Does AI pay off for small business? How to measure it

What did the Apollo economist actually say about AI?

In late October 2026, Torsten Slok, chief economist at Apollo Global Management, wrote that AI has delivered "zero financial payoff" to corporate America outside the tech sector. His evidence is what he calls the S&P 493: the S&P 500 minus the seven tech giants. In that group, profit margins and revenue per worker have barely moved, despite heavy AI spending. The Magnificent Seven have pushed their margins close to 30%, while everyone else sits around 10%. Productivity data shows no clear speed-up either. The line people remember: AI may be a great calculator, but in business, intelligence is mostly judgment. It's a sharp, credible reading of aggregate numbers. It also reached every owner who's been wondering whether that monthly AI subscription is earning its keep.

Read the note closely and you'll find a second point that matters more to you. Spending data cited by Apollo shows that the top 10% of firms account for 99.5% of AI inference spending. Put simply, most companies haven't started using AI seriously, day to day. That isn't proof that AI can't pay off. It's evidence that the payoff hasn't spread yet, and Slok himself treats the question as open. Big-company AI projects also come in a big-company shape: a pilot, a steering committee, a vendor contract and a dashboard nobody opens after week three. When the goal is "transform the business," nobody can tell on a Friday afternoon whether it worked. That vagueness is exactly where the money disappears.

A small business doesn't live inside those averages. You don't have a transformation budget, and you don't need one. Your question is narrower and much easier to answer. Did this tool give me back hours I'd otherwise spend on admin? And were those hours worth more than I paid for it? That's how Slok's note connects to your week. For a five-person agency, a plumber or a solo consultant, AI will pay off through boring, specific jobs, if it pays off at all. The reminder that fires on time. The client who gets asked about a late invoice. The contract that gets read before you sign it. The rest of this guide covers how to measure that honestly and how to spot the spending you won't get back.

Why do big-company AI projects fail to show a return?

Big-company AI projects usually fail to show a return because they aim at fuzzy goals instead of specific tasks, which leaves nothing concrete to measure. You can't check "use AI to boost productivity." You can check "stop missing the Friday timesheet deadline." Large firms also bolt AI onto old processes instead of removing a step, so people end up doing the same work with a new tool on top. Then there's adoption. An unused license costs exactly as much as one used every hour. And when the output needs judgment, like pricing a deal or calming an angry customer, someone still has to check every line, so the time saved quietly vanishes in review. None of this is inevitable. It's a habit of scoping projects badly.

Here's my contrarian take: right now, most small businesses should spend less on AI, not more. The pitch you'll hear is "AI strategy," "AI transformation," a custom chatbot for your website or an all-in-one platform with twelve modules. That's the S&P 493 mistake on a smaller scale. You pay for features you'll never switch on and a learning curve you never finish. The owners who get real value pick two or three annoying chores, hand them off and ignore everything else. It sounds unambitious, and it is. It's also how you end up with a tool that earns its keep in the first month. The alternative is a subscription you quietly cancel in month six, after hoping it would start paying off.

Slok's point about judgment is fair, and it works in your favor. Admin rarely needs judgment. Remembering to send the tax paperwork, asking a supplier when a delivery will arrive, nudging the team to log their hours: none of it takes wisdom. It takes someone who doesn't forget. AI handles that job well today, and it's the job small business owners most want off their plate. Keep the judgment calls for yourself: which client to let go, which quote to accept, what to say when a deal goes sideways. Hand off the reminding, the asking and the reading. If a tool promises to make your decisions for you, be skeptical. If it promises to stop you dropping balls, you can test that within a week.

How do you measure whether AI pays off in hours saved?

The simplest way to measure AI payoff is hours saved per week, multiplied by what an hour of your time is worth, minus what the tool costs. Everything else is decoration. Take a baseline before you buy anything. For one normal week, write down every admin chore and roughly how long it takes: chasing paperwork, setting reminders, rereading supplier emails, digging through PDFs, nudging the team. Be honest, and count the interruptions too. Getting back into an invoice after a phone call costs more than the two minutes the invoice itself takes. Then pick a fair hourly value. For most owners, that's what you'd bill a client or what you'd pay someone else to do the task. Take the lower figure if you want to be conservative.

