This week around $300 billion was wiped off global software stocks.

Not because of interest rates.

Not because of inflation.

Not because of macro panic.

Because of one prompt.

Anthropic’s Claude Code dropped a 25-line “legal work” prompt on GitHub. Contract diffs. Risk flags. Compliance checks. The type of structured work that junior lawyers and mid-tier legal software tools monetise every day.

The output was strong enough that investors joined the dots fast.

If agents can handle legal workflows, which are regulated and auditable, what happens to CRM, invoicing, HR stacks, procurement tools?

The illusion cracked.

RELX dropped 14%, its worst day since 1988.

Thomson Reuters fell 16%.

LegalZoom dropped 20%.

Public markets overreact. They always do.

But directionally? This move makes sense.

Seat-based SaaS just met agent math.

See below to continue…

The London Defense Tech Network is the premier hub for connecting startups, investors, primes, government officials, and industry stakeholders across the UK’s defense technology ecosystem. Our mission is to foster collaboration, knowledge sharing, and innovation by bringing together the key players driving the future of defense technology.

Members gain access to curated networking opportunities, timely industry updates, and a collaborative space to share insights, explore partnerships, and accelerate growth. Whether you’re a startup seeking guidance, an investor evaluating emerging technologies, or a government or prime contractor looking to engage with innovators, this group offers a trusted, professional platform to connect with the right people.

Join us to be at the forefront of defense tech innovation, expand your professional network, and participate in shaping the future of the industry.

Continue…

For 15+ years SaaS followed a clean model.

Growth was tied to hiring.

If a company added 100 employees, SaaS vendors celebrated.

Agents break that link.

No seats. No dashboards. No onboarding plan.

Just execution.

If one system performs the work of ten employees, what happens to ten seats?

That’s the question investors reacted to.

Traditional SaaS economics looked like this:

Agent reality looks different:

The business doesn’t have to die to lose value.

It just has to look less defensible.

SaaS is not dying.

Lazy SaaS is.

Over the last decade, many products were thin layers on top of structured data. A clean dashboard. Filters. Notifications. A workflow wrapper.

The real value sat in the data.

Now large language models read that same data directly.

An agent does not need to click through ten tabs. It asks one question and gets a structured answer.

If your product is:

You are exposed.

Zoho’s CEO said it well. Industries that spend more on sales and marketing than on R&D were always fragile.

This is the maturity test.

Yes.

Enterprises won’t rip and replace overnight. Risk teams. Procurement. Internal politics. Audit trails.

That protects incumbents for longer than markets assume.

But pilots move fast.

General Counsel won’t fire lawyers because of Claude.

They will route repetitive work through it.

First-pass reviews. Redlines. Summaries. Risk scoring.

If your SaaS monetises first-pass workflows, your volume shrinks first.

And shrinkage kills growth multiples.

This feels dramatic because of stock price moves.

But this isn’t extinction. It’s repricing.

We’ve seen this before.

Cloud didn’t erase on-prem software overnight. It compressed multiples. It shifted power. It forced adaptation.

Mainframes still run banking systems today.

They just don’t trade at growth premiums.

That’s the risk here.

A SaaS company trading at 40x ARR doesn’t need to fail.

It just needs to look like a utility.

Suddenly it’s 12x.

Your £100M valuation becomes £30M without a single customer leaving.

That’s what investors are pricing in.

This one hurts.

Agents do not care about your UI.

They don’t admire your dashboard.

They don’t get confused by navigation.

They execute.

If your edge is design, layout, or workflow convenience, and not data ownership or system control, that edge erodes fast.

The question shifts from:

To:

That’s a different product philosophy.

The companies that survive share traits.

Deep integration.
They sit inside core systems. Removal creates operational risk.

Regulated complexity.
Healthcare, finance, energy, defence. Audit and compliance matter.

Proprietary data.
Unique datasets are the new moat. AI enhances them instead of replacing them.

Adaptive pricing.
Seat-based pricing weakens. Usage, outcome, and workflow pricing aligns better with agent economics.

If you own the workflow, you win.

If you rent the interface, you lose.

Founders need to stop debating and run a test.

Write your product’s core job in one sentence.

Now run it through an agent workflow.

Time it. Measure accuracy.

If the agent hits 80% on day one, your product strategy needs to evolve.

This doesn’t mean shut down.

It means move.

Maybe you stop selling seats and start orchestrating workflows.

Maybe you build the control layer that manages agents across systems.

Maybe you monetise validation and audit trails instead of access.

But pretending agents won’t improve is not a strategy.

There’s another layer most people ignore.

AI doesn’t remove work. It raises expectations.

If one founder produces in one hour what used to take a day, the baseline shifts.

More output.
More experiments.
More internal tooling.
Faster cycles.

That benefits companies that integrate AI deeply.

It hurts companies that bolt it on for marketing.

AI-native SaaS has a future.

SaaS that adds “AI-powered” to a weak core does not.

For UK and European founders, this is even more relevant.

Growth capital is tighter. Multiples are under pressure. You don’t have room for illusion.

If your business depends on headcount expansion inside clients, stress test it.

If your differentiation is UI, rethink it.

If your moat is real data and regulatory complexity, double down.

And consider this.

Government and regulated sectors won’t move overnight.

But they will test.

If you can position as the “agent orchestrator” inside regulated environments, you build defensibility and narrative at the same time.

The $300B wipeout wasn’t a collapse.

It was a filter.

Seat-based value got commoditised.
Thin wrappers got exposed.
Sales-heavy growth got questioned.

Strong products with real data, deep workflows, and adaptive pricing remain powerful.

The SaaS apocalypse is not the end of software.

It’s the end of lazy software.

If you’re building, don’t panic.

Audit your product.
Audit your pricing.
Run the 48-hour test.

Then move.

Because agents aren’t coming.

They’re already here.

✅ If this hit a nerve, send it to someone.

(👉 Vote now — we’ll share the results in next week’s issue. All votes are anonymous.)

(Last week’s poll results)