What a Small Business Should Automate First

What should a small business automate first? The process with paying customers already waiting in it, and for most service businesses, from an HVAC shop or a dealership service lane to a Las Vegas law firm, that is the phone. The Federal Reserve's 2025 Small Business Credit Survey, published in March 2026, found just under half of small firms now use AI, with the top uses being writing and marketing, individual productivity, and planning and analysis. Answering customers is not on that list, even though phone calls convert 10 to 15 times better than web-form leads. This post ranks the uses the Fed found against the phone and gives you a scoring sheet to set your own order.

Small businesses mostly automate writing first, the Fed survey found

The Fed Communities summary of the 2025 Small Business Credit Survey, published March 3, 2026, draws on more than 6,500 small employer firms and more than 5,200 nonemployer firms, all with fewer than 500 employees, surveyed from September to November 2025. Its headline on AI is plain: just under half of firms use it. The top reported uses are writing and marketing, individual productivity, and planning and analysis.

Two more findings matter for anyone deciding what comes next. Most firms had not seen any change in labor costs from adopting AI, and a majority reported productivity increases. Put together, the typical small business has used AI to make existing work faster for the people already doing it. It has not used AI to take on work nobody was doing, like answering the call that rings while everyone is busy.

That pattern makes sense. Writing tools are cheap, need no integration, and one employee can start using one before lunch. The phone is harder: it touches your calendar, your pricing, and the customer's first impression of you. Easy to start is a good reason to try something. It is not a good reason to call it your automation strategy.

Treat the adoption number with care. The survey is a convenience sample, adoption is self-reported, and 'uses AI' is a low bar that can mean one person occasionally drafting an email. We looked at a different measure of the same question in our September post on small business AI use.

Owners say reaching customers is their top problem, then automate something else

The companion 2026 Report on Employer Firms, also published March 3, 2026, covers 6,525 responses fielded September 3 to November 14, 2025. Asked about operational challenges, employer firms most often named reaching customers and growing sales. Hiring and retaining qualified staff came second.

The mood behind that answer is cautious. The revenue expectations index fell from 39 to 33 and the employment expectations index fell from 26 to 23, both the lowest since the 2020 survey. More than four in 10 firms reported higher costs from tariffs as a financial challenge, and 77 percent reported tariff costs, rising costs, or both. The Fed Communities summary adds that nearly half of firms source some inputs from outside the United States, and a large majority of those saw input prices rise from 2024 to 2025.

Here is the tension. An owner whose top problem is reaching customers, whose costs are rising, and who expects slower revenue growth has the most to gain from not losing the customers who already reached out. Yet AI use goes mostly to producing more writing and marketing output, which is a bet on reaching new people. The cheaper win is usually answering the people who already called.

The staffing finding points the same way. If hiring is the second biggest challenge, the answer to a ringing phone during a busy afternoon is rarely another hire. Like the summary, this report is a convenience sample weighted to the national small employer population, not a probability sample, so read the levels as directional and the year over year changes as the stronger signal.

The phone ranks first on revenue evidence and last on adoption

Here are the uses the Fed found, plus the phone, scored on what matters for sequencing: does the use touch revenue directly, can you measure what it produces, and is there evidence it moves money. Adoption figures for individual uses are not broken out in the summary, so the first column records where each use stands, not a percentage.

Use case	Reported adoption (Fed, 2025 survey)	Touches revenue	Measurable output	Evidence it moves money
Writing and marketing	Named among the top uses	Indirectly	Posts, emails, ads drafted	None in the survey; gains reported as productivity
Individual productivity	Named among the top uses	Indirectly, through time	Hours saved	Majority report productivity gains; most saw no labor cost change
Planning and analysis	Named among the top uses	Indirectly	Forecasts, reports	No revenue measure in the survey
Answering the phone	Not named among the top uses	Directly	Calls answered, jobs booked	Calls convert 10 to 15 times web forms; 75% of unreached callers go to a competitor

The phone figures come from vendor and trade studies, summarized on our missed-call statistics page. The conversion figure is BIA/Kelsey's. In a 411 Locals study that monitored 85 businesses across 58 industries for 30 days, 62 percent of calls went unanswered: 37.8 percent were answered, 37.8 percent went to voicemail, and 24.3 percent got no response at all. Roughly 80 percent of callers who reach voicemail hang up without a message (Forbes, Marchex, BIA), 75 percent of callers who cannot reach a business call a competitor instead (BrightLocal), and 78 percent of customers buy from the first company that responds (Velocify).

Those studies are small or do not disclose their samples, so treat them as direction, not precision. But notice what the table says. The three uses owners adopted first carry no revenue evidence in the survey; the reported gains are productivity. The one use with revenue evidence attached sits outside the top of the list. If you fix one thing this quarter, the evidence leans toward the channel where money arrives, not the channel where content leaves.

Score each process by volume, completion rate, and revenue per unit

Adoption surveys tell you what other owners did. They cannot tell you what to automate in your business. For that you need four numbers per candidate process, and you can collect them in a week.

The score is the dollar value of work that currently does not get finished: monthly volume, times the share left uncompleted, times the revenue each completed unit brings in. It is a ceiling, not a forecast, because no fix completes everything. But it ranks processes in the right order, and it forces the revenue question onto tasks that usually dodge it.

