How Soar rebuilt for a market that stopped rewarding the old playbook

We've spent the last six months rebuilding Soar from the inside — the operating model, the strategy frameworks, the way our teams spend their time. There wasn't a single moment that triggered it. It was a direction of travel.

Through the back half of 2025, a handful of signals stacked up that made the same point from different angles. Y Combinator formally backed a new category of AI-native agencies. The cost of building software collapsed by an order of magnitude. Meta, who spent the previous year shipping little of consequence for the advertisers pouring billions through its platform, started releasing features that mattered again: Partnership Ads, incremental attribution, a real step-change in Advantage+.

The conclusion was unavoidable. A new standard was forming for what a best-in-class performance agency looks like, and the operating model that had worked for the previous six years wasn't going to carry us into the next six.

What I saw most agencies do, and are still doing, is try to graft AI and automation on top of the way they already work. That's the wrong instinct. You don't get a modern agency by adding tooling to an old process. You get one by rebuilding from the fundamentals of operational excellence and data, then letting tools strip out the layers of executional work that were never where the value lived in the first place.

What a strategist actually does now

The headline numbers: strategists are spending over 50% less time in ad accounts, and around 70% more time on strategy.

The more important question is what "strategy" actually means in that reclaimed time, because it's not the same job it was 18 months ago.

Eighteen months ago, a strategist's week was dominated by the mechanics of the account. Things like:

• Pulling reports.
• Pacing budgets.
• Screenshotting performance.
• QC on launches.
• Writing the Monday recap.

Most of the intellectual work that would actually move a client's business — persona research, funnel design, creative strategy, offer development, account structure planning — was getting squeezed into whatever hours were left.

Now that work is the job. Strategists are running customer and market research, leading persona workshops, designing funnels, briefing creative against specific awareness phases, and planning account structures around forecasted creative output rather than the other way around.

The most underrated effect is the one that's hardest to put on a slide: they have more time to think. Most of what separates good strategy from bad isn't effort — it's whether anyone had enough space to diagnose the actual constraint and aim effort at the point of maximum leverage. The old model eliminated that space by default.

The visible change from the client's side is that we direct the growth process with more clarity and conviction. Teams on both sides get held to account. The conversation has levelled up.

Meta has changed more than most operators realise

I look at this from the seat of a group managing over £100 million a year in Meta spend, so I'll put it plainly.

The platform has continued to shift away from a campaign-led model toward a persona- and creative-driven one. Advertisers who have rebuilt their process, their people and their inputs around that shift are extracting better performance than they ever have. Most pure performance agencies have not kept up.

Meta is now the largest advertising platform in the world, a position it reached this month, on the back of the most tangible AI-driven upside of any major tech company in recent earnings cycles. The advertisers compounding on that infrastructure and the ones falling behind are diverging fast, and the gap is widening rather than narrowing.

We saw this early. Our acquisition of Hambi Media in 2024 gave us a true performance creative capability inside the group before "Andromeda" became a term people threw around in DTC Slack channels. That timing matters: the creative operation is no longer a service line that sits next to media buying. It's the backbone of the entire engine.

Forecast creative the way you forecast spend

Every brand should now be forecasting creative the same way it forecasts ad spend. If you can't tell me how many concepts you need to ship next quarter to hit your spend target, you don't have a plan, you have a wish.

Done properly, the full operation runs roughly like this:

1. A forecast of how many assets you need, derived from the spend level you're trying to hit.
2. A deep understanding of the customer and the market, built from AI-driven research.
3. Personas translated out of that research, backed by data that validates them.
4. A strategist bridging forecast and research into a plan split by awareness phase, persona, format and vehicle.
5. Execution split across UGC creators, static designers and higher-production work.
6. Filming, production and post.
7. Testing and feedback loops that make next month's work sharper than this month's.

Most brands dramatically understate their content requirement. Performance creative is an outlier game. Even best-in-class teams see hit rates around 20%, which means volume and variety aren't vanity metrics — they're the mechanism by which you generate enough signal for Meta to find the next pool of customers you haven't reached yet.

Only around 10% of the brands we meet walk in the door with a mature process in place. The ones that do almost always blend internal and external sources of creativity. The honest answer on in-house vs agency is that brands absolutely can build this internally, but the right lens is return on capital.

If a creative partner is paying back in spend and growth, you add more sources of creativity. You don't look at it as a cost line; you look at it as a production input to a business outcome.

One client we work with has gone from roughly 100 assets a month to well over 4,000. Another is running 500+ new ads a day in the US market. That's what it takes now to set the pace.

One client we work with has gone from roughly 100 assets a month to well over 4,000.

The 97%, and why influencer and paid are converging

The old TOFU/MOFU/BOFU funnel has been replaced, in our model, by creative awareness phases: unaware / problem-aware, solution- and product-aware, most-aware, in that order.

Most of any brand's addressable market sits in the first group. They don't know your product exists and haven't articulated the problem it solves. We build for all three phases, but the counterintuitive truth is that most of our clients' outlier ads — the ones that absorb the majority of spend and do the heaviest lifting in the account — live in the unaware and problem-aware segments. That's where the TAM expansion actually happens.

Partnership Ads have been a step-change in how we operate here. Running paid media through a creator identity is driving roughly 30% improvements in new-customer ROAS for us. Influencer and paid media, in that sense, have genuinely converged at the execution layer.

Running paid media through a creator identity is driving roughly 30% improvements in new-customer ROAS for us.

But I'd push back on the idea that influencer and paid should just collapse into one discipline. The nuance matters. In this new world, creativity is king and follower count matters very little to performance. Our focus is less "influencer marketing" and more working with a growing pool of high-quality creators, on Meta and TikTok Shop, as a creative input into paid — rather than as a separate channel with its own vanity metrics.

Three things survive automation

Meta's direction of travel is clear. More real-time model adaptation. More agentic optimisation inside Ads Manager. More AI tooling, native and via third parties like Manus, for analysis, creative generation and account management. More of the executional work that agencies have historically billed for is going to be stripped out of the stack entirely.

I don't think that's a death sentence for the agency model. Three things remain genuinely valuable, and get more valuable as the rest automates away:

• Strategy. In an environment of higher uncertainty and more capability, the premium on diagnosing the right problem and aiming effort at the right lever goes up, not down.
• Creative generation. I don't believe this reaches 100% AI, if only for policy reasons. The brands compounding on Meta are the ones pairing AI scale with human taste and judgement.
• Certainty. Many years of collective domain-relevant experience, combined with the context generated via our data set, lets us be precise with recommendations and stay a step ahead of what's possible from an in-house team alone. It reduces the risk when allocating significant budget in pursuit of a goal.

The agencies that survive are the ones that rebuild their operating model around those three things, and let software and automation handle the rest. The ones that bolt AI on top of the old model will get commoditised.

That rebuild is uncomfortable. It reallocates people, rewrites job descriptions, and forces you to admit that a lot of what you previously billed for was executional work the machine now does faster and cheaper. But it's the only way the maths works in the next cycle.

This is a moment of high uncertainty, and that's exactly what makes it a moment of strategic opportunity.

What we're building at Soar is a business designed to compound on uncertainty — one that puts the client's business outcome at the centre of a system that gets sharper every quarter. The next three years are going to reward operators who rebuilt honestly, and punish the ones who didn't.

We decided some time ago which side of that line we wanted to be on.

Olly Hudson is CEO of Soar Group, a UK-based performance agency managing over £100m/year in Meta spend across its client portfolio.