bureau projects overall employment in the occupation to decline 2 percent from 2025 to 2035 while still producing around 314,400 openings a year. That hourly figure is the fair benchmark for the administrative half of a small company's week.

Now do the arithmetic on your log. A handoff that takes forty minutes a week consumes about thirty five hours a year. At the median administrative rate that is roughly 805 dollars of labour, every year, for one line on your list. Three such lines and you are spending the price of a used van on retyping. Your own hour costs more than the benchmark, which is why the jobs an owner personally performs should be priced highest and cut first.

Step three. Pick the first job with a rule, not a feeling

Three tests, and a candidate has to pass all of them.

It repeats at least weekly. Anything quarterly can wait, because you will have forgotten how the automation works before it runs twice.

Its decision can be written down as a rule that a new hire could follow without asking you a question. If you cannot write the rule, the job is judgement, and judgement is the last thing to hand over, not the first.

A failure is visible and cheap. Start where a mistake shows up the same day and costs an apology, not a customer. Payroll and tax filing are not first automations.

One candidate usually survives, and it is almost always dull. Dull is correct. The first automation exists to prove the method inside your own business, not to impress anyone.

Step four. Buy the smallest thing that does the job

Three platforms cover most small company work, and all three publish prices you can read without a sales call. These were read from the vendors' own pricing pages on 12 September 2026, and prices move, so check them before you buy.

Zapier is the one to try first, because it connects the widest set of ordinary business software and its free plan runs 100 tasks a month, which is enough to prove a single handoff. Paid plans start at 19.99 dollars a month for Professional when billed yearly, and 29.99 dollars billed monthly, with Team from 69 dollars a month billed yearly.

Make suits work with more branching, where one trigger has to fan out into several conditional steps. It publishes a free plan of up to 1,000 credits a month and a Core plan from 12 dollars a month, with Pro at 21 dollars and Teams at 38 dollars. Its paid tiers are priced on a credit slider, so the figure you see moves with the volume you select.

n8n is the choice when you want to keep the work in your own hands. A standard self hosted version is free on GitHub, and the hosted plans are priced in euros: Starter at 20 euros a month billed annually for 2,500 workflow executions, Pro at 50 euros for 10,000, Business at 667 euros for 40,000. Self hosting costs you a server and someone competent to run it, which is a real cost, but it keeps your data on your own infrastructure.

The rule underneath the prices: start on the free tier, and upgrade when the run count forces you to, never because a feature list is flattering. If you are still under 100 tasks a month after six weeks, you did not automate a real handoff. You automated a hobby.

Step five. Give it an owner and a way to fail loudly

The expensive failure in small company automation is not the automation that breaks. It is the one that breaks quietly. A connection expires, nobody is told, and for five weeks the orders that used to copy themselves across simply do not, until a customer asks where their delivery is.

So every automation you build gets three things before you call it finished. A named person who owns it, by name, not by department. An alert that reaches a human when a run fails, in a channel that person actually reads. And a calendar entry, monthly, to confirm the process it automates is still the process you run. Businesses change faster than their scripts, and an automation that quietly encodes last year's workflow is worse than no automation.

What not to automate in the first month

Do not automate a process you know is broken. You will get the wrong outcome faster, and at scale.

Do not start with anything customer facing that requires judgement. The Census research is blunt about how firms are actually using this: 66 percent of users rely on it solely to augment tasks rather than replace them, and reductions in employment linked to artificial intelligence were rare, occurring in only 2 percent of firms. The pattern that works is a person, made faster, not a person removed.

Do not buy six tools at once. The same research found most adopters using three or fewer functions, and the failure pattern in small companies is not picking the wrong tool, it is paying for several and configuring none.

If what you actually want is a list of named software by job, with prices, we published that separately as nine jobs artificial intelligence does for a small business. If the work you are trying to automate lives in the finance function, the specific version is where agents actually work in a finance department.

When it stops being a subscription problem

A connector platform can only join things that already have doors. The moment the job spans a system with no usable integration, or depends on a rule that exists only in your business, or touches data nobody else holds, you have left the subscription market. That is not a bigger plan. That is engineering, done inside your company, against your systems, by somebody answerable when it breaks.

That is the work Hayat Amin and Beyond Elevation do: a forward deployed engineer who sits inside a small or mid sized company in New York City or elsewhere in the United States, builds the automation in your own stack, and leaves it running and documented rather than leaving a deck behind. You can book a call about that engineering work, or read what it involves on the forward deployed engineering page. When the harder question underneath is what any of it does to the numbers, there is also the CFO seat.

If you want to know which of your own processes are worth automating first, we run a free audit: one call, then a written list of what to automate first, what it saves and what it costs, at beyondelevation.com/call/audit.

Frequently asked questions

How to automate my business?

Log every handoff for one week, price the hours it eats at the Bureau of Labor Statistics median administrative rate of 23.23 dollars an hour, pick the single job that repeats weekly and runs on a written rule, build it on a free tier first, and give it a named owner and a failure alert. One job, finished and measured, then the next.

How can AI automate my business?

It handles the parts a rules based connector cannot: reading a message and deciding what it is about, pulling the figures out of a document, drafting the routine reply. The Census supplement found writing, document analysis and information search to be the leading uses at the level of the worker, which is exactly that. The plumbing between your systems is still ordinary automation. The judgement inside a step is where a model earns its place.

How can I use AI for small business if I am not technical?

Use what is already inside the software you pay for before you buy anything new, because it is configured against your data and costs nothing extra. Then add one connector platform on its free tier. Being non technical is not the constraint people assume. Knowing your own process well enough to write the rule down is the constraint, and you already have that.

How to automate an Airbnb business?

The handoffs are unusually clear in short lets, which is why it is a good first case. Booking confirmed triggers the message sequence, the cleaner's calendar entry and the door code. Checkout triggers the review request and the restock check. All of that is trigger and rule work for a connector platform, and none of it needs a model. Keep pricing and any dispute with a guest in human hands.

How to learn AI for small business?

Not through a course. Take the job at the top of your handoff log, build it yourself on a free plan, and break it a few times. A fortnight of that teaches more about what these tools do and do not do than any curriculum, because the failures are yours and they happen in your own data. Learn the second time by reading what the vendor documents about limits and pricing before you commit to a plan.