The agency business model has not fundamentally changed in decades. You sell time, expertise, or a combination of both. You hire people to do the work. You try to charge more than it costs. The details vary, but the structure is the same.
AI is breaking that structure. Not overnight, but fast enough that agencies still running on the old playbook are already feeling the pressure. The models that worked when labour was the primary cost driver do not work the same way when AI can do 40-60% of the labour.
Here is what is thriving, what is dying, and where the smart money is going.
The traditional models, stress-tested
Hourly and time-and-materials billing
Verdict: dying.
This was already a weak model. AI accelerates its decline. When delivery time drops by 30-50%, your revenue drops with it. You can try to pad the hours, but clients are increasingly AI-literate and they know a 2-hour task when they see one.
The only scenario where T&M still works is highly specialist, unpredictable work where scope is genuinely impossible to define upfront (complex integrations, emerging technology implementations). For everything else, billing by the hour is a race to the bottom.
Project-based pricing
Verdict: strong, if you adapt.
Fixed-fee project pricing has always aligned incentives better than hourly billing: you commit to an outcome, not a time commitment. AI makes this model significantly more profitable because your delivery cost drops while your price stays anchored to the value of the outcome. The profit margin improvements are significant.
The risk is commoditisation. If clients can see that “a website” or “a brand identity” is being AI-produced, they will push for lower project fees. The defence is to price based on the strategic value you deliver, not the deliverable itself.
Retainer models
Verdict: needs restructuring.
The traditional hours-based retainer (30 hours per month at £X per hour) has the same problems as hourly billing. If AI means you can deliver the same value in 18 hours instead of 30, you are either overstaffing the retainer or having an awkward conversation with the client.
Value-based retainers survive and thrive. “We manage your entire content marketing function and target 40% organic growth” is a value proposition that does not erode when you get more efficient. In fact, your margin improves. See our detailed guide on restructuring retainers for the AI era.
The emerging models
These are the models that AI enables or dramatically improves.
Productised services
Verdict: thriving.
A productised service is a standardised offering delivered at a fixed price with a repeatable process. Think “SEO audit for £2,500” or “monthly content package: 8 posts, 4 newsletters, 1 report for £3,000.”
AI makes productised services explosively scalable. The standardised process lends itself perfectly to AI automation. You build the workflow once, refine it, and then deliver it to dozens or hundreds of clients with marginal additional effort per client.
The margin profile is extraordinary. A productised service that costs you £800 to deliver (including AI tools and human oversight) and sells for £3,000 gives you a 73% gross margin. Try getting that with a bespoke retainer.
Agencies that want to explore this should read our guide on building productised services with AI. It is, in our view, the single biggest opportunity for agencies in 2026.
Outcome-based pricing
Verdict: high-risk, high-reward.
Charging based on results (leads generated, revenue influenced, rankings achieved) has always been attractive in theory and terrifying in practice. AI shifts the risk profile.
With AI-driven analytics, attribution, and optimisation, you have more control over outcomes and better data to prove your impact. Agencies that are confident in their delivery and have strong AI systems can credibly offer outcome-based pricing because they can predict and influence results more reliably.
The model works best when:
- You control enough of the funnel to influence results meaningfully
- You have historical data to set realistic targets and pricing
- The client has realistic expectations about timescales and external factors
Do not offer outcome-based pricing on a capability you have not proven. It is a model for your strongest services, not your newest ones.
Revenue share and equity models
Verdict: selective opportunity.
Some agencies are moving to revenue share arrangements, taking a percentage of the revenue their work generates in exchange for reduced or zero upfront fees. AI makes this more viable because your delivery costs are lower, so you can absorb the deferred revenue.
This works for agencies that are selective about clients (you need to bet on businesses that will grow), have strong commercial acumen (you are effectively investing in the client), and are comfortable with variable income.
It is not a model for every client. It is a model for 2-3 clients per year where the upside is significant and the risk is manageable.
AI-as-a-service
Verdict: emerging and promising.
This is a genuinely new model. Instead of (or alongside) delivering traditional agency services, you build and operate AI systems for clients. Custom chatbots, internal knowledge bases, automated reporting dashboards, AI-powered content pipelines.
The beauty of this model is recurring revenue with declining marginal cost. Once you build the system, maintaining it takes a fraction of the time. The client pays monthly for the operational value it provides.
Agencies with technical capability and a willingness to productise their AI knowledge are finding this a lucrative complement to traditional services. It also positions you as indispensable; clients become dependent on systems you built and maintain.
Which model suits which agency?
The right model depends on your agency type, your clients, and your appetite for change.
Creative agencies (brand, design, campaign work): project-based pricing anchored to strategic value. AI speeds up production but the creative and strategic thinking justifies premium project fees.
Performance and digital agencies (SEO, PPC, content, data): productised services plus outcome-based components. The work is measurable, repeatable, and ideal for AI automation.
Consulting and strategy agencies: value-based retainers or project fees. The expertise is the product. AI enhances the depth and speed of your analysis but does not change the core value proposition.
Full-service agencies: a hybrid. Productised services for standardised work, value-based retainers for ongoing relationships, and project fees for bespoke work. The key is unbundling your offering so each component is priced appropriately.
How to transition
You do not need to overhaul your business model in one move. Here is the practical path.
- Start with new clients. Apply the new pricing model to every new engagement. Keep existing clients on their current model while you build confidence.
- Productise one service. Pick your most repeatable, most AI-automatable service and turn it into a productised offering. Test it with 5-10 clients before rolling it out broadly.
- Shift existing retainers at renewal. When retainers come up for renewal, propose the new model. Frame it as an upgrade: “We are moving to a value-based model that better aligns our incentives with your growth.”
- Track the right metrics. Revenue per head, gross margin per service line, and client lifetime value. These tell you whether your new model is working far better than utilisation rates or billable hours.
The agencies that act on this now are the ones building the AI-first model that will define the next era of the industry. The ones that wait will find themselves competing on price in a race they cannot win.
This is part of Margin Watch, a series on how AI is reshaping the business of running an agency. Subscribe to the newsletter to get new articles weekly.
