AI automation services for small businesses.
You tell us what eats your week. We give you a free consultation, then build the automations around the tools you already use and hand them over running. No new platform to learn, no developer to hire.
- Free consultation
- Fixed quote, in writing
- You own what we build
Built by A8gent, Automation practice
You talk to the person who does the building, not a sales team.
Free consultation request
Tell us what you want to automate
Prefer email? hello@a8gent.com
Businesses that want the work done, not another tool to learn.
Size is not the qualifier. A one-person trade business and a fifty-person operations team both have a process that runs on copy-paste, and it gets scoped the same way: pick the workflow that costs the most hours, build it against the real systems, prove it, move on. What changes as the company gets bigger is the surrounding work - more people have to agree on what the automation should do, it has to fit software you already run rather than replace it, and there is often a security or procurement review before anything touches live data. We plan for that in the quote instead of discovering it halfway through.
We build in any industry, but five come up often enough that we have written the workflows out in detail: dental practices, HVAC and plumbing, cleaning companies, salons and barbershops, and accounting firms. If yours is not on that list, the underlying workflows repeat more than owners expect, and the consult is the fastest way to find out which one applies to you.
Owner-operated businesses
Trades, clinics, agencies, e-commerce, professional services. Two to fifty people, no in-house developer, and a growing pile of admin.
Teams drowning in repetition
If someone on your team does the same sequence of clicks every day, that is the work we take off them first.
Businesses losing money to delay
Leads that go unanswered overnight, invoices sent late, customers waiting on a reply. Speed is usually the cheapest thing to fix.
Concrete examples, not “AI transformation”.
We start with one workflow that costs you real time or money, ship it, then move to the next one.
Lead capture and follow-up
Website enquiries, missed calls, and form fills get an instant reply, get qualified with a few questions, and land in your calendar or CRM with the details already filled in.
Invoicing and payment chasing
Jobs turn into invoices without re-typing, invoices go out on schedule, and overdue ones get a polite chase sequence until they are paid or escalated to you.
Customer support and email triage
Common questions get answered from your own documents and pricing, everything else is sorted, summarised, and routed to the right person with context attached.
Reporting and dashboards
The weekly numbers you assemble by hand from three systems, pulled together and delivered to your inbox or a dashboard on a schedule.
Data entry and document handling
Details read off invoices, PDFs, forms, and emails, checked against rules you set, and written into your spreadsheet, accounting tool, or CRM.
Scheduling and onboarding
Bookings confirmed, reminders sent, no-shows chased, and new customers or staff walked through the steps without anyone remembering to do it.
Quotes, proposals, and content
First drafts built from your own templates, pricing, and past work, ready for a human to check and send rather than write from scratch.
Whatever is specific to you
Most engagements are none of the above. If a task repeats, has rules, and lives in software you already pay for, bring it to the call.
Not sure which one to start with? Run the free automation audit or just bring the question to the call. If you would rather see one of these end to end before you decide, our missed-call text-back build guide is the whole thing written out, and the audit template is the sheet we score candidate workflows against.
Four steps, and the first one is free.
- 1
Tell us what eats your time
Send a short description of the job that keeps landing on someone's desk. No technical detail needed, plain English is what we want.
You get: A reply within one business day - 2
Free consultation
A call where we walk through how the work happens today, which parts can be automated safely, which should stay human, and what it would take to build.
You get: An honest recommendation, free - 3
We build it
We build the automation against your real tools and data, with approval steps wherever a mistake would cost you. You see progress as it goes, not at the end.
You get: A working automation, plus a fixed quote agreed up front - 4
Handover
We switch it on, watch it for a while, and document how it works. You own the accounts, the workflows, and the documentation. Ongoing support is optional, not a lock-in.
You get: It runs, you own it
What actually happens, from first message to handover.
