👋 Greetings {{first_name|earthlings}},

Emma here, co-founder of Spark AI. Nearly every agency I speak to has plenty of AI activity going on, and a hard job pointing to where it's paid off. They can see the time it saves and the new things it makes possible, but client work comes first, and the internal work that would turn those saved hours into something the business can feel keeps sliding down the list.

Last week I was on a call with the MD of a mid-sized independent. Everyone has a ChatGPT licence, and a handful of people are really building with AI. When I asked what had changed commercially, there wasn't an easy answer.

Sorting that out is what we do. So we've built the Spark AI Maturity Score. It’s a free five-minute benchmark, built from two years of working inside more than 80 agencies. It's the same assessment we run with clients before we start a programme, and it shows you where your AI effort is getting stuck and what to do about it.

As a thank you for being a Spark Intelligence reader, you're getting it first. It opens to everyone tomorrow. What five minutes gets you:

Free · 12 questions · 5 minutes · report emailed instantly

First we ask a few basics about your agency, size and discipline mostly, so your score gets compared against businesses like yours. Then it's twelve questions, and your report lands in your inbox the moment you finish.

Feeling capable and being capable are two different things, and most of the industry is sitting in that gap without knowing it.

It makes sense when you think about how confidence gets built. Someone uses ChatGPT every day and it saves them an hour. Someone else has built a very clever workflow for briefs. The tools are good, the wins are real, and the feeling that follows is "we're doing well with this."

What that feeling doesn't tell you is whether any of it has changed how the agency runs. Whether it's shared beyond the people who happen to enjoy it. Whether anyone owns it. Whether it's altered a single line of how you price or deliver work.

The score measures that second thing: what AI has actually changed in how your agency operates.

There's no shortage of AI maturity models about at the moment. Most are built from opinion, or from a survey where people self-report how well they're doing – which, given everything above, is close to the least reliable data you could use.

Ours came from the work. When we started with agencies in 2024, we expected the industry to split cleanly into two groups: the ones using AI and the ones who weren't. The reality was much messier. Agencies that looked identical from the outside were in completely different places – some with brilliant individual users and no shared standards, others with AI embedded in daily workflows who hadn't changed a thing about how they charged for the work.

We built the Spark AI Maturity Model® to describe that pattern properly, refined across two years and 80 agencies. It's the same framework behind the Spark Report, our free study of AI maturity across the industry, with the fifth edition publishing this October.

Every question in the score maps back to it.

Your score is built from four parts of your approach to AI – Vision & Strategy, People, Skills & Culture, Data & Tools, and Governance & Accountability – and places you at one of four stages.

Stage 1: Experimentation – individual use, no shared direction
Stage 2: Adoption – agreed goals, trained staff
Stage 3: Optimisation – agents changing how you work
Stage 4: Transformation – AI shapes what you sell

Most of you will land at Stage 1 or Stage 2, and there's no bad score here. In two years of measuring, no agency has entered a Spark programme above Stage 1. Stage 3 already puts you ahead of most of the industry, and we don't expect many organisations to reach Stage 4 for some time yet.

Most agencies score unevenly across the four, and it's the distance between your lowest and your highest that is the most interesting indicator of where you are.

You can't move forward without closing that distance. For example, you won't reach Stage 3, where AI starts shifting your commercial model, while Governance & Accountability is still stuck at Stage 1. Your weakest area is the one setting your ceiling.

Most leaders are starting to think about 2027 budgets in the next few weeks. Knowing which part of your approach to AI is weakest will help you target that budget in the right places. Five minutes now gives you an evidence-based answer to take into that budget conversation rather than an instinct.

Take the score, then send it to two people on your leadership team and ask them to take it separately, without discussing it first.

Leadership alignment on the role AI should play in your business is crucial. Nothing answers that faster than three of you filling in the same twelve questions and comparing what comes back. If you all land in the same place, you've got alignment worth building on. If you don't, you know where you need to go to focus first!

