Guide · 9 min read
How AI automation agencies price their work
Most AI automation agencies price in stages: a scoped assessment, a fixed-scope build for an agreed workflow and a monthly retainer to run, support and improve it. Software and AI usage costs usually sit on top. What moves the price is the number of systems, the state of the data, security needs and how complex the workflow is.
Key takeaways
- Expect pricing in stages: assessment, build, then run and support.
- Running costs such as software subscriptions and AI usage sit outside the agency fee, so ask for them explicitly.
- The biggest cost drivers are the number of systems, data quality, security requirements and workflow complexity.
- Be wary of a fixed price given before anyone has seen your systems and data.
- Agree what counts as a change, and how changes are priced, before the build starts.
What's in this guide
Which pricing models do AI automation agencies use?
Agencies usually combine a scoped assessment, a fixed-scope build, a monthly run and support retainer, and sometimes day rates for work that is hard to scope.
| Model | What it pays for | When it fits | What to check |
|---|---|---|---|
| Assessment or discovery | Mapping the workflow, systems, data and risks, and agreeing a first scope | Before any build, especially when several systems are involved | What you receive at the end, and whether you keep it |
| Fixed-scope build | Designing, building, testing and launching an agreed workflow | Once the workflow and systems are understood | What is in scope, what counts as a change and the acceptance criteria |
| Run and support retainer | Monitoring, fixes, small improvements and user support after launch | Any workflow the business depends on | Response times, what is included and the notice period |
| Day rate or time and materials | Work that is hard to scope in advance | Exploratory or changing requirements | Estimates, caps and reporting |
| Outcome-linked element | Part of the fee tied to an agreed result | When the result can be measured cleanly | How the result is measured and attributed |
Most engagements mix these. In our view, pricing based purely on outcomes rarely suits automation work, because results also depend on things the agency does not control, such as data quality and how the team adopts the workflow.
What makes an AI automation project cost more?
Cost rises with the number of systems, poor data quality, security and compliance requirements, workflow complexity, approval steps and the level of support needed.
- Number of systems. Each connection adds mapping, permissions, testing and monitoring.
- Data quality. Duplicates, missing fields and inconsistent stages add clean-up work before automation can be trusted.
- Security and compliance. Access reviews, data processing agreements, hosting requirements and audit logging all take time.
- Workflow complexity. Exceptions, branching logic and human approval steps add design and testing effort.
- API limits and older systems. Some systems need workarounds or custom code.
- Adoption. Training, documentation and changes to how people work.
- Support level. Response times and hours of cover after launch.
Which costs sit outside the agency fee?
Software subscriptions, middleware usage, AI model usage, hosting and any new licences usually sit outside the agency fee, so ask for an estimate of each.
The agency fee is only part of the total. Ask who pays for each of these, and whose name the account is in:
- Automation platform or middleware subscriptions and usage.
- AI model usage, which often grows with volume.
- Hosting for any custom services.
- Additional CRM or tool licences the workflow needs.
- Monitoring and logging tools.
AI usage costs can grow quickly once several workflows run on their own schedules. Our guide to controlling AI agent costs covers how to track and cap them, and what a self-running content engine costs to operate shows the running-cost side in practice.
What does an example engagement structure look like?
A common structure moves from assessment to a fixed-scope first workflow, then launch and training, then a monthly retainer to run and improve it.
| Stage | Pricing shape | What you should receive |
|---|---|---|
| Assessment | Fixed fee or scoped discovery | A mapped workflow, a systems and data review, risks, a baseline and a proposed first scope |
| First workflow build | Fixed scope and price | A tested, launched workflow with agreed permissions and approval steps |
| Launch and training | Included in the build or priced separately | Trained users, documentation and a named owner |
| Run and improve | Monthly retainer | Monitoring, fixes, small improvements and a regular review |
| Next workflows | A new fixed scope each time | Each expansion scoped on what the first workflow proved |
This structure keeps the commitment small until the first workflow has proved itself, and gives you natural points to stop, change direction or take ownership in-house. It is illustrative: any agency will adapt it to the systems and risks involved.
What should you ask before you sign?
Ask what is in scope, what counts as a change, which running costs to expect, what support includes, who owns the work and how you can exit.
- What exactly is in scope, and what would count as a change?
- How are changes estimated and approved?
- Which running costs should we budget for, and whose name are the accounts in?
- What does support include, and what are the response times?
- Who owns the workflow logic, code and documentation?
- What notice period applies to the retainer, and what happens at handover?
Our buyer's checklist for choosing an AI automation agency covers the wider questions, and agency vs in-house compares the cost of building the capability yourself.
How does The Sparked Group price its work?
We agree scope and timing after an assessment, and make ownership, hosting, third-party software and AI usage costs clear before any build starts.
Every engagement starts by understanding the workflow, the systems and the data, so that the scope we propose is one we can stand behind. Ownership, monitoring, support and ongoing software and AI usage costs are agreed before work starts. We do not publish standard prices; we quote after assessment, once the systems and risks are understood.
To discuss a specific workflow, find your starting point or read how we approach AI automation.
Frequently asked questions
How do different AI automation agencies price their services?
Most combine a scoped assessment, a fixed-scope build for an agreed workflow and a monthly retainer to run, support and improve it. Some use day rates for work that is hard to scope, or tie part of the fee to an agreed outcome. Software and AI usage costs usually sit outside the agency fee.
Why does AI automation pricing vary so much?
Price depends on the number of systems involved, the quality of the data, security and compliance requirements, the complexity of the workflow and its exceptions, the approval steps needed and the level of support after launch.
Should we accept a fixed price before an assessment?
Be cautious. Without seeing your systems, data and exceptions, an agency either adds contingency to the price or discovers extra work later and raises change requests. A short assessment first usually produces a more reliable scope and price.
What ongoing costs come with AI automation?
Expect subscriptions for automation platforms or middleware, AI model usage that grows with volume, hosting for custom services, any extra tool licences and a support retainer if the workflow is business-critical.
Cited and further reading
- A practical guide to controlling AI agent costs · The Sparked Group
- What a self-running content engine actually costs to operate · The Sparked Group
- How to choose an AI automation agency: a UK buyer's checklist · The Sparked Group
Want help choosing your first workflow?
Bring one process, one bottleneck or one idea to a first call. We will help you find a practical starting point.
Select a time →