| Stage | What we need from you | What you get | Typical elapsed |
|---|---|---|---|
| First message | Three or four sentences about the job that keeps landing on someone's desk. No form with twenty fields, no NDA required to have a conversation. | A reply within one business day and a link to book the call. | 1 business day |
| The consult call | Thirty minutes, and ideally a screen-share of the inbox, spreadsheet, or CRM where the work actually happens. Watching the real thing beats any description of it. | An honest read on whether this is worth automating at all, and what the first build would be if it is. | 30 minutes, free |
| Written scope and fixed quote | One round of corrections. You are the only person who knows the exceptions, and the exceptions are what the estimate hangs on. | A written scope naming the trigger, the steps, the systems touched, the approval points, the failure path, and the definition of done. One fixed price under it. | 2 to 4 days after the call |
| Access and data check | Read access to the systems in scope, created in your own accounts, plus a sample export of the records the automation will read. | A short data-readiness note. If cleanup is needed, it is quoted as its own visible line rather than folded into the build fee. | 1 to 3 days |
| Build against your real data | Someone available for short questions. Not a project manager, just whoever actually knows the answer. | A working build you can watch progress on, running in test mode against real records with nothing sending, paying, or deleting. | 1 to 2 weeks for one workflow |
| Supervised go-live | One named person to approve the first batch of outputs before they leave the building. | The automation live, with a human approval step in front of every action that touches a customer or money. | 2 to 5 days |
| The two-week watch | Tell us which outputs looked wrong. That list is the entire point of this stage. | Fixes for the failure patterns nobody predicted, and the approval step loosened only where the log justifies loosening it. | 2 weeks |
| Handover | Nothing. This part is on us. | Credentials in your accounts, the workflow itself, written documentation, a runbook for the known failure modes, and a recurring plain-language summary of what ran and what needs attention. | 1 day |
The scope document matters more than the build
Almost every automation that disappoints was scoped badly, not built badly. So the deliverable we spend the most care on is a document nobody thinks to ask for: a written scope that names the trigger, every step in order, each system the workflow reads from or writes to, the points where a human has to approve something, and what the workflow does when a step fails. That last item is the one that separates a demo from something you can leave running. If a quote you receive from anyone does not describe its own failure path, it has not been scoped.
The definition of done goes in the same document, in your words rather than ours. “Every missed call gets a text inside sixty seconds and appears as a lead in Jobber before the office opens” is a definition of done. “Improve lead response” is not. The proposal template is the format we use, published in full so you can hold anyone quoting you to the same standard, and the discovery call script is the list of questions the call works through.
Why the first build runs with nothing switched on
We build against your real records, not against invented test data, because invented test data is always tidier than reality and hides exactly the cases that will break the thing. But we build with the outbound side disconnected. You get to read the hundred messages the automation would have sent before a single one goes out, which is the cheapest possible moment to discover that the tone is wrong, that forty of them are duplicates, or that six of them would have gone to people who asked never to be contacted again. Duplicate protection and opt-out handling are built in from the start rather than bolted on after the first complaint.
For anything that sends text messages to US numbers, registration under A2P 10DLC has to be in hand before delivery is reliable. It is a small recurring fee and a real amount of paperwork, it can take days rather than hours, and it is the single most common reason a messaging build slips. We start it at the access stage rather than the launch stage, and the registration walkthrough is published free if you would rather handle it yourself.
The two weeks after go-live are part of the job
Every automation mishandles something in its first fortnight in a way nobody predicted. That is not a sign of a bad build, it is a property of real customers. So the watch period is scoped and priced in rather than sold afterwards as support. During it we read the logs, you tell us which outputs looked wrong, and the four or five genuine failure patterns get fixed. Only then does the human approval step get loosened, and only on the specific actions where two weeks of records say it is safe. Some approval steps never come off, and for anything that issues a refund or cancels a booking, they should not.
What handover actually contains
Handover is a list of specific artefacts, not a phone call where we say it is finished. You receive the workflow itself in an account you own, every credential created under your own billing, written documentation of what runs when, a runbook naming the known failure modes and what to do about each, and a recurring summary in plain language covering what ran, what it saved, and what needs a human. Nothing we build lives in an account we control, which is the practical test for vendor lock-in: if the answer to “what happens if you disappear” is anything other than “it keeps running”, the arrangement is wrong. Ongoing support exists, and it is optional. The hiring checklist has the exact wording to use when asking anyone else the same question.
Monitoring is a deliverable, not a nice-to-have.
The kind of build we do, by industry.
HVAC and plumbing: after-hours call capture
Trigger: an inbound call rings out, hits voicemail, or arrives outside dispatch hours. Systems touched: your business phone number, a messaging provider, the on-call technician's phone, and your field service platform, which is usually ServiceTitan, Housecall Pro, or Jobber. Realistically: the caller gets a text within seconds asking whether this is an emergency, genuine emergencies escalate to the on-call phone with a second contact behind them, everything else gets a booking link, and by the time the office opens each conversation already exists as a lead in the field service platform rather than as a voicemail nobody has played. The hard part is not the code. It is writing down precisely what counts as an emergency and what the on-call technician may attend without waking the owner. The full build is on the HVAC and plumbing page, with the running-cost detail in the AI receptionist cost breakdown for HVAC.
