When Good News Becomes Bad News

What if your operational excellence is exactly what's destroying your pricing power? If you're deploying AI to make things faster and cheaper, you're building a time bomb into your business model. This week, it started detonating.

Microsoft beat operating results expectations last week, the stock dropped 10%, wiping out $357Bn in market cap.

On Tuesday this week, Publicis beat operating results expectations, and their stock dropped 9%.

The pattern of drops in software and services-exposed industries continued yesterday.

What's going on? How did this good news become bad news?

AI is a double-edged sword. And after 18 months of swinging one way, it has just started swinging back. There will be more.

AI is being deployed in a the vast majority of use cases to make things faster and cheaper. It worked brilliantly. And that's precisely the problem.

When you make things faster and cheaper, your clients have a reasonable expectation (actually, a requirement) that the savings get passed through to them. Maybe not immediately, but the cushion will be absorbed by smart Procurement teams soon enough. Operational efficiency creates a pricing paradox. The better your AI gets at optimisation, the more pressure you face to reduce fees. You've automated yourself into a race to the bottom.

As I've written previously, AI eats from the tail.

It starts with the tasks furthest from strategic value such as reporting, analytics, optimisation, execution, coordination and consumes them systematically. You can let it eat its way toward your vital organs, or you can surgically sever the tail and reposition upstream while you still control the narrative.

In the ad industry, most agencies are feeding the tail to the predator and calling it transformation.

The stock market didn't punish for poor execution. It repriced where defensible value actually sits in an AI-saturated world. When AI compresses the value of software, dashboards, and execution layers, technology itself stops being differentiation. It becomes a commodity where advantage lies with the largest, specialist players: the platforms that build the foundation models.

The value that remains defensible moves upstream to capabilities AI fundamentally cannot replicate: creative judgment that determines what should exist, cultural intelligence that understands human context machines miss, strategic narrative that imagines futures not present in historical patterns, senior counsel that holds up under board-level scrutiny when everyone has access to the same optimisation tools.

These capabilities were always valuable. AI just makes their scarcity visible by commoditising everything else.

There are exactly two strategies for applying AI to your business. Most people think they're pursuing both. They're not. They're choosing the first by default.

Strategy 1 is operational efficiency.

Deploy AI to automate existing workflows, reduce headcount, compress timelines, cut costs. Deliver the same output with fewer resources. Compete on efficiency and price. This path is easier. The technology exists. The use cases are obvious. The ROI calculations are clean. Every vendor is selling this path because it's demonstrable in a 30-minute demo.

And it's doomed. Because every efficiency gain you achieve must be passed to clients as fee reduction. You're optimising toward commoditisation. When everyone has access to the same models running the same optimisations, efficiency becomes table stakes, not differentiation. The value flows through you to your clients. You become a pipe, not a partner.

And not just a pipe, but a pipe with a bomb in it. You're building this efficiency model on tools that are artificially cheap right now. OpenAI's API pricing has dropped from $36 to $2 per million tokens. Every major model provider is burning venture capital to gain market share. When those subsidies end, and they will, your margin gets compressed from both directions simultaneously. Clients expect the fee reductions you promised based on current AI costs, but your actual tool costs rise dramatically. You've contractually committed to pricing based on temporary subsidies. Boom.

This is how good news becomes bad news: your efficiency achievements create the expectation of fee reductions. Your margin improvements become your clients' pricing leverage. Your operational excellence becomes your strategic vulnerability.

Strategy 2 is to become undeniably better.

Deploy AI to enable capabilities that were impossible before, and to to scale the beautiful and unique capabilities of humanity. Create outcomes clients couldn't achieve elsewhere. Solve problems they didn't know were solvable. Compete on insight, imagination, and strategic value. This path is harder. The use cases aren't obvious, you have to discover them. The ROI is ambiguous until you prove it.

But it's the only path that sustains premium positioning. When you make something undeniably better, clients don't expect fee reduction. They expect to pay more for access to capability that creates competitive advantage. The value stays with you. You remain the strategic partner.

Being the best of the worst isn't enough anymore. You need to be fundamentally better at something that matters, access new pools of value that aren't priced with "agency holdco" gravity, and celebrate that difference openly enough that the market recognises you're playing a different game entirely.

Operational efficiency alone doesn't get you there. It just makes you a more efficient version of the category everyone's exiting.

To be clear, this is not an OR choice between the two. It's an AND. Do strategy 1 to release cash and capability to build strategy 2. Most stop at 1 and take a breather. Mistake.

