A8gent
Decision guide . 2026

AI consultant vs agency vs DIY: what to hire.

Five ways to get automation built, compared on the four things that actually differ: what it costs, how fast it lands, whether you own it, and who keeps it running afterwards. Including an honest section on when not to hire us.

  • 5-way comparison table
  • Budget decision tree
  • When not to hire us
The options

Five routes, side by side.

Everyone selling one of these will tell you it is the right one. The differences are real but they are not about quality - a good freelancer beats a bad agency every time. They are about risk, ownership, and where your money goes.

OptionTypical costSpeedOwnershipMaintenance
DIY / no-code toolsSubscriptions only, commonly $30-$200/mo, plus your time. The time is the real cost and it is easy to underestimate.Slow at first, fast later. Expect weeks on your first workflow, days on the third.Total. You built it, you understand it, nothing to hand over.Yours, permanently. Fine while you remember how it works; a problem when the person who built it leaves.
Marketplace freelancerCheapest paid route. Often a few hundred to low four figures for a narrow task.Fast on a well-specified brief. Slow and painful on a vague one.Depends entirely on the brief. Ask for source and documentation explicitly or you will not get them.Weakest link. Continuity is unreliable; the same person may not be available in six months.
Solo consultantMarket-typical low-to-mid four figures per workflow; high four to low five figures for pipelines.Fast on one or two workflows. A queue forms if you want several at once.Should be total, if you insist on it. Source, credentials, and documentation in your accounts.Handed to you, with optional support. Key-person risk is the honest downside.
AgencyHighest. Expect a meaningful premium over a consultant for the same build, covering account management and delivery overhead.Fastest for several workstreams at once, because they can parallelise. Slower to start, more process.Varies widely. Some hand over cleanly; some build on proprietary platforms. Ask before signing.Usually a retainer, often good. Check whether it is optional or the automation depends on it.
In-house hireA salary, permanently, plus tooling. Only rational above a certain volume of automation work.Slowest to start - hiring takes months - then continuously available.Total, and institutional. The knowledge lives inside the business.Continuous and immediate. The risk moves to retention: when they leave, undocumented workflows leave with them.
Cost figures are market-typical planning ranges for small businesses in 2026, not quotes.

For a full per-workflow cost breakdown behind these ranges, including tool subscriptions and monthly running cost, see AI automation cost in 2026.

Decision tree

Start with your budget and the number of workflows.

Under $1,000Do it yourself, or switch on what you already pay for.

Before buying anything, check whether your CRM, booking tool, or accounting package already does it. An enormous share of small-business automation requests are features nobody enabled. If not, a no-code platform plus a weekend is a genuinely good answer for a simple workflow.

$1,000 - $2,500One narrow workflow, built for you, or DIY with a paid review.

Enough for a single-trigger automation on tools you already have - missed-call text-back, booking reminders, an inbox triage rule set. Not enough for data cleanup or multi-system work, so pick the narrowest useful thing.

$2,500 - $8,000A solo consultant. This is the sweet spot.

One real multi-step pipeline plus a smaller second workflow, with approval steps, logging, and a proper handover. At this level agency overhead would eat a large share of the budget without adding anything you need.

$8,000 - $25,000Still a consultant, unless you need things in parallel.

A connected set of three to five workflows with real data cleanup underneath. The question that flips this to an agency is not budget, it is whether the work must happen simultaneously or can happen sequentially.

Above $25,000, several parallel workstreamsAn agency, or a specialist firm.

Multiple builders, guaranteed cover, formal contracting, and someone whose job is coordination. You pay for that, and above this size it is worth paying for.

Continuous six-figure automation workHire in-house, and use consultants for spikes.

Once automation is permanent work rather than a project, the salary is cheaper than the equivalent in project fees, and the institutional knowledge is worth more than either.

Two questions that override the budget bands

Does the workflow touch money or customers without a human check? If yes, do not DIY it as your first project. The failure mode is not a broken workflow, it is a wrong invoice or an inappropriate message sent to a real customer.

Will anyone maintain it in six months? If nobody in the business owns automations, buying more of them makes things worse. Fix the ownership question before the build question.

