background

Thomson Reuters may be down 18%, but here’s what you can do next as a SaaS founder.

Published about 1 month ago • 7 min read

The Numbers That Should Worry You

Last month, Thomson Reuters dropped 18%. LexisNexis fell 14.4%. Wolters Kluwer shed 13%. This all happened within days of Anthropic launching Cowork, a browser-based AI tool that automates the kinds of workflow tasks these companies charge per seat for.

Those weren’t just the outliers; over $2 trillion has been wiped from software company valuations in the weeks since. The software sector’s weight in the S&P 500 dropped from 12% to 8.4%. Figma is down 65% from its peak. Duolingo lost 68% of its market cap in six months. SaaS indices are down 22% while the broader market gained 17.6%.

If you’re a SaaS founder reading this, I know the instinct. You look at those numbers and think: “That’s public market noise. My product is good, my customers are still happy.”Maybe. But most commentators have the story wrong: this is a business model crisis that AI just made more urgent. There is something more fundamental at play here.

A trend that didn’t start with AI

The fact is, SaaS growth has been declining every single quarter since 2021. Before ChatGPT, before Claude, and before anyone used the word “agent” outside a spy novel.

What changed was the budget. AI is now consuming the IT spend that used to buy SaaS seats. Meta is spending $135 billion a year on AI infrastructure. Microsoft, $75 billion. The hyperscalers collectively will pour $470 billion into AI in 2026. Every one of those dollars is a dollar that never reaches another Salesforce seat or Workday module.Meanwhile, the “best-of-breed” era is collapsing. CIOs don’t want five-point solutions for marketing, sales, support, finance, and ops anymore. They want fewer vendors, less overhead and risk, not more.

The “growth” most SaaS companies reported over the past three years wasn’t real growth. It was price hikes on captive customers, expansion revenue from existing accounts, and contract uplifts with no connection to genuine product adoption. Strip those away, and net-new customer acquisition, the metric that actually matters, is the weakest line in public earnings.This sounds like more of a structural crisis, and AI just turned up the volume on it.

Why one seat no longer means one pound or dollar

The classic SaaS formula (seats × ARR per seat × growth multiple) assumed that each employee remained the unit of work. AI agents have severed that assumption and the charging model.

An AI-driven sales development agent can replace an $80k-a-year SDR and a $1,800-a-year license with a few hundred quid of compute each month. That’s an 80–90% drop in revenue per account, while the customer’s productivity actually increases. The vendor’s revenue collapses while the customer’s output rises, and the entire model breaks.

About 35% of SaaS companies have responded by hiking seat prices. I call this late-stage extraction, by squeezing more rent from a dying revenue stream. Ultimately, it’s self-defeating: the higher you push seat prices, the faster CFOs approve the AI replacement project.

The remaining 65% are experimenting with hybrid pricing: seat-based base fees plus a usage or outcome layer. It’s a sensible transitional move, but don’t mistake it for a solution. The sales teams, compensation plans, and billing systems are all built around seat counts. Re-engineering those structures is an operational problem, not just a product one.

And if you want to see the contradiction playing out in real time, look at Salesforce. They launched Agentforce at $125 to $550 per user per month, on top of already-elevated base pricing, at the exact moment their enterprise customers are cutting headcount because AI is handling more work. The CFOs noticed.Per-seat pricing is now becoming a liability, not an asset.

What AI has commoditised, and what it can’t touch

I’ve spent the past six months testing native AI workflows: browser automation, skills-based systems, and agent orchestration. Over that time, the pattern has become clear.

Anything probabilistic is in danger. Pattern matching, content generation, recommendations, and simple workflow automation. If a foundation model can replicate 90% of your core value at 1% of the cost, if that’s your moat, it's gone. Workflow automation is the number one category being replaced right now. Retool’s 2026 report says 35% of teams have already swapped out at least one SaaS tool for a custom-built solution. 78% plan to build more this year.

Most founders miss this: AI agents don’t use your beautiful UI. They call APIs directly. Your interface, the thing your design team spent two years perfecting, becomes optional when the user isn’t a human clicking buttons. This is the area to invest in. I heard one founder say he chose Salesforce or HubSpot purely because the APIs are more robust and feature-rich.

