Most agencies are using AI internally. Fewer are selling AI as a service. The ones that are selling it are adding £10,000 to £30,000 per month in new revenue without hiring a single person.
The difference between using AI inside your agency and packaging it as a client-facing service line is not technical. It is commercial. You already have the skills. You just need the structure.
Internal use vs. external service: the gap
Using AI to speed up your own workflows is operational efficiency. Selling AI services to clients is a revenue stream. They require different thinking.
When you use AI internally, the bar is “does this save us time?” When you sell it externally, the bar is “does this deliver a measurable outcome the client will pay for?” That distinction changes everything, from how you scope the work to how you price it.
The agencies making real money from AI services are not selling “AI” as an abstract capability. They are selling specific, packaged outcomes that happen to use AI in the delivery.
Three service categories worth building
1. AI consulting and audits (£2,000 to £8,000 per project)
Walk a client through their business, identify where AI can reduce costs or improve output, and deliver a prioritised implementation plan. This is the easiest entry point because you are selling your expertise, not a deliverable.
What it looks like: a two-week engagement. One discovery workshop. One week of analysis. A 20-page report with specific tool recommendations, workflow redesigns, and an estimated ROI for each recommendation.
Why it works: clients are drowning in AI noise. They have seen 50 LinkedIn posts about ChatGPT and have no idea what applies to their business. Your audit cuts through that. You have already done the experimentation. They are paying you to skip the learning curve. For guidance on positioning this to existing clients, see how to sell AI services to agency clients.
2. AI implementation and training (£5,000 to £15,000 per project)
Help clients set up AI tools, build custom workflows, create prompt libraries, and train their teams. This is the highest-margin work because delivery cost is low once you have built the playbook.
Typical scope: tool selection and setup, three to five workflow automations, a prompt library tailored to their industry, two training sessions for their team, 30 days of post-implementation support.
Reusability is the key margin driver here. Your first client implementation takes 60 hours. Your fifth takes 20. The playbooks, templates, and training materials transfer across clients with minimal customisation.
3. AI-powered managed services (£1,500 to £5,000 per month, recurring)
Ongoing services where AI is embedded in the delivery: enhanced reporting, content production, research packages, or monitoring services. This is where the recurring revenue sits.
For example, an AI-powered competitive intelligence service that monitors a client’s competitors, analyses their activity, and delivers a monthly strategic briefing. Your cost to deliver: two to three hours of analyst time plus AI tools. Client value: strategic insight that would previously have required a dedicated hire.
Pricing models that protect your margin
Project-based for consulting and implementation. Fixed scope, fixed price. Do not sell hours. The value of an AI audit does not change because you got faster at delivering it. If anything, your speed is part of the value proposition. This aligns with the broader shift toward value-based pricing in AI-augmented agencies.
Retainer for managed services. Monthly fee for ongoing delivery. Structure it so AI handles 70% of the production and your team handles the 30% that requires judgement and strategy. Your effective hourly rate should be significantly higher than your traditional service lines.
Do not itemise “AI” as a line item. Clients should not see a cost for “AI tools” or “AI processing.” They are buying outcomes. The technology is your business, not theirs.
A well-structured AI services division should operate at 60 to 70% gross margins, compared to the 40 to 50% typical in traditional agency services. The difference is leverage: your delivery cost does not scale linearly with output.
Which team members to invest in
You do not need to hire AI specialists. You need to upskill the people you already have.
Operations people make the best AI service leads. They understand processes, they think in systems, and they are comfortable with tools. A senior project manager or operations director with six months of AI training will outperform a newly hired “AI consultant” who does not understand agency work.
Strategists make the best AI consultants. They already know how to diagnose business problems and recommend solutions. Add AI literacy and they can run audits and workshops immediately.
Producers and coordinators are your implementation team. They are used to configuring tools, documenting processes, and training people. AI implementation is a natural extension.
The investment is not enormous. Budget for tool subscriptions (£200 to £500 per person per month), dedicate one day per week to closing the AI skills gap for the first three months, and start with a single pilot client.
Revenue potential: realistic numbers
Based on agencies we work with, here is what a focused AI services push looks like in year one.
- Months 1 to 3: Two to three AI audits at £3,000 to £5,000 each. Total: £6,000 to £15,000.
- Months 4 to 6: Two implementation projects from audit clients at £8,000 to £12,000 each. One new managed service retainer at £2,500 per month. Total: £23,500 to £31,500.
- Months 7 to 12: Three to four managed service retainers running at £2,000 to £4,000 per month each. Plus ongoing project work. Total: £48,000 to £100,000.
Year one total: £77,500 to £146,500 in new revenue. That is conservative. It assumes a small team, modest pricing, and a slow ramp. Agencies that move faster and price on value regularly exceed these numbers.
Structuring the case study pipeline
Every AI project you deliver is a case study in the making. Build this into your process from day one.
Before starting any engagement, agree metrics with the client: time saved, cost reduced, output increased, revenue generated. Measure the baseline. Measure the outcome. Get a testimonial.
Three strong AI case studies are enough to build a credible service page and a compelling pitch deck. Five puts you ahead of 95% of agencies who are still talking about AI theoretically.
The format that works best: problem, approach, result, client quote. Keep each one under 300 words. Specificity sells. “Reduced report production time from 12 hours to 3 hours” is worth more than “implemented AI across their marketing operations.”
The bottom line
Selling AI services is not a pivot. It is an expansion. Your existing skills, relationships, and delivery capabilities are the foundation. AI is the multiplier.
The agencies that wait for the market to mature before offering AI services will find that the market matured without them. The ones building now, even imperfectly, are locking in client relationships and case studies that compound over time.
Start with an audit for your best client. Price it at £3,000. Deliver it in two weeks. Use the results to scope an implementation project. Use the implementation to build a managed service retainer.
That is not a strategy deck. That is a 90-day plan.
This is part of The Pitch Stack, a series on winning new business with AI. Subscribe to the newsletter to get new articles weekly.