The bottom two bands deserve a fair test before you rule them out. Our missed-call text-back, quote follow-up, and appointment reminder guides are complete builds, free, and each one is a fair sample of how much of your own time the DIY column in the table above really costs. The build versus buy tool and the hire-versus-automate comparison put your own numbers against the same bands.

Trade-offs

What each option is genuinely bad at.

DIY is bad at reliability

First versions work on the happy path and fail silently on everything else. There is usually no logging, no alerting, and no fallback, so you find out from a customer.

Freelancers are bad at handover

Not through malice - the engagement structure does not reward documentation. Unless it is in the brief, you get a working workflow and no way to change it.

Solo consultants are bad at parallelism

One person builds one thing at a time. If you need four workflows live this quarter, you either wait or you hire more than one person.

Agencies are bad at small budgets

Their cost structure needs a certain project size. Below it, you are paying for account management and process rather than building, and you often get their most junior builder.

In-house is bad at starting

Months to hire, and a real risk of hiring someone who can build but has never had to maintain what they built.

Everything is bad at bad data

No option on this list fixes inconsistent records for free. Whoever builds it, the cleanup work exists and someone pays for it.

Whichever route you take, two decisions get made on your behalf if you do not make them. The first is the workflow platform, which sets your recurring bill for as long as the automation lives - n8n versus Make covers the difference that matters most, which is whether you can move the workflow elsewhere later. The second is whose model runs the AI steps, and the per-provider breakdowns are worth a look if you have opinions about cost or where your data goes. Ask any bidder to answer both in their proposal - our proposal template has the fields, and it is free to use with whoever you hire.

Honesty

When not to hire us.

We are one practitioner doing done-for-you automation, and we take work from businesses of any size - the scoping does not change with headcount. The shape of the practice does matter though: it is an advantage for some projects and the wrong answer for others. Here is where you should go elsewhere:

  • You need several workstreams delivered in parallel. We build one workflow at a time. If four things must go live this quarter, an agency can staff that and we cannot.
  • You need guaranteed cover independent of one person. Key-person risk is real with any solo consultant, including us. If your operation cannot tolerate it, that is a legitimate reason to pay agency rates.
  • You have enterprise procurement or compliance requirements. Vendor security questionnaires, formal SLAs, insurance thresholds, and audit programmes are a firm's job, not ours.
  • You need on-site presence or change management. Getting forty people to adopt a new process is organisational work, and it needs someone in the room.
  • Your budget is under about $1,000. Do not spend it on us. Check what your existing tools already do, then learn to build the first one yourself - the same method we use on client projects is what the courses teach.
  • The payback maths does not clear about twelve months. We will tell you this on the free consult rather than after you have paid.
Where we fit

And when we are the right call.

Owner-operated businesses, roughly 2 to 50 people, no in-house developer, one or two workflows that genuinely cost time or money. You want the thing built and handed over, not a platform to learn or an account team to manage. You talk to the person who does the building, the price is fixed in writing after a free consult, and you own the source, the credentials, and the documentation at the end.

Most of that work has been in appointment-driven and job-driven trades: dental practices, HVAC and plumbing firms, cleaning companies, salons, and accounting firms. Reading the page closest to your business is the quickest way to judge whether we understand the work before you spend a call finding out.

If that sounds right, the free automation audit is the way to test it: a 30-minute call and a written report within 48 hours, yours to keep whether or not you hire us. If you want to vet us properly first, run the 15-question hiring checklist - our answers to all of them are printed on that page.

The framework

Five dimensions, scored honestly.

Cost is the dimension everyone leads with and the one that decides fewest projects. These are the five that actually separate the options, with the winner named on each.

DimensionWhat it really meansBest on this dimensionWorst on this dimension
CostTotal spend to get one workflow live, including your own hours at whatever they are worth.DIY, if you enjoy the work. Nothing else comes close on cash cost.In-house hire, until automation is permanent work. A salary against one project is not rational.
SpeedTime from decision to a workflow running against real cases.A solo consultant for one or two workflows. No onboarding process, no account team, no queue.In-house, because hiring takes months. An agency is second slowest to start.
DepthHow well the person building it understands your process rather than the tooling.In-house, over time. Nobody learns your business faster than someone inside it.Marketplace freelancer, structurally. The engagement is too short to build context.
ContinuityWhether the thing is still maintainable in eighteen months by someone who is not the original builder.Agency, if handover is clean. Cover is the thing you are paying the premium for.Freelancer, then solo consultant. Key-person risk is real and worth pricing.
RiskThe chance of a wrong output reaching a customer, and what happens when it does.Consultant or agency, because approval steps and a failure path get designed in.DIY first attempts. They work on the happy path and fail silently on everything else.
Structural comparison. Individual suppliers vary far more than the categories do.