Below, the phone is worked as the filled example for an assumed HVAC and plumbing shop. Assume 400 inbound calls a month, 240 answered live (a 60 percent completion rate, better than the 37.8 percent in the 411 Locals study), 30 percent of answered calls booking a job, and a $450 average ticket. Revenue per completed call is 0.30 times $450, or $135.

Process	Monthly volume	Current completion rate	Revenue per completed unit	Priority score: volume x (1 minus completion) x revenue
Inbound phone calls (worked example)	400 calls	60% answered live	$135 per answered call	400 x 0.40 x $135 = $21,600
Marketing posts and emails (worked example)	8 pieces planned	50% published	$200 assumed, not measured	8 x 0.50 x $200 = $800
Quote and estimate follow up	Your count	Your rate	Your number	Your score
Your next candidate

The phone row scores about 27 times the marketing row ($21,600 divided by $800). Even if you halve the phone's revenue per call and double the marketing value, the phone still scores more than six times higher ($10,800 against $1,600).

Two adjustments make the phone number more honest. First, not every unanswered call is lost. Roughly 80 percent of voicemail callers hang up without a message, so if all 160 missed calls reached voicemail, about 32 would leave one and some would book on callback. Counting only the 128 who leave nothing, the score is 128 times $135, or $17,280. Second, Marchex's 2025 call analytics found 34 percent of calls to service businesses arrive outside standard business hours. On 400 calls that is 136 after hours, so a shop that only answers during business hours cannot complete more than 264 calls, or 66 percent, no matter how attentive the front desk is.

That second point is why the phone is a different problem from writing. Faster writing makes the existing person more productive. A phone gap is often a coverage gap, and productivity tools do not cover hours when nobody is working. That is the job an AI receptionist or after-hours answering is built for, and it is why the score, not the adoption chart, should set your order.

A five person law firm scores lower but still puts the phone first

Smaller volume gives a smaller number, not a different order. Take an assumed five person family law office with 120 calls a month and half answered live, because the attorneys are in court and the one assistant is often on another line. Assume 1 in 20 answered callers retains the firm at an average first fee of $3,000. Revenue per answered call is $3,000 divided by 20, or $150.

The score is 120 times 0.50 times $150, or $9,000 a month. Against the same $800 marketing row, the phone is still more than 11 times larger. Apply BrightLocal's 75 percent figure to the 60 missed callers and roughly 45 people a month reached the firm, got nobody, and moved on to the next name on the list.

Now scale up to an assumed dealership service lane taking 1,500 calls a month and answering 70 percent. At an assumed $90 per answered call, averaged across simple maintenance and large repair orders, the score is 1,500 times 0.30 times $90, or $40,500. Bigger operations usually answer a larger share, but the volume keeps the gap large in dollars. In all three cases the ranking is the same: phone first, writing tools later.

Where the phone does not belong at the top of the list

The ranking flips in some businesses, and the score shows when. A restaurant that takes most reservations online might get 150 calls a month, answer 80 percent, and earn an assumed $25 per completed call. Its phone score is 150 times 0.20 times $25, or $750, about the same as the $800 marketing row. For that restaurant the phone is not the obvious first project, and starting with marketing or scheduling is reasonable.

The phone also drops when revenue per completed unit is low or unknown. A retailer whose calls are mostly about store hours, or an online brand where the phone is a support line rather than a sales line, will see small numbers in the third column. If you cannot estimate what an answered call is worth, measure that before buying anything.

It drops when completion is already high. If you have a full-time receptionist plus after-hours coverage and rarely miss a call, the gap is small, and your best next process is probably quote follow up, intake paperwork, or something internal.

Finally, the outside numbers here are weak on their own. The Fed adoption data is self-reported from a convenience sample, and the phone studies are vendor and trade-association work with small or undisclosed samples. That does not change the method, because the method runs on your counts, not ours. It does mean you should not quote the 10 to 15 times conversion figure to a lender as a fact about your business.

Questions owners ask

What should a small business automate first if money is tight?

Automate the process with the largest score: monthly volume, times the share left unfinished, times revenue per finished unit. For most service businesses that is the phone, because calls convert 10 to 15 times better than web forms and missed callers rarely leave a message. If your phone already gets answered around the clock, move down the list to the next highest score.

Is using AI for writing and marketing a waste of money?

No. A majority of firms in the Fed survey reported productivity increases, and drafting is where many started. The issue is order, not value. Writing tools make existing staff faster but produce output whose revenue is hard to measure. If the people who already called are not getting answered, more marketing simply pushes more callers into the same gap.

How do I find out how many calls my business misses?

Pull a month of call logs from your carrier or VoIP system and count total inbound calls against calls answered live. Separate the after-hours share, since roughly a third of service calls arrive outside business hours in Marchex data. Then look at voicemail: most callers who reach it hang up without a message, so the voicemail count understates the real loss.

What to do this week

1. Count your calls. Pull 30 days of logs and record total inbound, answered live, and after hours. That fills the first two columns of the phone row.
2. Put a dollar figure on one completed unit. For the phone, multiply your booking rate by your average ticket. For marketing, write down what one post earned last quarter, even if the honest answer is that you do not know.
3. Score three processes and start with the top one. Use the sheet above. If the phone wins, read our AI receptionist cost breakdown before calling vendors; if it does not, our AI strategy consulting page explains how we sequence the rest.

If you want someone to check your scores with you, the free 15-minute audit on our business AI solutions page is a quiet place to start.