👋 Greetings earthlings,

Emma here, co-founder of Spark AI. This week: how AI is reshaping the middle of your org chart. But first – two weeks ago I asked you to score yourselves against Anthropic's AI fluency benchmark (catch up and score yourself here). Nearly half of you scored high fluency, a third said you had room to grow, hardly any in the middle. If you're in the ‘room to grow’ camp, jump back to the last newsletter – once people start doing a few of the higher fluency behaviours consistently, we find the rest tends to follow.

Now to today. Every panel I've been on recently circles the same question: what does AI mean for junior roles? It's hugely important. And we cover how agencies are tackling this in our AI skills and hiring report. But focussing only on junior roles risks missing the bigger picture. Your account directors, strategy leads and project managers are sitting at the most interesting intersection in your business right now - between what AI is taking on and what only humans can do. Today I'm looking at how to reorient that middle layer as the key unlock to navigating your team through the AI era.

Plus – it's one of our busiest weeks of the year at Spark, with programmes running at agencies including Publicis and JKR, three industry events, and our team on stage at the British Library. More on that below.

What's inside:

• AI and the middle of your org structure

• How to reorient gradually

  • 1. Use AI to do things better, not just faster

  • 2. Human coordination still matters

  • 3. Build the capabilities the new role needs

• Monks and the billable hour

• Tool updates for everyone

  • Microsoft: Copilot Cowork

  • OpenAI: GPT-5.4

  • Google: Gemini in Google Workspace

• Where to find us this week

The most interesting strategic question in your agency right now

The teams navigating AI well aren't making dramatic restructuring decisions. They're making small, deliberate shifts – quietly changing what their best people spend time on, bit by bit. And the place they're doing it most is the middle of the org chart.

Your account directors, strategy leads and project managers sit at the exact point where AI capability meets human judgement. As AI starts to help with the mechanics – the status reports, the briefing documents, the scheduling, the information routing – what gets freed up is something more valuable to your business. It’s the ability to read a client relationship. To find the insight that no report will surface. To know which creative team is burning out before anyone says a word. To walk into a pitch and feel totally confident to hold the room.

That work doesn't disappear when you introduce AI. It has the opportunity to expand. The question is whether you're intentional about that expansion – or whether the time saved elsewhere just gets reabsorbed into more volume.

Deloitte found that managers currently spend nearly 40% of their time on administrative problem-solving, with only 13% devoted to developing people. That 13% is where your agency's real value lives – in the mentoring, the judgement calls, the client knowledge built up over years. AI creates the conditions to do far more of it. Most agencies aren't being deliberate enough about making that happen.

Gartner's November 2025 research projects that 32 million roles a year will need to be significantly redesigned from 2028 onwards. For an independent agency, that's not a distant macro trend. It's a decision about your team that you're probably already being nudged towards – and one that's far easier to make thoughtfully over the next year or two than in a hurry when the pressure peaks.

Tsedal Neeley at Harvard Business School has a term for what this actually requires: “change fitness” – the ability to anticipate disruption and redesign your value before the market forces you to. The agencies developing it now won't look dramatically different tomorrow. But in a year or two, the distance between them and everyone else will be visible.

McKinsey's research from way back in 2023 still holds true: middle managers are the people who determine whether AI actually lands across an organisation. They're the ones applying judgement to outputs before they reach clients, coaching teams on what good looks like, and helping everyone else figure out what to do with the time AI gives back. In 2023 McKinsey thought nearly half of managerial work could be handled by AI - I’d love that to be re-validated now. But the point I’m trying to make is whatever the percentage is - are you actively steering what your best people do with the other 50%?

How to reorient gradually

One thing I've noticed in life is that tasks have a habit of filling the time we give them. Without actively deciding to spend your time in different ways, any time you free up simply gets absorbed in the mundane or unmemorable. Same goes for the time we ‘save’ with AI.

With that in mind, these are the areas we'd suggest looking at first.

1. Use AI to do things better, not just faster

The opportunity for middle managers is to use AI to raise the quality of what they do. Status reports written from transcripts aren't just quicker; they're more consistent. Briefing documents built from structured inputs aren't just faster to produce; they're more relevant. The question to ask is "what does my work look like when I'm not spending half my day on the mechanics of it?".

2. Recognise that human coordination still matters

AI can orchestrate workflows, but it can't hold a room together. The judgement calls that make coordination work – knowing when to flag something before it becomes a problem, reading the mood of a client relationship, deciding what doesn't need escalating, mentoring and growing junior team members – are deeply human. The middle layer is the glue for all of this. Don't restructure it out of existence; design it around what it does best.