The point of scoring them separately is that almost nobody needs to win on all five. Work out which two matter for the specific workflow in front of you, and the answer usually falls out. A workflow that sends messages to customers is a risk problem. A workflow that has to be live before a busy season is a speed problem. A workflow you will change every month is a continuity problem, and that one has a different answer from the other two.

The dimension most people forget

Continuity. Everyone evaluates the build and almost nobody evaluates the eighteenth month. Ask every bidder, in writing, what a different builder would need in order to take over. The completeness of that answer predicts more about your experience than the price does, and vendor lock-in is the term for what happens when the answer is thin.
Agency

When an agency is plainly the better buy.

We are a solo practice and we still tell people to hire an agency several times a year. These are the situations where it is not a close call.

You need things to happen at the same time

One person builds one workflow at a time. If four workflows must be live this quarter because a system migration or a season forces the date, sequencing them is not an option and no amount of goodwill fixes that. An agency can put four builders on it. That is the single clearest reason to pay the premium, and it is a capacity argument rather than a quality one.

You cannot tolerate key-person risk

If the automation is load-bearing for revenue and a two-week absence would hurt, you want an organisation with cover rather than an individual with a calendar. Ask what the cover arrangement actually is, in writing, and what response time is committed to. An uptime commitment from a firm is worth something. The same words from a solo practitioner are worth less, including from us.

Procurement, security review or compliance paperwork

Vendor security questionnaires, insurance thresholds, formal contracting and audit programmes are a firm's job. If your organisation requires them, a solo consultant will either fail the process or spend your budget clearing it. This is not a comment on the quality of the build, it is a comment on who can absorb the overhead.

Change management inside a bigger team

Getting forty people to adopt a new process is organisational work and it needs someone in the room. The technical build might be two days of the project. The other three weeks are meetings, and a firm with a delivery manager is genuinely better at that than a builder is.

If you are going down that road, price it properly first. The agency pricing guide explains how agencies build their numbers, and the agency pricing calculator lets you rebuild a quote from the inputs to see what you are being charged for. Then apply the same hiring checklist to the agency that you would apply to an individual. Firms fail the ownership questions more often than freelancers do, because building on a proprietary platform is a retention strategy.

In-house

When hiring beats paying anybody.

The comparison is not a fee against a salary. It is a fee against a salary plus everything a permanent person does that a project cannot: noticing the workflow that broke quietly, maintaining the ones already live, and knowing why a rule exists. Those are worth a lot, and they are worth nothing at all if there is not enough of the work to fill a role.

The threshold is not a budget number, it is a shape. Hiring wins when automation has become continuous work rather than a sequence of projects, when enough systems exist that someone has to own them permanently, and when you can keep that person technically engaged enough to stay. Fail the last condition and you will hire, train, and re-hire, which costs more than any consultant.

Two honest warnings. First, plenty of people who can build an automation have never had to maintain one, and the interview does not surface that difference. Ask a candidate what broke in something they built and what they changed afterwards. Second, an in-house hire concentrates the continuity risk rather than removing it: undocumented workflows leave with the person who wrote them, so the documentation standard has to be enforced internally exactly as you would enforce it on a supplier.

The middle path works well and is under-used: hire for ownership, and use consultants for spikes and for the first build of an unfamiliar pattern. If you go that way, teaching the internal person the method directly is cheaper than teaching them by watching. That is what the courses are for, particularly the operators course, and the pricing page lists what each one contains. The implementation roadmap and what is an AI agent are the free versions of the same groundwork.

Identical either way

The costs that do not change with who you hire.

Three lines on the bill are the same whoever builds it. Comparing quotes without separating them is how people conclude that one supplier is cheap when they have simply excluded something.