Every completed score is anonymised and collated into the free Spark Report, so the more of you who take part, the better the peer comparison gets for everyone. Complete it and you'll get early access to our biggest benchmark to date in October before we publish it out in the wild.

And you won't have to wait until October to see something back. Take it this week and I'll publish where this list landed in the next issue, so you can see how your score compares with everyone else reading this.

Free · 12 questions · 5 minutes · instant results

So many updates this week that despite this newsletter being a giant already I would be remiss to skip this section!

Big news for media agencies. This is not strictly just a tool update – more an operating model update. Omnicom's Omni platform has been stitched into Adobe's creative and marketing stack, and combined with Acxiom's 2.6 billion verified IDs underneath it. 

They’ve built five use cases across four verticals – retail, financial services, pharma and automotive – deployed over twelve months, taking a client from brief to activation without switching tools, with compliance baked in for the regulated sectors.

On the outside it looks like a brilliant new toolset, but what Omnicom is really packaging is an operating model, and selling it as the product. This is a live example of AI reshaping services and the commercial model themselves. If the holding companies can productise how the work gets done, independents will need an answer to that in the next eighteen months, and the answer is unlikely to be a better tool stack. If I haven’t said it enough already (sorry), identifying your AI Maturity Score is a good place to start. Read more →

You used to have to upload documents into a notebook by hand. Now you can point it at a folder and it reflects file changes automatically – added documents, permission changes, edits. For us this is huge - keeping notebooks up to date was a big friction point. Worth knowing before you set it up: delete a file from Drive and it disappears from your notebook sources too. Read more →

Previously, you mainly worked with the data through rows, columns, charts and dashboards. With the new Sheets Canvas, you can ask Gemini to turn that data into things like a Kanban board, interactive dashboard or tracker. You can then interact with that view directly, and any changes are reflected back in the original spreadsheet. We’ve been playing around with it and it’d pretty good - like creating a Claude Artefact or something similar, but your changes actually reflected in the original data. It’s on a phased rollout right now. If you have a pricing spreadsheet, a interesting place to start is turning it into an interactive tool for the team. Read more →

Last week my co-founder Jules took over our monthly Spark Session to break down what an agent really is – a narrowly defined job that runs on a trigger, reaches into your systems, and leaves the result somewhere useful – then built five live across Claude, ChatGPT and Gemini.

His rule for anyone starting out is point agents at your internal work first. Most of the value sits in replacing the human glue between systems, and it's the lowest-risk place to learn before anything goes near a client.

We also run an Agent Build Sprint – two days building and refining agents with your team. See what's involved →

We've been flat out this summer running AI Accelerators with clients including Pearlfisher, Radley Yeldar, Big Group and ACC in the US and Kontrapunkt and Everland in Denmark, and we are off to Ireland this week to see Javelin. We’ve been loving guiding these incredible businesses on their AI journey, though it's also the reason the Spark AI Maturity Score nearly missed the date we'd set for its launch!

We built the tool, but we were running out of road, so our friends at Kreante got it production-ready and secure. I met Jorge, who runs Kreante, at an Agency Hackers conference in Miami – a really lovely and seriously skilled team. Thank you Jorge and Dario, and worth knowing if you're building something of your own.

If someone on your leadership team is asking where your agency stands with AI compared to your competitors, forward this over.

That’s all for today, see you next time.

Emma
Co-founder, Spark AI

P.S. The Maturity Score opens to everyone tomorrow, 25th August. Until then it's just you brilliant lot. Get in there first and let me know what you think. Get your score →

Spark AI is the AI performance partner that builds organisation-wide AI capability for agencies and brand teams. How? Through role-specific AI training, leadership coaching, AI consultancy and agent building.

We’ve worked with 80+ agencies, published the #1 bestselling book on AI for agencies and publish the AI in agencies benchmark, and teach AI Strategy at Oxford University.