👋 Greetings {{first_name|earthlings}},

Emma here, co-founder of Spark AI. Jules and I are packing for Cannes soon – the Lions festival kicks off later this month, and if you’re heading out too I’d love to see you (grab a free AI audit using the link at the end). But before we go, I want to pick up a thread we started a few editions back: if AI makes the work faster, should that change what you charge for it?

A few weeks ago I wrote about Monks moving off the billable hour (edition #44). Since then we’ve watched the same shift show up again and again – in our own research, in the agencies we work with, and most recently on Spark Sessions, where we sat down with a team who turned a production problem into a licensed AI product. So this edition is about the tricky subject of pricing: why the hourly model can now work against you, the routes agencies are taking instead, and a way to think about which one fits you.

By the way, we are taking a pause for the next issue of Spark Intelligence for team holidays and getting stuck in to events at Cannes - I’ll be back with a full Cannes download here in the newsletter and on our podcast in a few weeks.

There’s a trap that’s easy to walk into. You invest in AI, you rebuild a workflow, and you free up a chunk of time. Then your client says: hang on, if it took you half as long, shouldn’t I pay half as much? Clients are expecting you to invest in AI, but then try to hammer you on price when you do. It isn’t fair – but if you’re still pricing by the hour, you’ve built the trap yourself.

And this is no longer hypothetical. In the latest Spark Report, AI use in content and creative generation jumped from 20% to nearly 50% in six months, and research, strategy and analysis climbed from 25% to more than a third. AI is firmly in the creative work itself, and the thinking around it. At the same time, almost 90% of agency staff say they are saving time using AI. When you charge for time, every efficiency you find becomes a discount you hand to the client.

The fix is to price for what the work is worth, not how long it took. Jules frames it as a ladder of options to weigh up:

A word of realism before you redraw every contract: none of these is a switch you flip overnight. Existing clients are used to buying your time, and asking them to change how they pay mid-relationship is a hard conversation – often harder than it’s worth. The easier route is to treat these as models to test rather than a wholesale move. Start a new client, or a new project, on new terms, and let the proof build from there. The ladder is something to experiment your way up, not a leap you have to make in one go.

And when a client wants you to innovate with AI, remember what you’re really being asked: to do something new, that nobody has done, without any track record to lean on. Everybody wants innovation, but they want it without any risk – and you can’t have both. You’re carrying that risk and investing to build the capability. That isn’t a cost to swallow quietly. It’s part of what you’re charging for.

This is also where the client conversation has to change. Value pricing only works if the client can see the value – and the polished concept they sign off is only the tip of the iceberg. Beneath it sits everything that got you there: the strategic directions explored, the assumptions challenged, the alternative routes considered and quietly ruled out. AI lets you do more of that thinking, and do it faster – so show it. “We researched 200 websites, analysed 100 reddit threads, explored eight strategic directions, pressure-tested all of them with a synthetic audience built from focus group interviews, and landed on the one with the strongest commercial case. Here’s why.” That’s what they’re paying for.

None of this means abandoning the craft. A great example we’ve seen is Conturae, a 20-person copywriting agency who pride themselves on every word being written by a human. They’ve automated everything around the writing instead – a bot scans incoming briefs against their template and chases anything missing, and a research agent assembles what the writer needs before the brief gets passed to the copywriter. Then once they’re done, a final check runs the draft against the client’s tone of voice. The copywriter still makes every call, and lands 95% client approval first time. They protected what makes them valuable and used AI to streamline everything that didn’t.

POD LDN and their sister agency Umbrella take exactly the same approach in promo production – the trailers and sizzle reels for upcoming shows and films, where the same video gets recut over and over for different channels, slots and audiences. Their tool, PromoTool, analyses a broadcaster’s past promos to draft creative briefs, sorts the footage into clips and assembles a first cut on the timeline. The editor then does the bit only they can do – swapping clips, reworking transitions, shaping the story – and once it’s right, PromoTool handles the versioning, compliance and language variants. A sizzle reel that used to take three or four weeks now comes together in a day or two, and the editor still makes every creative call. The tool does the parts that were never really the job; the editor does the part that is. As Jules likes to say (a big racing fan in his youth), a Formula 1 driver doesn’t become less valuable in a faster car – they become more valuable, because they can perform at a higher level.

We dug into all of this on the latest Spark Sessions, where POD LDN and Umbrella walked us through how they took PromoTool to market. Rather than using it to make promos faster for clients, they license it to broadcasters and streamers to make their own – they own the IP and earn from the same product many times over. Watch the recording here.

Being clear on the role you want AI to play in your business helps define your pricing model, because helps you be clear about exactly what you are pricing. Through our work with over 70 agencies, we distilled four archetypes that categorise agency goals with AI. You’ll recognise bits of yourself in several, but one will be dominant – and it points to how you should charge.

There’s no wrong archetype, and you may be a blend. But don’t attempt to be all of them at once, which spreads your AI investment so thin it shifts nothing. Concentrate on the one that makes you distinctive.

If you commission agencies rather than run one, this shift matters to you too of course. As pricing moves from hours to outcomes, what you’re buying changes – you’re paying for the thinking, the systems and the results, not a timesheet. The agencies you want to work with are those investing to make your budget work harder and your campaigns land faster. When you next review a roster, ask how each agency is using AI to raise the standard of the work. The answer tells you who’s building for your future, and who’s simply defending their margin.

The pace doesn’t let up! Here’s what you need to know this fortnight.

OpenAI has started rolling out ads inside ChatGPT across the UK, Japan, Brazil, Mexico and South Korea, on top of the US. The barrier to entry has dropped fast – early on only big brands could play, with minimum spend around $250,000, but in the US there’s now no minimum at all, so smaller businesses can test it too. Two things to keep in mind: ads only show on the free and cheaper plans (Plus, Pro and Enterprise stay ad-free), and there’s a live question about bias, since the same company runs both the answers and the ads. OpenAI says ads are labelled and kept separate, and it’s working on independent measurement. What this means: as we always advise, keep your team on a a paid plan designed for businesses – for the privacy of your work and your clients’, and to stay clear of the ad layer entirely. OpenAI’s announcement.

Anthropic shipped another new model on 28 May. It’s a step up rather than a leap, and the price hasn’t changed. You can now set the effort level – low, medium, high, extra or max – so you can dial it up for tricky work or keep it light and fast for quick jobs. We’re pretty exciting Extra is the sweet spot for knowledge work. And it’s more honest about the limits of what it knows: rather than guessing confidently, it’s more likely to flag when it’s unsure. Anthropic’s update.

It’s a busy early summer for team Spark AI. Here’s where you can catch us – the full line-up and sign-ups are on our events page.

Heading to Cannes? Grab a free audit slot here. Can’t make it? We’re also running free AI audits on a call for when we’re back – same thing, a clear read on where you actually sit and what to do next. Just hit reply.

It’s a busy summer for clients on our AI Accelerator too, with brilliant teams including Pearlfisher, Avantgarde, Everland and ACC in the current cohort. If you’d like to find out more book a call to talk it through.

That’s all for this edition – I hope it’s sparked some thinking on where AI fits in your pricing. See you next time,

Emma

Co-founder, Spark AI

Spark AI helps you lead your team through the biggest shift since digital, with AI training, transformation and tools. We’ve worked with 70+ agencies, published the #1 bestselling book on AI for agencies, and teach at Oxford University.

👉 Find out more