The platform subscription

The workflow runs somewhere, and that somewhere bills monthly for as long as the automation lives. Published cloud pricing as of August 2026: n8n lists Starter at 20 euro a month billed annually for 2,500 executions and Pro at 50 euro for 10,000. Make.com lists a free tier at 1,000 credits a month, Core at $9 and Pro at $16 a month at 10,000 credits on monthly billing. Zapier prices on a different unit again. The full platform directory covers the rest including self-hosted, and n8n versus Make is the comparison to read, because the difference that matters is portability rather than price.

The model usage

Any AI step costs per use and scales with volume rather than sitting flat. The model provider pages carry published token pricing, including OpenAI, and the cheapest LLM API comparison is worth an hour if the workload is high volume and simple. Whoever builds it should give you a projected monthly figure at your volume before you commit; the running cost breakdown shows how that projection is assembled and the workflow cost calculator lets you do it yourself.

The data cleanup

Nobody automates past inconsistent records for free. Duplicates, three spellings of one status, dates stored as text: the work exists and somebody pays for it. A quote that does not mention your data has either inspected it and found it clean, which is rare, or has not looked. Two businesses asking for the same automation can differ by a factor of three on this line alone, which is the main reason quotes for apparently identical work vary so widely. What AI automation costs and the build cost page itemise it, and the receptionist cost breakdown is a worked example for one build in one trade.

Run it yourself

A comparison you can actually finish this week.

Reading about this is less useful than doing it once. Four steps, and none of them require talking to a salesperson.

  • Write the workflow down. Trigger, every step, the exceptions, where a human must approve. The workflow brief template is the shortest version of this and the audit template is the longer one. Everything downstream gets easier once this exists, and it is yours whoever you eventually hire.
  • Score the risk before the price. The risk scorer and risk checklist take fifteen minutes and will tell you whether this is a DIY-able workflow or one that needs a human approval step designed in by someone who has done it before.
  • Work out the payback. Hours saved times loaded cost against the build fee plus the running cost. The ROI calculator does the arithmetic and the payback period entry explains what number to aim at. If it does not clear roughly twelve months, the answer is usually not to buy.
  • Try the cheapest route for one evening. Follow a free build guide end to end. The guide library has complete builds, and the invoice chasing guide is a fair test of whether the DIY column is realistic for you. An evening spent this way is the most reliable way to price your own time, and the readiness quiz turns the result into a recommendation.

Then send the same brief to every bidder, including yourself as an option. Use the discovery call script to run the conversations consistently, and the rest of the template library for the paperwork. Comparing five quotes written against five different understandings of the job is not a comparison, and it is how most people end up choosing on price.

Failure modes

How each route actually goes wrong.

Not the marketing objections. The things that genuinely happen, and the specific thing to do about each.

RouteThe realistic failureWhat prevents it
DIYIt works for six weeks, then a silent failure goes unnoticed until a customer complains.Logging and an alert on failure, decided before the build. A workflow with no failure path is not finished.
FreelancerIt works, and nobody can change it. The builder is unreachable and there is no documentation.Handover contents written into the brief, not requested afterwards. Source, credentials, docs, walkthrough.
Solo consultantIt is built well and then the builder is unavailable at the moment you need a change.Documentation good enough for a stranger, plus an agreed response expectation. Judge it by asking what a different builder would need.
AgencyBudget goes into coordination, the build is done by their most junior person, and it lives on their platform.Ask who specifically is building it, and what happens to the workflow if you leave. Get both answers before signing.
In-houseOne person owns everything undocumented, and then resigns.The same handover standard you would demand from a supplier, enforced internally from the first workflow.
Every one of these is a process failure rather than a technology failure.

The common thread is that all five failures are about what exists after the build, not during it. That is why what a consultant actually does spends more space on handover than on building, and why the glossary and the agent archetype library exist: understanding what you are commissioning is the cheapest insurance available, and it costs nothing.

The hybrid

The mixture most businesses end up with anyway.

Framing this as one choice is a simplification that suits sellers more than buyers. What actually happens in a business that gets good at this is a mixture, arrived at over two or three years, and it is worth aiming at deliberately rather than stumbling into.

