AI automation pricing: one-time fee or subscription?
By NoWorkr · Published
The most important decision in AI automation pricing is not how big the number is. It is whether you pay once for an automation you own, or pay every month for access to one you rent. With us, a scoped build starts from $1,500 as a one-time fee, and an ongoing partnership starts from $750/month.
The two models can look alike in the quote and sit very far apart after a year. Still, the choice is rarely written down anywhere, and most owners only notice it when they think about cancelling.
Here we set the two models side by side, work out when one overtakes the other, and use our own prices as the example instead of a range with no author behind it.
AI automation pricing: the two models side by side
The market uses many words for the same thing: fixed price, project price, sprint, retainer, partnership, licence. Behind them there are only two ways to pay.
| Price per automation | Subscription | |
|---|---|---|
| What you pay for | A piece of work that gets finished | Access and operation, for as long as you pay |
| When you pay | Once, typically split into two instalments | Every month, with no end date |
| Our level | From $1,500 | From $750/month |
| Who owns the result | You: source code, documentation, access | Usually the vendor |
| What happens if you stop | Everything keeps running | Everything stops |
| Best for | Stable workflows you already know | Unclear needs and ongoing development |
The bottom row decides the most. Everything else is arithmetic, and you can do the arithmetic yourself in a morning.
Price per automation: you pay for a piece of work
A price per automation covers one workflow from end to end. The quote, the invoice, the weekly report or the standard enquiry that arrives twenty times a day.
You pay for mapping, building, testing and handover, and then the task is closed. The day we finish, you could stop talking to us for good, and the chain would still run every morning.
The model has three advantages that matter to an owner.
The number is fixed. You can put it in the budget and calculate with it. A vendor who has understood the task can give you a fixed price. If they cannot, the task is not scoped yet, and the next step is a mapping session, not a quote.
The cost stops. A build is a one-off item. The saving keeps coming every month afterwards, and that is the whole reason automation can pay off at all.
You own it. Source code, documentation and access are yours. It is written into the agreement, not promised in a sales meeting, and you can read more about how we work.
The downside is just as real: you need to know what you want. If the workflow is not decided yet, you pay to automate a disagreement, and that gets expensive whatever the pricing model.
AI automation subscription: you pay for access
A subscription typically covers three things in one price: use of a platform, a certain amount of ongoing development, and someone watching that the chains keep running.
It is a good model in two situations. The first is when you do not know how many automations you will need. The second is when new ones keep appearing: a new webshop, a new field in the CRM, a new report for the board.
With us, that model is called a partnership, and it starts from $750/month. You get several chains built across the year instead of one agreement per task, and you get someone to fix them when a system changes underneath you.
So far there is nothing wrong with the model. The problem shows up in two places.
The price is the same when development stops. Most subscriptions cost the same in the month where three new chains are built as in the month where nothing happens. If you pay for development you do not get, you pay for nothing.
You rarely know what you own. If the platform and the account belong to the vendor, it is a rental. Ask directly: whose name is on the account, and who holds the source code? The answer moves the total bill more than any hourly rate.
The crossover point: when the subscription becomes the most expensive
Here is the calculation none of the big pricing guides make. Take the same task and set the two models side by side over time.
Put in our own numbers. A scoped build from $1,500 is paid once. A partnership from $750/month has cost two times $750 after two months, and that is where the two are level.
In month three the subscription is the more expensive one. By month twelve you have paid twelve times $750 for the same single task, and you still own nothing.
That calculation is only fair if the subscription covers exactly one workflow. If your vendor builds four new chains a year within the subscription, the picture looks completely different, and it is a reasonable purchase. The point is that you should do the calculation, not that the answer is always the same.
If you want to put numbers on your own task first, you can convert the hours into money in our ROI calculator.
What keeps running whichever model you pick
Some costs do not disappear just because you paid a one-time fee. They are small, but they belong in the quote.
Model usage. Every time an AI model reads or writes something, it costs a little. For a typical small business workflow, that is a few hundred dollars a month, and you only pay for what gets used. The prices are public, for example at OpenAI.