Dental: recall and reactivation from dormant lists
Trigger: a patient's last completed hygiene visit passes a threshold you set, commonly 7, 9, 12, and 18 months, and no future appointment exists on the record. Systems touched: the practice management software, a messaging channel, and the front desk's task list. Realistically: the overdue list is pulled on a schedule, filtered for anyone already booked, flagged do-not-contact, or messaged inside your cooling-off window, then worked as a short sequence with replies routed to a human and bookings written back so people drop out the moment they act. It is not new demand, it is people who already chose you, which is why it tends to be the first build worth doing in an established practice. Scope discipline matters here: this is front-office automation, and nothing we build reads an image, suggests a finding, or writes into a clinical note. See the dental practice page and the Open Dental recall build, or the cost side in the receptionist cost page for dental practices.
Cleaning companies: quote reply and cancellation backfill
Trigger: a quote form submission, or a recurring job that cancels, skips, or moves through any channel including a phone call logged at the office. Systems touched: the website form, a messaging provider, and a booking or field service platform such as Jobber, Housecall Pro, or in plenty of cases a shared calendar and a spreadsheet. Realistically: enquiries get a priced or bracketed reply while the person is still on your site rather than the next morning, and a freed slot is offered to a waitlist in order instead of staying empty. Backfill is the one most owners underrate, because an unfilled slot is not a deferred cost, it is revenue that cannot be recovered later. Speed to lead is doing most of the work in the first half. The workflows are on the cleaning company page and the build is written out in the quote-to-booking guide.
Salons and barbershops: confirmation cascade and rebooking
Trigger: an appointment approaching, with each stage firing only if the previous one did not resolve, or a client passing their service-specific rebooking window with nothing on file. Systems touched: the scheduling platform, which is usually Fresha, Vagaro, Booksy, Square Appointments, Phorest, Mindbody, or Boulevard, plus messaging and sometimes Instagram or WhatsApp. Realistically: the point is not the reminder, it is the cascade. A silent client at the second stage gets a different message than one who confirmed, and a slot released early goes to a waitlist rather than to nobody. A no-show is an hour of chair time you cannot resell, which is what makes no-show rate the number worth attacking first here. What limits the build is almost always the scheduling platform's API, which varies more in this category than in any other, so it is the first thing we check rather than something we assume. The detail is on the salons page and in the win-back build guide.
Accounting firms: document chasing and deadline escalation
Trigger: an engagement needing specific documents that have not arrived, or a statutory deadline approaching on any open engagement. Systems touched: practice management such as Karbon, TaxDome, Canopy, Senta, or Jetpack Workflow, ledger software such as Xero, QuickBooks, or Sage, a document portal, and an e-signature tool. Realistically: chasing becomes per-document rather than per-client, so a client who has sent four of six items gets asked for the two that are missing instead of a generic nudge, and the internal escalation fires before the deadline rather than after it. The value here is risk rather than hours: the cost of a missed statutory deadline is not measured in admin time. Integration depth differs sharply between those platforms, which is checked against your actual account rather than the documentation. See the accounting firm page and the invoice chasing guide.
Everywhere else: quote follow-up and inbox triage
Two patterns repeat across every trade we have written up, including restaurants, law firms, e-commerce and retail, and real estate agencies. The first is that quoted work is lost to silence far more often than to a competitor's price, so a structured quote follow-up cadence that pauses on any reply recovers deals nobody ever actually decided against. The second is that most inboxes contain three or four message types that arrive constantly and need the same answer every time, which is the natural first target for a drafting step grounded in your own documents and pricing rather than in a general language model's guesswork. If your industry is not written up yet, the industries hub shows the shape and the consult finds your version of it.
What none of these examples claim.
Custom quote after a free consult.
We do not publish package prices, because we would only be guessing at your workflow. What we do publish is how we price, so there are no surprises.
To go into the call with a number in your head, our breakdown of what AI automation costs gives market-typical build and running ranges per workflow, and the ROI calculator turns your own time-saved figures into a payback period.
You should compare us with other options rather than take our word for it. The consultant versus agency versus DIY comparison is honest about where an independent builder is the wrong choice, and the hiring checklist lists the questions we would want you to ask us. If you just want the written recommendation with nothing attached, start with the free automation audit.
- The consultation is free, and you are under no obligation afterwards.
- We quote per project, in writing, after the consult - once we know what the work actually involves.
- Fixed scope and fixed price, so there are no open-ended hourly bills.
