We’ve heard “AI will eat software.” The bigger story:AI is going to eat services.
By “services,” I mean asset-light, information-driven businesses whose core moat is business process logic embedded in software — underwriting, booking, logistics coordination, CRM, claims processing.
AI Eats Information Work — and Services Are Information Work.
My thesis is that AI will eat the legacy services companies—insurance (health, auto, property, life), CRM, e-Commerce, travel & leisure, banking, brokerage, legal, real estate and a lot of others.
Why? Because at the core of these services companies, there are business processes that were codified in software over the past decades. And now GenAI can enable startups to spin up service businesses that codify in software the entire redesigned AI-centric business with all its business processes, and do it at low cost and do it rapidly.
While the incumbent service business spend time using GenAI for automating one workflow at a time, startups can rebuild entire service businesses from scratch using GenAI.
Legacy Services vs Clean-Slate AI
1. Legacy systems at incumbent service businesses are not moats. They’re anchors.
Architectural Rigidity. Old software. Siloed data. Manual integrations. Multi-year vendor contracts.
These are not competitive advantages — they are drag.
A small team of sharp business process engineers who understand underwriting, logistics, claims, travel booking, CRM, etc. can now use GenAI to spin up:
- AI-centric business processes
- Clean data pipelines
- Agent-based workflows
- Entire operational platforms
In months. Not years.
2. Incumbents’ cost structures are anchored to maintenance.
Just keeping legacy systems alive consumes enormous operating budgets.
Now imagine:
A new GenAI built entire operational platform that costs roughly what incumbents already spend on maintenance — but delivers better service with a fraction of the labor.
That’s not incremental improvement. That’s economic displacement.
Incumbents aren't just paying for software; they are paying a complexity tax. When a health insurer spends $100M on maintenance, they are paying to keep thousands of brittle, manual integrations from breaking. An AI-native competitor doesn't 'maintain' integrations; it builds the entire service using redesigned AI-centric processes. This shifts the cost structure from Labor-plus-Legacy to Compute-plus-Logic, effectively reducing the cost of service by 90%.
3. Customers and partners aren’t sticky. They’re captive by friction.
Bookstores were “sticky.” Travel agencies were “sticky.” Brick-and-mortar insurance agents were “sticky.”
Until Amazon. Until Booking.com. Until online comparison engines.
History teaches a simple lesson: Customers aren’t loyal to process complexity. They’re trapped by it. Remove the friction — the “stick” disappears.
4. Can AI-generated software displace 20 years of legacy systems?
Yes.
The core moat of most service companies isn’t brand. It isn’t distribution.
It’s encoded business process logic sitting in legacy systems.
Mainframe actuarial code. Claims rules engines. Freight pricing tables. CRM workflow automation.
For decades, rewriting these systems was nearly impossible. With GenAI? That assumption is breaking.
Anthropic and Palantir focus on automating slices of business processes. But startups are beginning to think bigger: Not “automate a function.” Build a new AI-native service company.
5. Can incumbents pivot and rebuild themselves?
The probability is extremely low.
Rewriting core systems means:
- Re-engineering processes
- Challenging entrenched internal stakeholders
- Accepting short-term margin compression
Large organizations struggle to do this. They pilot GenAI for a few use cases. While startups design entirely new operating models.
Let’s Talk Specifics
Insurance
Can GenAI write systems for:
- Actuarial pricing
- Underwriting
- Claims processing
- Fraud detection
- Onboarding
- Customer service
Yes.
And not as disconnected tools — but as a unified AI-centric platform. Startups can build workflows incumbents would love to have — but can’t justify rebuilding.
Travel Booking
AI agents can:
- Optimize itineraries
- Price dynamically
- Personalize recommendations
- Handle service issues
All in a single conversational interface. What once required teams, patches, and millions of lines of code can now be redesigned cleanly from scratch.
Logistics & Supply Chain
Today’s systems are buried under:
- Thousands of relational tables
- NoSQL layers
- Data lakes
- Hundreds of APIs
GenAI enables something different: Re-architect the entire business around AI-centric decisioning — not bolt AI onto legacy databases.
CRM & Customer Service
AI already replaces routine contact center workflows.
But take it further: Can GenAI help create a new Salesforce or ServiceNow in 6–12 months at a fraction of current operating costs?
Not unimaginable.
Five smart engineers and $10 to 50M in VC funding can now challenge what once required billions.
Are We There Yet?
Not yet. But the constraints that protected incumbents are dissolving.
For the incumbents, the track you are on is not safe, the GenAI locomotive is coming.
Last Week’s Market Tremors Weren’t Irrational
The recent sell-off in service companies wasn’t random volatility.
Investors are starting to ask: What happens when AI doesn’t just assist services — but builds the entire service platform from scratch? Which moats survive that? Which don’t?
Conclusion: The New Frontier for Builders
The era of "AI-as-a-feature" is coming to a close, and the era of AI-native services is beginning. For decades, the "moat" of a service company was its encoded business process logic—hundreds of thousands of lines of code buried in legacy systems. Today, those systems have turned into anchors, dragging down incumbents with massive maintenance costs and siloed data.
While established giants spend years trying to automate a single workflow, the path is clear for a new breed of entrepreneur.
The Opportunity for "Sharp Engineers"
The next wave of disruption will not look like the SaaS revolution; it will look like the total economic displacement of legacy service providers. If you are a business process engineer who understands the intricacies of underwriting, logistics, or claims processing, you now hold the keys to the kingdom.
· Speed to Market: You can now spin up entire operational platforms in months, not years.
· Cost Advantage: You can run a service business at 10% of the operating cost of a legacy incumbent.
· Clean-Slate Design: You aren't "bolting on" AI; you are re-architecting the entire coordination layer around AI-centric decisioning.
A Call to Action
Startups are no longer limited to "automating a function". You have the tools to build a new Salesforce, a new UnitedHealth, or a new Booking.com from scratch with a fraction of the labor and a fraction of the capital.
"Incumbents are shackled by what they built. Startups are not."
If you are building an AI-native company designed to displace an entire service business, the world is waiting for your operating model. Don't just improve the process—replace it.