A client rang me in January. Good bloke, runs an e-commerce brand in Manchester, we'd built his WooCommerce site back in 2022. He wanted a full redesign. Same scope as last time: custom theme, product filtering, checkout optimisation, the works. When I quoted him £6,800 he went quiet for a moment, then said, "Isn't AI supposed to make this cheaper now?"
Honestly? That question keeps coming up. And it deserves a real answer, not a PR response about "value" and "expertise."
So here's what I've actually observed, building sites at Seahawk Media, watching our team use Cursor, GitHub Copilot, Midjourney, and a dozen other tools daily. The honest answer is: AI changed agency pricing, but not the way most people assume.
---
The Hours Shrank. The Rates Didn't (Straightforwardly) Drop.
When Copilot first became part of our actual workflow, around mid-2023, I tracked it obsessively for about three months. Custom WordPress plugin work that used to take a developer 6 hours was landing closer to 2.5. Repetitive CSS theming? Cut by maybe 40%. First-draft copywriting for landing pages, which we sometimes help clients rough out, got faster by an embarrassing margin.
So you'd think: fewer hours billed, lower invoice, happy client.
It didn't work like that.
Why rates held (or went up)
The tasks that got faster were already the cheap tasks. The expensive part of any project was never the code typing. It was the back-and-forth, the brief interpretation, the "we thought we wanted X but actually we want Y" client archaeology. That part takes exactly as long as it always did.
What AI actually did was compress the mechanical execution layer. And when your mechanical execution gets faster, one of two things happens: you lower prices, or you take on more projects. Most agencies, including ours, did the second thing. I'm not apologising for it.
The Freelancers Union's 2024 survey found that the majority of independent contractors who adopted AI tools reported no reduction in their rates. A meaningful chunk raised rates, citing the higher quality of output they could now deliver. That tracks with what I've seen.
---
Where Rates Actually Collapsed
Let's be honest. Some work got commoditised fast. Genuinely fast.
Content writing is the obvious one. I know agencies that charged £150 per blog post in 2022. That market is largely gone for standard-tier content. Clients with any technical awareness know they can run a decent first draft through ChatGPT or Claude themselves, and if they're paying an agency £150 they expect something that reads like it came from a person who thought hard about the subject.
Logo design for small businesses. Basic landing pages with no custom functionality. Simple explainer copy. These categories took a real hit. If you were running a one-person shop in 2021 charging £80/hour for this kind of work, the past two years have been uncomfortable.
The spec work problem got worse
Here's a thing I didn't predict: AI made spec work more common, not less. Clients started asking for full mockups before signing. Because they'd seen on Twitter that someone built a prototype in Figma using AI in 20 minutes, they assumed the same effort applied to production-grade work. It doesn't. But the perception shift is real, and it's put downward pressure on smaller agencies who haven't figured out how to push back.
---
Niches Where Rates Went Up
This is the part nobody writes about.
Technical architecture work. If you know how to prompt AI tools intelligently, review what they output, catch the subtle bugs, and integrate everything into a production environment without it falling over, you are worth more money than you were in 2020. Not less. The ceiling on what a senior developer can actually produce in a day went up, and smart clients are willing to pay for that ceiling.
At Seahawk, we built a headless WordPress setup for a fintech client last year. The front-end scaffolding that would have taken our lead dev two weeks got stubbed out in three days using Cursor with a solid system prompt. But the actual decisions, the API security architecture, the edge caching strategy, the review of what Cursor got wrong (and it did get things wrong, repeatedly), that still required a senior engineer thinking hard. We billed the project on value, not hours, and the client paid more than they would have in 2022 for the same outcome.
Complex UX consulting rates also went up, from what I can see. The thinking work, the user research synthesis, the decisions about what to build at all, that's where humans still have a real edge and clients are recognising it.
---
The Hourly Model Itself Is Breaking
I want to be direct about this because it affects every freelancer and small agency reading this.
Charging by the hour was always a bit of a strange construct. You're essentially penalising yourself for getting better at your job. Ten years ago it was tolerable because efficiency gains were incremental. AI made them non-incremental.
If I told a client, "This took me 45 minutes because I used Copilot," and charged accordingly, I'd be billing £90 for work that used to be a £400 line item. The work has the same commercial value to the client. I just delivered it faster. Why should I get paid less?
The shift toward fixed-price, value-based, and retainer models has genuinely accelerated. I started moving Seahawk's new client work to mostly fixed-price proposals back in 2023, and the conversations got easier almost immediately. Clients care about the outcome, not the timecode.