- You pay any third-party software costs directly, in your own accounts, so nothing is marked up.
- If we do not think an automation will pay for itself, we will tell you on the call instead of selling you one.
What a given budget buys, in scope rather than adjectives.
| Workflow | Typical build effort | Tool subscriptions | Typical monthly running cost |
|---|---|---|---|
| Missed-call text-back | Roughly $800-$2,500. One trigger, one message, a handoff rule. | Messaging or telephony provider plus a workflow platform, commonly $30-$120/mo combined. | $30-$120. Effectively just the subscriptions. |
| Lead follow-up sequence | Roughly $2,000-$6,000. Capture, qualify, write to CRM, schedule, escalate if no reply. | CRM seat, workflow platform, AI model usage. Commonly $80-$300/mo. | $100-$350. Copy and questions need revisiting as your offer changes. |
| Invoice chasing | Roughly $1,500-$5,000. Reads the ledger, sequences reminders, stops on payment, escalates. | Accounting tool API access plus workflow platform, commonly $40-$150/mo. | $50-$200. Low churn once tone and escalation rules settle. |
| Support and email triage | Roughly $3,000-$12,000. Classification, drafting from your own documents, routing, review step. | AI model usage dominates, plus a helpdesk seat. Commonly $100-$400/mo. | $150-$500. The knowledge base needs updating and misclassifications need reviewing early on. |
Read the table as a shape rather than a price. The dominant driver of build effort is the number of systems the workflow touches, not how much AI is involved, which is why an automation built on tools you already pay for is dramatically cheaper to own than one that introduces a new platform. The full breakdown, including the payback arithmetic and the data-cleanup line most quotes omit, is on the AI automation cost page, with the one-off side isolated on the build cost page and the recurring side on our write-up of what an agent costs to run each month.
What each budget band includes, and what it does not
| Band | What the scope typically includes | What is explicitly not in it |
|---|---|---|
| Around $2,000 | One narrow workflow with a single trigger and a single outcome, built on tools you already pay for. Basic documentation and a walkthrough. | Data cleanup, a new platform, multi-system branching, or a long support tail. Treat it as proving the idea rather than finishing the job. |
| Around $8,000 | One multi-step workflow across two or three systems plus a smaller second one. Some data cleanup included. Approval steps, logging, and a full handover with source and documentation. | Custom software, a bespoke integration where no API exists, or a standing retainer. This is the band most first serious projects land in. |
| Around $20,000 | Three to five workflows that share data and hand off to each other, meaningful data cleanup, custom integration where no connector exists, monitoring, error handling, and a defined support period. | A transformation programme, parallel delivery streams, or contractual uptime commitments. Those need a firm with a bench, and we say so. |
Structure matters as much as the total. Fixed price per scoped workflow puts the estimation risk on the builder, which is where it belongs, and it is what we quote. Hourly billing moves that risk to you and makes the discovery work you cannot supervise the very thing you are paying for. The structure to be most careful with is a subscription where the provider hosts and owns the workflow: the meter never stops, and if it lives in their account you are renting a process you depend on. Ask directly whether you receive the source, the credentials, and the documentation, and check the answer against total cost of ownership rather than the monthly number.
The number that decides it is payback, not price.
Where we are the wrong choice, named specifically.
Work we turn down
- Anything regulated as advice or as a medical device. Clinical AI that reads an image and suggests a finding is a regulated pathway involving your insurer and, in most jurisdictions, an approval process. The same goes for automated legal or financial advice. We build in the administrative layer only, and a general automation consultancy should not be selling you anything else.
- Processes with no stable rules yet. If the exception rate is higher than the happy path, or the process changed twice this quarter, automating it means rebuilding it. Write the process down and run it manually for a month first. That advice is free and it is frequently the whole recommendation.
- Workflows that fail the payback test. Something that runs twice a month and takes ten minutes is not worth a build fee, whatever it costs you emotionally. We say so on the call rather than selling it to you.
- Features you already pay for. A large share of businesses are sitting on automation capability inside their existing CRM, booking platform, or accounting tool that was never switched on. Turning that on is cheaper than anything we could build beside it, and checking is part of the free audit rather than a paid discovery phase.
- Hosting your workflow in our account. We do not resell subscriptions, mark up third-party software, or run automations you cannot access. If a provider will not tell you where the workflow lives, that is the answer.
- Strategy decks with no build attached. An AI readiness assessment that ends in a slide deck is a genre, not a service. If you want the analysis without the build, take the free automation audit and keep the report.