The upstream approach starts with strategic formation, not tactical execution. Use AI where human insight matters most: understanding cultural nuance that quantitative data misses, identifying opportunity spaces that don't exist in historical patterns, imagining brand positions that have no precedent, forming strategy before optimising tactics.

There's another critical dimension to getting upstream: you need to stay out of the scope that platforms are coming for. Google, Meta, Amazon and others are building the execution layer. Campaign optimisation, media buying, performance reporting, audience targeting… each of these capabilities are becoming platform features, not agency services.

Agencies will roll their eyes are this assertion and say they aren't there yet. It's true, but I'd focus on the 'yet'. That 'yet' closes sooner than you think and they are ruthless, resourced and focused and they will take this scope. The platforms have structural advantages you cannot match: proprietary data, integrated infrastructure, zero marginal cost for additional features.

I have experienced this myself in the past year building Simbioniq. We don't make traditional human insight faster or cheaper as that would be Path 1, doomed to price compression and eventual platform integration. We enable strategic exploration that wasn't possible before: dialogue with psychologically complete simulated populations about decisions that don't exist yet, things that real people can't imagine yet. The insight doesn't get faster. It gets better. Undeniably, demonstrably, premium-commanding better. And critically, it sits upstream from what platforms can productizse as it requires deep domain expertise, psychological sophistication, and strategic judgment about what questions matter before you optimise how to answer them.

That's the reframe. That's upstream. That's where defensible value lives in an AI world, and where platforms won't follow because it doesn't serve their business model.

The McKinsey story I wrote about last week revealed something critical. When Managing Partner Bob Sternfels described deploying 20,000 AI agents alongside 40,000 humans, he didn't claim AI replaced strategic thinking. He identified three capabilities AI fundamentally cannot replicate, and the phrasing matters: these aren't capabilities AI threatens, they're capabilities AI liberates from tactical work.

First is aspiration.

AI optimises toward patterns in training data. It cannot imagine futures that don't exist in historical precedent. Large language models work by predicting "given everything I've seen before, what's the most likely next word?" That's interpolation, not extrapolation. It's finding patterns within known territory, not imagining beyond it. Humans set the aspiration, the stretch goal, the category-creating vision, the discontinuous leap. AI can plot the next point on your current trajectory. Only humans can choose an entirely different trajectory.

Second is judgment.

AI has no "should," only "can." It cannot determine what your organisation should pursue given your values, risk tolerance, competitive position, and strategic intent. When an AI system makes a recommendation that fails, who's accountable? The vendor who built it? The data it trained on? The executive who approved it? This ambiguity isn't a bug, it's fundamental to the technology. Strategy requires judgment that cannot be automated because judgment requires accountability, and AI cannot be held accountable.

Third is discontinuous thinking.

When you ask an AI to generate something, it's fundamentally predicting statistically probable combinations from its training data. That's inherently sequential and incremental. Breakthrough innovation requires abandoning linear extrapolation entirely. AI can recommend better breeding programs for faster horses. Humans imagine abandoning horses for automobiles.

Here's what's optimistic about this: AI doesn't threaten these human capabilities, it liberates them.

You need to be fundamentally better at something that matters. Not 10% better at what everyone does. Categorically different at something clients can't get elsewhere. This means building capabilities, not implementing technology. Simbioniq chose psychological completeness in simulation over speed in survey administration. That's a capability distinction, not a feature advantage.

You need to access new pools of growth that aren't priced with your current category's gravity. As long as you're compared to other agency holdcos, you're trapped in their multiple. You need to reframe what business you're actually in. Are you an agency that uses AI, or are you a strategic intelligence company that happens to have grown from agency roots? The framing determines the valuation.

The market movements this week weren't about poor performance. It was about the market pricing in that operational good news like efficiency, automation, optimisation (all of which would have been welcomes with fanfare before) creates strategic bad news when it's not paired with upstream repositioning.

The agencies and brands that survive 2026 won't be the ones who automated fastest. They'll be the ones who repositioned most deliberately and who severed the tail before the predator consumed the body, who elevated human capabilities while AI handled optimisation, who chose undeniably better over marginally cheaper, who stayed upstream from platform scope.

This is about using AI properly.

Not to replace strategic thinking, but to liberate it from tactical execution.

Not to eliminate human judgment, but to create space for judgment that actually matters.

Not to compete on efficiency in territory platforms are claiming, but to compete on imagination, insight, and the strategic value that only humans can create in domains platforms won't pursue.

Most agencies will choose the first path only without realising they chose it. The market just told you what happens when you do.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 5 February 2026 in the Brandflow newsletter on LinkedIn.