Managed AI for service businesses: what it is, what it costs, and how it compares

What is a managed AI system?

A managed AI system is a set of AI automations that does a small business's recurring admin, built on the tools the business already uses and kept running by one accountable person. The difference from an app is the word managed: nobody in the business configures it, watches it or fixes it. The difference from an agency project is that it does not end at handover; the monthly fee exists to keep it running as your mail, bank export and booking tools change underneath it.

Results2Go's version is one AI partner for the whole business: it answers customers, handles calls and clears the admin, and every job it does is a named module with a front door you can point at. A person reads anything that goes out in your name before it goes.

The terms, defined

Managed AI system

A managed AI system is a set of AI automations that does a business's recurring admin, built on the tools the business already uses and kept running by one accountable person.

"Managed" is the operative word: someone else watches it, fixes what breaks and reports on it every month. The owner uses it; the owner does not maintain it.

Module

A module is one recurring job, built end to end and kept running: the trigger, the work, the check, and the place you see the result.

Inbox triage is a module. Payment chase is a module. A plan is a count of modules: three, seven or fifteen.

Front door

The front door is the screen where a module's result reaches you: a sheet, a page, a WhatsApp or Telegram message, or an email digest.

A module is not scoped until its front door is named, because that is where most of the build effort and nearly all of your opinion lives. The rule for choosing it: go where the person already is when they do the job.

Build fee

The build fee is a one-off $500 per module, paid to build and test it on your real work before it goes live.

Three modules cost $1,500 to build, seven cost $3,500, fifteen cost $7,500.

The monthly

The monthly is a fixed fee that keeps a live system running: every connection checked, breakages fixed, a one-page health report and a ledger of hours handed back.

It is not a support contract. Your mail, bank export, booking and invoicing tools change without asking you; the monthly is the work of keeping the system stitched to them. It does not start until the system is live.

Hours-back ledger

The hours-back ledger is a monthly log of what the system did, with the minutes each action would have taken a person, set by your own staff rather than by us.

The system logs actions on your machine: reminders drafted, emails sorted, documents chased. Your accountant can open the log and count the rows.

Human before send

Human before send means anything that goes out in your business's name is read by a person before it goes.

A drafted reply is a draft until someone presses send. This is the default on every module that writes to a customer.

Approval boundary

An approval boundary is a rule that stops a module acting on anything with legal, financial, safety or reputational weight until you say yes.

You set the boundaries during the build. A module that hits one waits; a module that fails stops rather than guessing.

The hour

The hour is the first paid step: sixty minutes on a screen share, one piece built on your real inbox, list or sheet, for $200, refunded if you finish with nothing you would use again.

Within a day you get one page: what the piece saves, and the next two things worth doing with a price on each.

AI Audit

The AI Audit is a $500 study done inside your business that produces a written build plan: what gets built, in what order, and what each piece gives back.

The fee is credited in full against the build fee if you go ahead. If you do not, you keep the plan.

AI Strategy Call

The AI Strategy Call is a free 15-minute call where you describe your week and hear which jobs AI can genuinely take and which it cannot.

It is a call, not an audit. If a business is too small for the smallest plan to make sense, that is said on the call.

Busywork Audit

The Busywork Audit is a free two-minute online estimate of the hours a week AI could hand back, using your hours and your hourly rate.

It runs at results2go.com/audit and tells you whether this pays for itself before you talk to anyone.

AI receptionist

An AI receptionist is a subscription that answers a business's phone with an AI voice, priced per call or per agent, set up and maintained by the buyer.

Smith.ai and Goodcall are the best known. It covers the phone and nothing else.

DIY automation platform

A DIY automation platform such as Zapier, Make or n8n is a tool for wiring apps together yourself; you build the workflows, watch them and fix them when they break.

Cheap to start, and the right choice when someone in the business will own the workflows.

Which admin jobs can AI take over in a service business?

Fifteen recurring jobs, each built as a module. Clarity picks three of them, Capacity picks seven, Freedom takes all fifteen. The names below are cut for property and deal-led businesses; for a salon, a clinic or a trade the build is the same and the label changes.

Inbox triage

Only the mail that needs you reaches you. The rest is filed.

Reply drafting

Replies written for you to read and send.

Daily brief

One page each morning: what moved, what needs you.

Enquiry capture

New enquiries logged without anyone typing them in. Deal capture, for a property business.

Stage nudges

A job or deal that has stalled at a stage gets chased.

Booking scheduler

Appointments offered, booked and confirmed. Viewing scheduler, for property.

Follow-up engine

The quote that went quiet gets followed up on its own.