Who should not hire us, and what to use instead
Your workflow is two steps between two mainstream apps. If the whole job is “when a form is submitted, add a row and send an email”, paying anyone to build it is waste. Zapier has the largest connector library in the category and you can wire that yourself in an afternoon. Our n8n versus Zapier comparison is honest about where it stays the right answer well past the point people expect.
You want visual multi-step logic at a lower per-operation cost. Make.com handles branching, iteration, and error routes in a visual canvas and prices per operation rather than per task, which changes the arithmetic considerably at volume. The three-way comparison covers where each one gets expensive.
You have someone technical in-house and want to own the platform. n8n is the right destination, and self-hosting removes the per-execution meter entirely. It also moves uptime, upgrades, and backups onto your side of the line, which is a real job rather than a free one. The true cost of self-hosting n8n prices that honestly, and the workflow cost calculator models your run volume against the hosted tiers. If you would rather compare the whole field first, the platform directory reviews forty-five of them.
What you actually need is software, not an automation. If the output is a product your customers log into, a bespoke database with its own interface, or anything with a roadmap, hire a developer. Automation platforms are excellent glue and poor foundations, and a workflow tool asked to behave like an application becomes unmaintainable faster than most people expect. The build versus buy tool walks the decision, and off-the-shelf agents versus custom builds covers the middle case where a published product already does four of your five requirements.
You need parallel delivery, continuity, or procurement-grade terms. A8gent is a small operation. There is no delivery bench, no account team, no contractual uptime commitment, and no compliance certification to hand your procurement department. If you need several workstreams running at once, a named replacement if someone is unavailable, or a formal SLA, an agency genuinely beats us and the extra cost is buying something real. The consultant versus agency versus DIY comparison puts the four options side by side without pretending we win every row.
You would rather learn it than buy it. A completely legitimate outcome, and for a lot of owner-operated businesses the correct one. The build guides publish the common workflows step by step for free, the glossary defines the vocabulary a vendor will use on you, and the readiness quiz is a five-minute check on whether your process is in a shape worth automating at all.
Where an off-the-shelf product beats a build.
Would rather build it yourself?
The same system we use on client projects is taught in our courses - picking the right workflow, building it, and keeping it safe once it is live. The $29 Starter Kit is the one-hour version.
What small businesses ask before they start.
What does AI automation actually mean for a small business?
It means software doing the repetitive parts of a job end to end: reading an email, pulling the details out, writing the reply or the invoice, updating your systems, and flagging anything unusual to a human. It is not a robot running your business - it is the boring 80% handled so your team can do the rest.
How much does it cost?
We quote per project after the free consult, because a single email automation and a full lead-to-invoice pipeline are very different pieces of work. What we can promise is a fixed price agreed in writing before we start, rather than an open-ended hourly meter.
How long does it take?
We deliberately scope one workflow at a time so you get something working quickly rather than waiting on a large project. The exact timeline depends on how many systems are involved and how fast we can get access, and we give you an estimate as part of the quote.
Do I need to change the tools I already use?
Usually not. We build around the systems you already pay for - your inbox, spreadsheets, accounting software, CRM, booking tool. If something genuinely cannot be connected, we will say so on the call before you spend anything.
Is my business data safe?
We only ask for access to what the specific workflow needs, keep credentials in your own accounts wherever possible, and build approval steps in front of anything that sends, pays, or deletes. You can revoke our access at any time.
What if the automation makes a mistake?
That is what the review steps are for. For anything customer-facing or financial, we default to a human approving the output until you are confident, and we build in logging so you can see exactly what happened and why.
Do I own what you build?
Yes. The accounts, the workflows, and the documentation are yours. You can keep us on for changes and support if you want to, but you are not locked in and nothing stops working if you do not.
Do you only work with small businesses?
No. The pages here lean on small-business language because that is the most common starting point, but the method does not change with headcount. Larger engagements get scoped identically - one workflow at a time, fixed scope, fixed price, you own the result. Being honest about what is different: with a bigger company there are usually more stakeholders in the definition of done, more existing systems the automation has to slot into, and a security or procurement review to clear first. Those go into the scope up front. A8gent is a small operation, so what you get is the person who builds it rather than an account team.
I would rather learn to do this myself. Can I?
Absolutely, and we would rather you did if you enjoy that. The same system we use is taught in our courses, and the $29 starter kit covers picking your first workflow and shipping it. Plenty of clients do both: learn the basics, then hand off the harder builds.
Tell us what eats your week.
The consultation is free and the advice is honest. If automation is not worth it for your business, we will say so on the call.