The Nielsen Norman Group has written about this from a UX consulting angle, and the same logic applies across the board: charge for the decision, not the hour.
---
What "An Agency Hour" Even Means Now
Here's a thing worth pausing on. When you pay an agency £120/hour in 2026, what are you paying for?
You're not just paying for a human sitting at a keyboard. You're paying for:
- The AI tool subscriptions (Copilot, Cursor, Midjourney, ChatGPT Pro, Perplexity, Claude, the list keeps going)
- The prompt engineering skill to get non-garbage output
- The review layer: a senior person checking what the AI produced
- The error correction when the AI confidently does something wrong
- The contextual knowledge about your industry, your existing codebase, your past decisions
Our monthly AI tooling bill at Seahawk runs to roughly £800-£1,000 across the team. That's a real cost. It doesn't appear as a line item on client invoices but it absolutely factors into what we charge.
---
How to Actually Price Your Work in 2026
I've made enough mistakes here to have a point of view.
- Stop anchoring to hours for fixed-scope work. Price the deliverable. "Landing page with A/B testing setup: £1,400" tells a client exactly what they're buying. "12 hours at £95/hour" invites negotiation and resentment.
- Separate strategy from execution in your proposals. Strategy (what to build, why, for whom) is still labour-intensive and human-dependent. Execution got faster. Price them differently.
- Be transparent about AI in your process if asked, but don't volunteer it as a discount reason. If a client specifically asks whether you use AI tools, say yes, and explain that those tools let you deliver faster without sacrificing quality. That's not a reason to lower your rate. It's a reason to hit deadlines.
- Raise your floor on new clients. If you haven't updated your rates since 2022, you're probably undercharging. The Bureau of Labor Statistics data on IT wages shows consistent upward movement even through the AI transition period, because the skill required to work effectively with AI tools is not zero.
- Track the new profitability, not just the revenue. If AI is saving you 30% of execution time on a fixed-price project, that's margin, not a signal to drop your quote.
---
The Clients Who Will Push Back (And How to Handle Them)
Back to that Manchester client. I talked him through it. Told him what the £6,800 was actually covering: the full discovery session, the custom Figma prototypes, the WooCommerce build with a bespoke product filter we'd write ourselves, performance optimisation to get his Core Web Vitals above 90, and a month of post-launch support.
He asked again about AI. I told him: yes, parts of this will move faster because of the tools we use, and that means fewer revision cycles and a tighter timeline. He's paying for the outcome, and the outcome is a site that converts. He signed off.
The clients worth working with understand that faster delivery is a benefit, not an excuse for a discount. The ones who don't? They'll go find someone on Fiverr and be back in 18 months with a mess to clean up. I've seen that cycle enough times.
---
FAQ
Has AI caused agency hourly rates to drop across the board?
No, not across the board. Rates for commodity work (basic content, simple static sites, boilerplate graphic design) have softened significantly. But rates for technical work, strategy, and senior-level consulting have held or increased. The average blended rate at mid-size agencies in the UK is higher in 2026 than it was in 2021, partly because simpler work has been filtered out of the agency model altogether.
Should I disclose to clients that I use AI tools?
I think you should be honest if asked directly. Beyond that, it's a business decision. What you shouldn't do is pre-emptively use AI tool usage as a justification for a lower rate. The tools are part of your professional infrastructure, like your laptop or your Figma subscription.
Is the hourly billing model dead?
Not dead, but declining for project work. Retainers and fixed-price engagements are becoming more common, and that's probably good for both agencies and clients. Hourly billing still makes sense for ongoing support, emergency work, or open-ended consulting where scope genuinely can't be fixed in advance.
How much should I factor AI tool costs into my pricing?
Practically speaking: tally your actual monthly AI subscriptions, divide by your billable hours, and make sure that number is buried somewhere in your effective rate. For most solo operators it's somewhere between £8 and £20 per billable hour. Not enormous, but real.
What types of agency work are most at risk from AI pricing pressure?
Junior copywriting, basic graphic design, templated SEO content, and simple landing-page builds. If your core offer sits in any of these areas, the pressure is real and it's not going away. The move is either up-market (more strategic, more technical) or toward volume plays with highly efficient AI-assisted processes. The middle is getting squeezed.
---
The rate conversation isn't really about AI. It's about whether you can articulate why your work is worth what you charge. AI just made that question harder to dodge. Answer it clearly, and the pricing sorts itself out.