What the $285 Billion Selloff Gets Wrong About AI Disruption

This week, $285 billion evaporated from software stocks in a single day.

The trigger was Anthropic’s announcement of Claude Cowork plug-ins, AI tools that automate tasks across legal, sales, marketing, and data analysis. Thomson Reuters dropped 18%. RELX fell 14%. LegalZoom slumped 20%. Analysts described software sentiment as “radioactive.”

The market’s logic was simple: If AI can do what these companies charge for, these companies are dead.

I spend my days building AI automation workflows for service businesses. I use Claude in client implementations every week. And I think the market is asking the right question but drawing the wrong conclusion.

The question; can AI replace the workflows that software companies monetize?; is legitimate. The conclusion; sell everything now; misunderstands how technology adoption actually works.

Let me explain what I’m seeing from the implementation side, and what it means for business owners trying to make sense of this moment.

The Three Stages of Disruption (and Why They’re Not the Same)

There’s a critical distinction the market is collapsing: capability, adoption, and replacement are three different things.

Capability means the technology can technically do something. Claude Cowork can review contracts, flag risks, draft NDAs. This is real. I’ve built workflows that do exactly this.

Adoption means organizations actually start using it. This requires integration with existing systems, employee training, workflow redesign, compliance approval, vendor evaluation, and budget allocation. None of that happens overnight.

Replacement means the new technology fully substitutes for the old one and the old one loses its revenue. This requires the adopters to cancel their existing contracts, which requires the new solution to handle 100% of their use cases including edge cases, compliance requirements, and audit trails.

The market priced in replacement on Monday. We’re barely at the beginning of adoption.

Here’s what I mean concretely: A law firm using Westlaw isn’t switching to Claude Cowork because a plug-in exists. They’re switching when:

  • Their compliance team approves AI for legal research

  • Their malpractice insurance covers AI-assisted work

  • The tool integrates with their document management system

  • Their associates are trained on the new workflow

  • They’ve validated accuracy on their specific practice areas

  • The partnership votes to change a system they’ve used for 20 years

That’s not a six-month process. For enterprise software with deep institutional integration, that’s a multi-year transition.

What AI Actually Disrupts First

This doesn’t mean disruption isn’t coming. It is. But it’s coming unevenly, and understanding the pattern matters if you’re trying to prepare your own business.

AI disrupts first:

New market entrants who would have bought legacy software but now don’t need to. A solo practitioner starting a law practice today might skip Westlaw entirely and build a Claude-based research workflow from scratch. They have no switching costs, no existing contracts, no institutional muscle memory. This is where incumbents lose future growth, not current revenue.

Workflows that were manual but tolerated. Many businesses have processes that should have been automated years ago but weren’t worth the investment. AI drops the implementation cost dramatically. These aren’t competitive losses; they’re greenfield automation opportunities.

The pricing power of tools that charged premium rates for commodity tasks. If you’re paying $500/month for software that does something Claude can now do for $20/month in API costs, that price pressure is real and immediate. Not replacement, compression.

AI doesn’t disrupt quickly:

Deeply embedded enterprise systems with compliance requirements. Regulated industries move slowly for reasons that have nothing to do with technology capability.

Workflows where “wrong” has legal or financial consequences. When the cost of error is malpractice or audit failure, organizations want vendors they can sue. “Claude made a mistake” is not a defense that general counsel accepts.

Anything requiring audit trails the vendor is liable for. Accountability infrastructure takes time to build.

The Integration Advantage

Here’s the part the selloff misses entirely: the companies best positioned to capture AI value aren’t the AI providers. They’re the incumbents who integrate AI into their existing products.

Think about it from a customer’s perspective. If you’re a Thomson Reuters client, what’s easier?

  • Switch to Claude Cowork, rebuild all your workflows, retrain your team, migrate your data, and hope the new system handles your edge cases

  • Stay with Thomson Reuters as they add AI capabilities to the interface you already know

The incumbents have distribution, trust, compliance infrastructure, and deep integration with customer workflows. Anthropic has capability. In enterprise software, distribution usually wins.

The companies that will actually lose are the ones that fail to integrate AI — not the ones who face AI competition. The selloff should have been a discrimination between “companies adapting” and “companies not adapting.” Instead, the market sold the entire sector.

That’s an opportunity for investors, but more importantly, it’s a signal for operators: the question isn’t whether AI will change your industry. It’s whether you’ll be the one integrating it or the one being disrupted by someone who did.

What This Means for Your Business

If you’re a business owner reading this, the $285 billion headline probably triggered one of two reactions: panic or dismissal. Neither is useful.

Here’s what I’d actually do:

1. Audit your current software stack for AI vulnerability.

Look at every tool you’re paying for. Ask: What is this software actually doing for us? If the answer is “organizing information and applying rules to it,” that’s AI-vulnerable. If the answer is “providing compliance infrastructure, audit trails, and institutional trust,” that’s more defensible.

This doesn’t mean cancel everything. It means know what you’re paying for.

2. Identify workflows you should automate before someone automates them for you.

The real disruption risk for most service businesses isn’t that AI replaces your software. It’s that a competitor figures out how to use AI to deliver your service faster and cheaper while you’re still doing it manually.

Map your highest-volume, most repetitive workflows. Client intake. Proposal generation. Reporting. Follow-ups. These are automation opportunities, not threats; if you get to them first.

3. Start building AI capability now, while the transition is gradual.

The worst time to figure out AI implementation is when you’re under competitive pressure and moving reactively. The best time is now, when you can experiment, learn, and build institutional knowledge before it’s urgent.

This doesn’t mean hiring an AI team or making a massive investment. It means picking one workflow, building one automation, and learning what’s actually involved. The implementation experience is the valuable part.

4. Don’t confuse “AI is coming” with “AI is here.”

Timelines matter. If you’re making three-year strategic decisions based on what AI can do today, you’ll over-rotate. If you’re making three-year decisions assuming AI won’t change anything, you’ll under-prepare. The right frame is: AI will significantly impact my business within 3-5 years. What should I be learning and building now so I’m ready?

The Selloff as Signal

The $285 billion selloff wasn’t irrational. It was the market catching up to a reality that practitioners have understood for a while: AI is moving from “interesting technology” to “competitive threat.”

But the market overcorrected because markets are bad at nuance. They see “AI can do legal research” and price in “Thomson Reuters is worthless.” The implementation reality is messier and slower than that.

For business owners, the signal isn’t “panic about AI.” It’s “the window for gradual adaptation is closing.” The companies that will thrive through this transition are the ones building AI capability now — systematically, strategically, without either hype or denial.

That’s the work I do with clients at BAMPT. If you’re trying to figure out where to start, I’m always happy to talk.


Chantal Emmanuel is the founder of BAMPT, where she helps service businesses implement AI-powered operations. She’s also CTO of LimeLoop and writes about automation, systems thinking, and building businesses that scale.

Previous
Previous

OpenAI Wants to Be Your Company's Operating System. Here's What That Actually Means.

Next
Next

BUILD LESS. MEAN MORE.