The pattern looks like this. Someone inside the business learns enough to build and maintain the simple workflows: the reminder that goes out on a schedule, the form that files itself, the report that assembles on a Monday morning. Nothing that touches money without a check, nothing that spans four systems. That person is usually an operations lead rather than a developer, and the material they need is method rather than programming.

Anything that touches revenue, customers or compliance gets built by someone who has done it before, with approval steps designed in and a tested failure path. It is handed over completely, so the internal person can change the message wording and the timing without calling anyone. That handover standard is the thing that keeps the mixture possible, and it is the single term worth being stubborn about in a contract.

An agency gets brought in only when the calendar forces parallel delivery, which for most businesses is once every few years rather than continuously. And an in-house hire arrives when the maintenance load has grown past what one operations lead can carry alongside their real job, which is a much clearer signal than any budget threshold.

The thing that breaks the mixture is not choosing wrong at any single step. It is accumulating automations nobody owns. Before adding the next one, be able to name the person responsible for the ones already running. If that name does not exist, fix that before buying anything else, from us or from anybody. The free automation audit is a reasonable place to have that conversation, and it produces a written roadmap you keep whether or not you hire anyone. If you are the person who builds this for other businesses, work with us is the relevant page, and about covers how this practice is set up and what it deliberately does not do.

FAQ

Consultant, agency, or DIY: FAQ

  • Is an AI consultant or an AI agency better for a small business?

    For a small business automating one or two workflows, a solo consultant usually puts more of the budget into building and gives you direct access to the person who does the work. An agency earns its higher price when you need several workstreams running in parallel, guaranteed cover if one person is unavailable, or a formal vendor relationship for procurement reasons.

  • What is the cheapest way to automate a small business process?

    Switching on a feature you already pay for. Most CRMs, booking tools, and accounting packages ship automation capabilities that nobody enabled. After that, DIY on a no-code platform costs only subscriptions and your time, which is genuinely the right answer if you enjoy the work and the process is simple.

  • When is DIY automation a mistake?

    When the workflow touches money or customers without a review step, when it spans three or more systems, when it needs to be reliable rather than merely working, or when the hours you would spend are worth more than the build fee. Half-built automations that nobody maintains are more expensive than never starting.

  • Should I hire a freelancer from a marketplace instead?

    It is the cheapest paid option and it can work well for a narrow, clearly specified task. The risks are handover and continuity: marketplace engagements often end without documentation, and finding the same person again months later is unreliable. Specify the handover contents in the brief, not after.

  • Does it make sense to hire someone in-house for automation?

    Usually only above a certain scale. A full-time hire is worth it when automation is continuous work rather than a project, when you have enough systems that someone needs to own them permanently, and when you can keep that person busy and technically challenged enough to stay.

  • How do I decide with a fixed budget?

    Under roughly $1,000, do it yourself or turn on features you already pay for. Around $2,000 to $5,000, one consultant-built workflow. Around $5,000 to $20,000, a consultant building a connected set. Above roughly $25,000 with several parallel workstreams, an agency starts to make sense. Above ongoing six figures of automation work, consider hiring in-house.

  • When should I not hire A8gent?

    If you need several workstreams delivered in parallel, guaranteed cover independent of one person, enterprise procurement and compliance paperwork, on-site presence, or work outside the small-business workflow types we do. In all of those cases an agency or a specialist firm is a better fit and we will say so on the call.

  • What should I compare suppliers on, if not price?

    Five things, and price is the one that decides fewest projects: cost, speed to live, depth of understanding of your process, continuity in eighteen months, and risk of a wrong output reaching a customer. Pick the two that matter for the specific workflow and the answer usually falls out. Continuity is the one people forget, so ask every bidder in writing what a different builder would need in order to take the workflow over. The completeness of that answer predicts your experience better than the quote does.

  • Can I mix approaches?

    Yes, and most sensible small businesses do. Learn enough to maintain simple workflows yourself, pay someone to build the ones that touch money or customers, and keep the option of an agency for anything that needs parallel delivery. Ownership of what you buy is what keeps the mixture possible.

Not sure which one you need?

Bring the question to the free consult. If the honest answer is an agency, in-house, or doing it yourself, that is what we will tell you.