But currently, deterministic systems of record seem to be holding up. Those platforms that store proprietary customer data, enforce compliance, or run mission-critical logic where a single error has material consequences. At the moment, AI can’t easily replicate that function. For example, Toast owns every menu item, pricing change, and supplier relationship for thousands of restaurants. Procore owns the project data for construction sites. That depth of integration is the moat.

The survival line is simple: if your core value is probabilistic, you’re exposed. If it’s deterministic, you’ve got a foundation to build on. Be honest about which side you’re on.

Six questions every SaaS founder should answer on Monday morning

Before you touch your pricing page or rewrite your pitch deck, sit with these. Answer honestly. If you flinch, you’ve found the vulnerability.

  1. “What would break if we disappeared tomorrow?”​
    High switching costs and deep integration indicate a platform. If customers can find alternatives within a week, your offering is a feature rather than a business.
  2. “Are we the tool or the workflow?”​
    Tools are easily replaced, while workflows become embedded. Platforms outperform features.
  3. “Who would fight to keep us?”​
    Communities create defensible advantages that AI cannot replicate. Users who actively contribute are not found in training data. If no customer would advocate for you, your product is a line item, not a partner.
  4. “Can AI replicate our trust layer?”​
    Governance, compliance, audit trails, human-in-the-loop oversight. With the EU AI Act landing in August, the companies that build clear accountability frameworks become the trusted layer between raw AI capability and real-world consequences. If AI can replicate yours, you’re exposed.
  5. “Is our core value probabilistic or deterministic?”​
    This is the survival line. If your core is pattern matching and recommendations, you’re in the danger zone. If it’s deterministic systems of record, you have a foundation. Be honest.Use these questions as a starting point for your Monday morning meeting. Any uncomfortable answer highlights an area that requires strengthening.

The business model that actually works

Per-seat may be dying, but outcome-based pricing isn’t plug-and-play either. When one AI-native SDR company offered customers both outcome-based and activity-based pricing, 90% chose activity-based. Defining “qualified lead” or “successful transaction” varies so widely across industries that negotiation friction kills deals.
​
I’ve struggled with this while launching my own product, Value by Design. Initially, my instinct was to charge a flat fee for the whole package. It was clean, simple and seemed like the right route. However, after prototyping, it was clear from the questions people asked that a tiered base and optional layers on top were what customers wanted.

The more practical path is a hybrid business model. Charge a base platform fee that covers your deterministic core. This should include the data governance, the compliance layer, and the community infrastructure. That will give you predictable revenue. Then a consumption-based AI layer on top, metered per inference, per API call, or per workflow execution, with transparent usage dashboards so customers control their spend.

This isn’t just an easy price tweak. It involves retraining sales, redesigning compensation, and rebuilding billing pipelines. The operational friction exists, but the founders who move first on hybrid pricing will capture the trust premium. The ones who wait will be forced into it on worse terms.

The strategy principles haven’t changed.

The technology shifted; however, the fundamentals of good strategy haven’t.

Know the real job your customer is hiring you for. If your product is the place where customers think, decide, and store institutional knowledge, you own something AI can’t easily replicate. Design your model to capture that value, not the number of seats you sell.

Always test before you scale. If you are not devoting 10% of your budget to testing, you are not learning fast enough. Pilot a hybrid pricing tier with a handful of anchor accounts. Measure churn, lifetime value, and unit economics. If that is successful, then you are safe to roll out.

It appears that the per-seat era is ending. But what replaces it rewards the same things that have always mattered: clarity about who you serve, depth of understanding about what they need, and the discipline to build something genuinely hard to copy.

I wrote something recently that I keep coming back to:

​
“The only thing that matters now is what can’t be generated. That’s where all the money is hiding.”
​

What’s your next move?

If you’re a SaaS founder staring at a spreadsheet of shrinking ARR and wondering where the pivot is, reply to this email. I read everyone, and I’ll give you an honest take on where your trust layers stand. Alternatively, book a call.​

No sales pressure, just a straight conversation to see what can be done to help.

Have a great weekend!

Much love to you all,

Chris

Source Links

background

Subscribe to Polything Marketing Consultancy