Pricing AI automation work is one of the least discussed and most consequential decisions you'll make as an agency. Here's what works, and what to avoid.
The one-off project trap
The most common mistake new AI automation agencies make is pricing everything as a one-off project. A client asks for an automation, you scope it, you quote a fixed fee, you build it, you deliver it.
The problem isn't the model — it's that it doesn't reflect the ongoing value you're delivering. AI agents require maintenance: prompts drift, APIs change, data sources evolve, and the client's requirements change as they see what's possible. A one-off fee doesn't compensate for any of that.
Worse, once the project is delivered, you have no commercial relationship. Getting more work means starting a new sales cycle from scratch. You're trading time for money with no leverage.
The agencies building real businesses on AI automation have moved to recurring models. The project fee still exists — but it's the entry point, not the product.
The recurring model that works
The most sustainable pricing structure for AI automation agencies has two components:
Setup fee
A one-time fee that covers scoping, building, and initial deployment. This should reflect the actual time and complexity of the build — not be artificially low to win the deal.
Typical ranges vary significantly by complexity:
| Project complexity | Setup fee range | |---|---| | Single agent, defined workflow | £1,500 - £5,000 | | Multi-agent system, 3-5 workflows | £5,000 - £15,000 | | Enterprise deployment, custom integrations | £15,000+ |
These are indicative — your market positioning and the client's perceived value should inform your actual numbers.
Monthly retainer
The ongoing fee that covers monitoring, maintenance, updates, and client access to the portal. This is where the business model compounds.
The key to justifying the retainer is having something to show for it every month. A client portal with task history, monitoring dashboards, and regular output gives clients a visible reason to keep paying.
Monthly retainer components to price into:
- Platform access — client portal, user logins, agent availability
- Monitoring — reviewing agent performance, catching failures before the client does
- Maintenance — updating prompts and workflows as requirements evolve
- Support — responding to bug reports, answering questions, making adjustments
- Reporting — monthly or quarterly review of agent performance and ROI
A reasonable baseline for a small deployment (3-5 agents, one client organisation) is £500-£1,500/month. Scale with the number of agents, users, and the complexity of the ongoing maintenance load.
Value-based pricing
The strongest argument for higher fees isn't your hourly rate — it's the value you're replacing. If your automation saves a client 20 hours of manual work per week at a fully loaded cost of £50/hour, you're delivering £4,000/month of value. A £1,200/month retainer is a very easy decision for them.
The challenge is making the value visible. This is where task history and monitoring earn their keep beyond operational usefulness. When you can show a client exactly how many tasks ran last month, how long each would have taken manually, and what the combined time saving was — the retainer conversation becomes a maths problem, not a negotiation.
Build the habit of tracking this from day one of an engagement:
Monthly value report format:
- Total tasks completed: X
- Estimated manual time per task: Y minutes
- Total hours saved: X x Y / 60
- Value at £[client hourly rate]: £Z
- Retainer cost: £[retainer]
- ROI this month: (Z / retainer) x 100%
Most clients, once they see this clearly, stop questioning the retainer.
Pricing tiers for agencies managing multiple clients
As you add client organisations, you can introduce tiering to reflect different service levels:
| Tier | What's included | Target client | |---|---|---| | Starter | 1-2 agents, up to 5 users, email support | Small business, single use case | | Growth | Up to 5 agents, up to 20 users, monitoring, monthly review | Growing business, multiple workflows | | Scale | Unlimited agents, unlimited users, SLA, dedicated support | Enterprise or high-volume client |
Tiers make it easy for clients to self-select and for you to upsell as their usage grows.
What not to price on
Per-task pricing creates anxiety. If clients feel like every task costs them something, they'll use the system less — which is exactly the opposite of what you want.
Per-API-call pricing is even worse — it introduces cost unpredictability and makes clients distrust AI agents that might "run up a bill."
Hourly rates for maintenance undermine the retainer model. If every small change requires a separate invoice, the client relationship becomes transactional.
The goal is a model where the client pays a predictable monthly fee and uses the system as much as they want. High usage is a good thing — it means they're getting value, which means they'll keep paying.
The expansion conversation
The best source of new revenue is your existing clients expanding their use of agents. After the first month or two, when you're in your regular review, ask:
"What else does your team do manually that takes up time and follows a consistent process?"
Every answer is a potential new agent. Each new agent justifies either an increase in the retainer or an additional setup fee — usually both.
Clients who started with one reporting agent and ended up with six agents covering sales, operations, marketing, and support are common in this space. The compounding nature of automation means that once a client trusts the system, expansion is the natural next step.
Agentic Vessel gives you the portal, monitoring, and task history you need to justify and demonstrate the value of your retainer every month. Start for free.
Build your AI agent workflows today.
Join developers already automating complex tasks with Agentic Vessel.
Register as a Developer