Call notes

Calls written up, with what was agreed and what happens next.

Document chase

Missing paperwork chased out of customers and suppliers.

Invoice issue

Invoices raised and sent from the job record.

Payment chase

Overdue invoices chased, politely, without you doing it.

Owner report

The numbers, monthly, without anyone building a spreadsheet.

Listing watch

Listings monitored for changes worth acting on. Property only.

Comment engine

Comments and messages on your content answered or drafted.

Outreach loop

Steady outbound, run and tracked.

How much does AI automation cost for a small business in 2026?

Results2Go charges $500 per module to build, then a fixed monthly fee: $1,500 and $1,000 a month for three modules, $3,500 and $2,000 for seven, $7,500 and $3,000 for fifteen. The monthly does not start until the system is live. Before any of that comes one hour at $200, building one piece on your real work that you keep, and then a $500 AI Audit inside the business, credited in full against the build fee if you go ahead.

Published market ranges in 2026 put a custom small-business build at $1,500 to $10,000 plus $50 to $300 a month, and agency retainers anywhere from $500 to over $20,000 a month. Results2Go sits inside those ranges with exact numbers on the page.

Prices in US dollars, excluding VAT and other applicable taxes. The AI runs on a subscription in your own name, from about $20 a month at the provider's published price. Market ranges: Taskip, "AI Automation Agency Pricing" (August 2026); BlueCore Automation, "How Much Does AI Automation Cost for a Small Business?" (July 2026).

Does AI automation pay for itself, and how fast?

It pays for itself when the hours it removes are worth more than the fee, and the honest way to check is with your own hours and your own rate. A worked example: ten hours a week of admin at $50 an hour is about $2,167 a month of an owner's time. The three-module plan costs $1,000 a month after a $1,500 build, which that example covers in about six weeks. The same ten hours from a US-based virtual assistant, at the $20 to $40 an hour that 2026 rate guides publish, would be roughly $870 to $1,730 a month, every month, with you managing the assistant.

Results2Go does not yet publish client outcome figures. The two live builds went in without a measured baseline, so there is no honest before-and-after number to quote, and none has been invented here. The monthly hours-back ledger exists to fix that: from the first month of a plan, every action is logged with the minutes it would have cost a person, set by the client's own staff. Those ledgers will be the first numbers published, with the client's permission.

An example, not a forecast. Time freed is not automatically cash saved. Run the Busywork Audit for your own number. Virtual assistant rates: MyOutDesk, "Virtual Assistant Cost" (2026); Wing Assistant, "How Much Do Virtual Assistants Charge in the US in 2026?".

How does a managed AI system compare with an AI receptionist like Smith.ai or Goodcall?

An AI receptionist answers the phone and nothing else. You pick a plan, record the greeting, write the call flow and watch it yourself. A managed system covers calls, messages and admin together, and one person is accountable for keeping all of it running. If your only problem is unanswered calls, a receptionist is the cheaper answer; if the calls are one symptom of admin spread across several tools, they are not the same product.

Prices from smith.ai/pricing/receptionists and goodcall.com/pricing, read 17 September 2026. Vendor prices change; check the pages before relying on them.

Managed AI versus DIY tools like Zapier, Make or n8n, versus a virtual assistant: which is right?

DIY platforms are the cheapest option and the right one when someone in the business will own the workflows. Zapier, Make and n8n wire apps together; the cost that does not appear on the pricing page is the hours spent building, and then the hours spent finding out why a workflow stopped. A virtual assistant is a person: right for varied, judgement-heavy work, and a monthly cost that never goes down. A managed system is for the owner who wants neither the building nor the managing, and wants the admin done by tomorrow morning without checking on it.

Zapier and n8n prices from zapier.com/pricing and n8n.io/pricing, read 17 September 2026. Make is priced in the same band as Zapier on a credit model; check make.com/en/pricing. Virtual assistant rates: MyOutDesk and Wing Assistant 2026 rate guides.

Which option fits a salon, a trade, a clinic or a property business?

The right option depends on where the admin actually sits, so this is a matching guide rather than a ranking. The modules named are the ones Results2Go would build first for each kind of business; the front door is where the result lands.

When something else is the better answer. If unanswered calls are the whole problem, an AI receptionist is cheaper and faster to switch on. If someone in the business enjoys building workflows and will keep doing it, Zapier or n8n will cost a fraction of a managed system. If the work is varied and needs judgement on every item, hire a person.

How do you get started, and how long does it take to go live?