3. Build the capabilities the new role needs

Workflow design, data fluency, and commercial confidence – the ability to explain to clients why they're paying for expertise and outcomes, not hours. Some of your middle managers will take to this quickly. Others will need more support. Both are fine. What isn't fine is assuming it happens without intention.

The data from our most recent AI in agencies benchmark - The Spark Report - points at exactly the problem you're trying to avoid: nearly 90% of agency staff are already recovering one to ten hours per week through AI – but that time is being reabsorbed into more volume rather than reinvested into doing things differently. To get real advantage you need to be deliberate about what to do with the time you get back.

Monks and the billable hour

David Wain Heapy from Prodigi (global talent - they are excellent) posted about this on LinkedIn, and it's been rattling around in my head since: S4 Capital's Monks has become one of the first major agency groups to move away from the billable hour. In early 2026, they began transitioning towards a subscription model – recurring fees giving clients ongoing access to senior talent, AI workflows, agents and institutional knowledge.

The logic is this: that the subscription is designed to get better over time. The idea being that as AI systems improve, the agency produces more work within the same fee structure. Wesley ter Haar, Monks' co-founder and chief AI officer, put it this way: something that once delivered 50 outputs a month might reach 70 as models improve; creating ongoing value inside the same contract.

Monks is targeting 25% of its revenue from the model by end of the year, with teams reorganised globally around clients rather than offices.

But what is Monks actually selling in this model? Perhaps it's not outputs anymore, but the systems that create the outputs. With that in mind the idea of a subscription makes sense. It shifts them from being a purely creative agency to a creative technology company. And if you're selling systems rather than outputs, your middle layer becomes the people who shape those systems, grow the talent around them, and make sure the work stays good – which brings us right back to the question above.

Tool updates for everyone

The relentless pace keeps on coming! Here’s the latest roundup for you of what’s coming across MS Copilot, ChatGPT and Google Gemini. Something for everyone:

Microsoft: Copilot Cowork

Microsoft built its newest Copilot feature on Anthropic's Claude, not OpenAI. Copilot Cowork lets you delegate multi-step work – competitive analysis, pitch prep, briefing packs – and it runs in the background, checking in at each stage. It draws on context across Outlook, Teams, and Excel so it's working with what it already knows about your clients and projects. Currently in research preview; broader access expected late March.

OpenAI: GPT-5.4

The most relevant feature for creative teams: mid-response transparency. GPT-5.4 outlines its plan before executing, so you can course-correct before it finishes – fewer wasted outputs when the AI misreads a brief. It's also OpenAI's first model with native computer-use capabilities. Available now on Plus, Team, and Pro plans. If you're on 5.2, you'll be migrated by June.

Google: Gemini in Google Workspace

Gemini can now pull from across your entire Google ecosystem – Gmail, Drive, Sheets, Docs – to produce a first draft, rather than working within a single document. The "Ask Gemini in Drive" feature lets you query across multiple files at once, which has obvious uses for research and client work. Rolling out in beta for AI Pro and Ultra subscribers; most Drive integrations are US-only for now.

Where to find us this week

Wow it's a busy week for team Spark! While we're running workshops and AI programmes with clients including JKR, Publicis (part of their brilliant AI festival), Epoch, Sampson May and Grande Cosmetics, we're also out in force across three industry events:

Today Jules is speaking at thenetworkone's Indie Summit in Bangkok – three days with independent agency leaders from around the world.

On Wednesday, Jules and I are both at BenchPress Live at Battersea Arts Centre, where The Wow Company reveal the latest financial benchmarks for £1m+ agencies. It's sold out, but if you're on the list we'll see you there.

And on Thursday, three of our AI Accelerator coaches – Dr Matthew Maxwell, Rob Crow and Simon Helm – are speaking at The Robots Are Coming at the British Library. Matthew, Rob and Simon lead the AI for Creatives stream of our AI Accelerator, our three-month agency transformation programme, and together they'll be showing how to turn AI into a joined-up creative capability using node-based workflows. I'll be there too – if you're going, come and say hello or drop me a line and we can grab a coffee.

Here's some of the dream team making it all happen (I actually have 6 more people to add - going to have to come up with a new team pic format to fit everyone in!)

Food for thought

Going back to the main feature of today, the middle management question and the commercial model question are really two sides of the same coin: what is your agency's value in an AI era, and how do you structure your team and your pricing to reflect it?

There's no single right answer, but the teams feeling most confident right now are the ones deliberately asking these questions rather than waiting for the market to answer it for them.

That's all for this edition. See you next time,

Emma

Co-founder, Spark AI

About 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 60+ agencies, published the only book on AI for Agencies, and teach at Oxford University.

👉 Find out more 

Not a subscriber yet?