The platforms under the chain. An automation platform, a database or a mail service. Most clients already have them, but they need to be in the calculation.
Maintenance. Systems change. A field moves in the CRM, a vendor updates its API, and someone has to fix the chain. For some it happens once a year, for others every month.
The difference between the two models is who pays for the third item. In a subscription, maintenance is included. With a one-time fee, you agree it separately, and that is far cheaper than making the whole price recurring just to get it included.
What you are left with the day you stop
This is the question that separates the two models most, and it almost never gets asked while the agreement is new.
If you stop a subscription, the automation stops running. The account closes, and the work lands back on the desk it came from. You have not built anything up. You have rented a solution for a while.
If you end the collaboration after a build, nothing happens. The chains keep running because they sit on your own accounts, and the code and documentation are with you. That is the difference between an expense and an investment.
That is why we build fully custom and without templates, and why ownership is written into the contract. At Markeity they save 20,000 DKK every month on systems we built once, and you will find more examples in our client cases.
Ask every vendor the same thing before you sign: what is still running in a year if we stop paying you?
A grant fits one pricing model, not the other
For a Danish small business there is one more argument, and it is rarely mentioned in a price comparison.
Key numbers
250.000 kr
upper limit for a grant from SMV:Digital
$1,500
typical start for a scoped build, as a one-time fee
15 hours
of manual work a week we typically find at a client
SMV:Digital is the state scheme for digitalisation of small and medium-sized businesses, and it can cover up to 250.000 kr of a project. Requirements and application rounds are on SMV:Digital's own page and change from round to round.
The scheme is built for a scoped project with a budget and an end date. An ongoing subscription is hard to fit into that shape. If you are choosing between the two models and your company is in the target group, the grant clearly pulls toward the one-time fee.
We have collected the process in our guide to a grant for an AI project. The most important rule is the order: you have to apply before the work starts.
How to choose a pricing model for your own task
You can settle it with four questions. Take them in order.
- Is the workflow decided? If you can write the rule in one sentence, the task is ready for a fixed price. If you cannot, buy advice or a short engagement first, not a build.
- Will more tasks follow this one? If the answer is one yes and then silence, that points to a one-time fee. If the answer is a list of ten, a partnership makes sense.
- How long does the chain need to run? If you are switching ERP next year, wait or rent. If the workflow has to last for years, it is worth owning.
- What happens if the vendor disappears? If you cannot answer, you are renting, whatever the agreement is called.
We use the same four questions ourselves before we send a quote, and they do not always point to the most expensive option. We have built 53 automations for Danish companies, and the pattern is almost always the same: the first chain is the one that needs the most decisions, and the third is the one you can really see in the accounts.
From mapping to launch, the first build typically takes about 3 weeks, and we work toward measurable ROI within 90 days. If it is not there, we keep going until it is. You can follow the process step by step in our process, and see where it typically pays off most in our guide to AI automation for small businesses.
FAQ
What does an AI agent cost?
An AI agent that solves one scoped task is priced as a build, and with us that starts from $1,500 as a one-time fee. If the agent needs access to several systems and has to handle many exceptions, the price goes up, because there are more links to build and test.
What does a chatbot cost?
A bot that only answers from the content on your own website can often be handled with a ready-made tool for a few hundred dollars a month. If it has to look things up in your order system and answer questions about specific cases, it is a build, and the price of a scoped automation applies.
Is a subscription always more expensive than paying per automation?
No. On one stable workflow, the subscription becomes the most expensive after a few months. But if you get new chains built continuously within the same monthly price, a subscription can easily be the cheapest route. Work out how many automations you realistically get built per year.
Can I start small without tying myself down?
Yes. Start with one task at a fixed price, and look at the result before you decide anything bigger. It is also the order we recommend: one scoped build, measured effect, and then a decision on whether you need more at all.
What is AI automation in practice?
It is when a piece of software carries out a workflow that an employee would otherwise do by hand, and makes decisions along the way based on a rule you have set. It reads an email, pulls data from two systems, creates a record and sends a message, without anyone having to click.
Start by working out the crossover point on the one task that annoys you most. That number decides which pricing model is right for you.