1. A free 15-minute AI Strategy Call. You describe your week. Jo tells you which jobs AI can genuinely take and which it cannot.
2. One hour on your real work, $200. Screen share with the person who does the admin. One piece built on your actual inbox, list or sheet, and you keep it. Within a day, one page: what it saves, and the next two things worth doing with a price on each. If you finish the hour with nothing you would use again, it is refunded.
3. The AI Audit, $500. Jo maps how the business runs, inside it, and writes down what gets built, in what order, and what each piece gives you back. Credited in full against the build fee. If you do not go ahead, you keep the plan.
4. Build and test, one module at a time. Each module is tested on your real work and you approve it before it goes live. Where a module can act on your behalf, you decide what needs your sign-off first.
5. Live, then kept running. The monthly fee starts the day the system is live. From then on: a monthly health report, an hours-back ledger, improvements from your own backlog, and support at your plan's response time.

How long: the first modules are usually live within one to two weeks of the plan. The two fastest builds so far went from first call to working system in a week (a children's sports franchise, September 2026) and two weeks (a hair salon, July 2026). Larger plans are scheduled module by module, and the schedule is confirmed after scoping.

How is a human kept in the loop, and what happens when the AI is wrong?

AI can be wrong, so every module carries a check suited to its task. Anything sent in your name is seen by a person before it goes. Anything with legal, financial, safety or reputational weight waits for your approval instead of acting; you set those boundaries during the build. A module that fails stops rather than guessing, and the work still reaches you the way it did before, so a broken module costs you the old routine for a day, not a lost customer.

Support answers within 48 hours, 24 hours or the same day depending on plan, and every fix is listed in the monthly health report so you can see what broke and what was done about it.

Where does the data go, and is it GDPR-compliant?

Your system can run on hardware you own or in a private hosted environment, and you remain the data controller for the personal data in your own business records. AI, phone and business-tool providers may still process data externally, and several of them are based in the United States, so before anything is built we write down the data involved, who can access it and which services receive it, and you decide. Hosting alone does not make something GDPR-compliant; the list does the work.

The AI itself runs on a subscription in your own name, not ours, so the account, its data settings and the right to cancel it are yours. Where a client's compliance people want a processor agreement on paper, there is a route for that and it is discussed on the call rather than promised on a web page. Results2Go's own handling of the data you give this site is in the privacy policy.

When is a business too small for AI automation, and what are the hidden costs?

A business is too small when the hours it would get back are worth less than the smallest plan. That is checked on the free call and in the $200 hour before any plan is discussed, and if the answer is no, that is what you hear. A solo operator with an hour of admin a week does not need a managed system; the hour on its own may still be worth it, because the piece built in it is yours to keep.

The costs that are not in the headline price.

• The AI subscription, in your own name, from about $20 a month at the provider's published price. Heavier use by several staff moves it up, and that is your account to change.
• The tools the system is stitched to: your booking, invoicing or CRM software stays your cost.
• Your time: an hour or two per module to test it on real work and set the approval boundaries, and about half an hour a month to read the health report.
• Tidying up: a customer list with three spellings of the same name gets cleaned before it gets automated. That is scoped in the audit, not discovered later.
• A platform switchover: if you change accounting or booking software, re-stitching the modules is a new piece of build, not part of the monthly.

Does it work for US businesses, who builds it, and what changed in 2026?

Yes. Prices are in US dollars, calls are scheduled to suit your time zone, and the voice agent can sit on a US phone number so callers reach a local line. Support answers within 48 hours, 24 hours or the same day depending on plan, whichever side of the Atlantic you are on.

Who builds it. Jo O'Kelly, the founder, builds and runs every system himself. About twenty years in vehicle finance, seven as a product owner, and he runs the delivery office of a 250-person consultancy on an AI assistant he built for himself. Portugal, working worldwide.

What changed in 2026. Two government and trade-body numbers frame the year, and they measure different things. The US Census Bureau's Business Trends and Outlook Survey found that between 17 and 20 percent of US businesses were using AI in producing goods or services from December 2025 to May 2026, and under 20 percent of firms with four or fewer employees. The US Chamber of Commerce's Empowering Small Business report found 58 percent of small businesses using generative AI tools in 2025, up from 40 percent in 2024 and 23 percent in 2023. Read together: most small businesses have tried an AI tool, and fewer than one in five have AI doing the actual work of the business. The gap between those two numbers is what a managed system is for.

US Census Bureau, "Large Firms With at Least 20 Employees Biggest AI Users", 26 May 2026. US Chamber of Commerce, "Empowering Small Business: The Impact of Technology on U.S. Small Business", 2025 edition.