Next, run the tool on the same chores for two to four weeks and log the time again. Don't count the novelty. The first few days always feel faster because you're excited, and slower because you're still learning. Week three gives you the real number. Count the misses that stopped happening, too. A forgotten follow-up that lost you a client, a late fee because a reminder got buried, a duplicate invoice you paid twice: those cost real money, not just time. If a tool prevents one missed payment a quarter, that alone might cover its cost. Write the comparison down in a single note. If you can't name a chore that got shorter or a mistake that stopped, it isn't paying off yet.

A worked example keeps this honest. Say you're a freelance designer billing by the hour. Your baseline week shows about forty minutes asking clients for feedback and files, half an hour juggling reminders and an hour reading a contract or two. Move all three into one assistant, and suppose you win back roughly half of that time. That's about an hour a week, or four hours a month. Multiply by your rate, subtract the subscription, and you have your answer in your own numbers, not an economist's averages. These figures are only an illustration, not a promise, and yours will differ. What matters is the method: three chores, one baseline, one month. Anything you can't measure this way goes on the "nice to have" pile.

Which admin tasks give a small business the clearest AI payback?

The clearest payback comes from tasks that are frequent, dull and easy to forget: recurring reminders, follow-ups with clients and suppliers, and reading documents you'd otherwise skim. Reminders are the easiest win to prove. "Remind me to call the accountant every Monday at 9" either fires or it doesn't, and you'll know within a week. Group reminders do the same for a team. A weekly standup nudge in the shared chat means you stop being the one who nags. With Agent Sonic, all of this happens in a normal WhatsApp chat, so there's no new app to learn and nothing you'll forget to open. If bookings are your main headache, compare WhatsApp tools that schedule appointments and send reminders.

Follow-ups are where hours really leak. Writing "Hi, just checking on…" to three clients, waiting, then remembering to check again eats more of your week than you'd guess. Sonic's outreach tasks (Pro plan, currently in Beta) handle this in plain words: "Ask the three clients who haven't paid March invoices when I can expect payment and let me know." Before anything is sent, you see the exact first message for each person and reply "yes," "no" or "friendlier, no emoji." Sonic writes from its own number, introduces itself as your assistant, follows up once at most and reports back. It asks a polite question. It isn't debt collection, and it doesn't take payments. Here's how an AI assistant messages people for you, step by step.

Documents make up the third bucket. Forward a supplier PDF and ask, "Is this a duplicate of anything I paid last month, and does the math add up?" Send a contract and ask for the deadlines, the auto-renewal terms and anything that looks one-sided. You still make the decision, but you're working from a clear summary instead of page eleven at midnight. For solo operators the gains stack up, because reminders, follow-ups and document reading all happen in the thread you already keep open for clients. That's the idea behind running a solo business from WhatsApp: fewer tabs, fewer apps, fewer things to remember. If your week is mostly meetings and you rarely touch paperwork, you'll gain less, and it's better to know that before you pay.

How do you spot AI spending that won't pay back?

AI spending won't pay back if you can't name the chore it replaces, the person who'll use it every day and the number you'll check after a month. If any of those three is missing, don't buy. Watch for the usual warning signs. Annual contracts before you've had a trial. Per-seat pricing for people who'll log in twice. A "platform" that needs setup, training and a weekly dashboard check, which is just more admin with an AI label on it. A custom chatbot for a website that gets a handful of visitors a day. And any tool whose main selling point is that your competitors are buying it. That's how the big firms in Slok's data got where they are: plenty of spending, not much to show in their margins.

Trust is the other hidden cost. An assistant that acts for you, whether that's sending messages, touching your calendar or reading contracts, only saves time if you don't have to double-check everything it does. Before you hand over anything a client will see, ask four questions. Does it show you the exact message before sending? Does it ask permission first? Does it report replies accurately? Can you stop it instantly? A single "no" means worry and cleanup will eat the time you saved. Read more on whether it's safe to let an AI send messages for you, or use this broader checklist of what an AI assistant should be allowed to read and do. For the record, Sonic previews every first message, and one command cancels a task.

So, does AI pay off for small business? It does when you treat it like a new hire, not a strategy. Give it a short job description, check its work for a month and keep it only if the hours add up. Start with one recurring reminder, one follow-up and one document. Measure, then decide. If you want to run that test inside WhatsApp, you can get access to Agent Sonic. Leave your details and you'll get Sonic's number and a short onboarding. For plan details, the pricing page on the Agent Sonic site shows what's included and which features need Pro. And if a month passes and you can't point to a single hour you got back, cancel. That's what honest ROI looks like, and Slok would probably agree.

Frequently asked questions