We're in the middle of the most significant platform shift since mobile. LLMs have gone from research curiosity to production-grade tools faster than any technology I can remember.
But here's what most people are getting wrong about the opportunity: they're building horizontally.
Every developer has seen the pattern. Generic AI writing tools. Generic AI customer service. Generic AI for anything-you-can-think-of. The playbook is the same: take a powerful model, wrap it in a clean UI, slap a $20/month subscription on it, and ship.
The problem? That playbook is dead.
The Horizontal AI Market Is Already Saturated
ChatGPT, Claude, Gemini — these are the best horizontal AI tools humanity has ever built. They're genuinely excellent. And they're free, or nearly free.
You are not going to win by building a better generic AI assistant. The foundation models win that war, and they'll keep winning it.
The opportunity is everywhere else.
What "Vertical AI" Actually Means
Vertical AI isn't just about adding AI to an existing SaaS product. It's about building systems that deeply understand a specific workflow, speak the language of a specific industry, and are integrated into the actual tools that industry runs on.
Think about what makes vertical AI genuinely defensible:
- Domain-specific training data — You're not just prompting a foundation model. You're building institutional knowledge of an industry.
- Deep integrations — You're connected to the actual systems of record: the PSA software, the ERP, the ticketing system, the scheduling tool.
- Workflow automation — You're not just answering questions. You're taking actions inside existing workflows.
- Industry-specific trust — Customers trust you because you understand their world. A generic AI tool doesn't know what a P1 ticket means to a managed service provider. You do.
The MSP Industry as a Case Study
I've been going deep on managed service providers — companies that manage IT infrastructure for small and medium businesses. Let me paint the picture.
The average MSP technician handles 8-15 tickets per day. The majority of those tickets are Tier 1 issues: password resets, connectivity problems, software installations, printer issues. Critically documented, repetitive, solvable by following a standard runbook.
These issues consume 60-80% of technician time.
Every MSP owner I've spoken to has the same pain: they can't grow because they can't scale their labor. The best technicians burn out doing repetitive work. New technicians take months to train.
The solution isn't just "add AI chatbot." That's the horizontal thinking.
The solution is:
- Deep integration with their PSA (ConnectWise, Autotask, HaloPSA)
- Automatic ticket classification and priority scoring
- Runbook retrieval and application for known issue patterns
- First-response automation for common issues
- Escalation to the right technician when human judgment is needed
- Continuous learning from resolution patterns
That's a vertical AI product. It speaks MSP. It knows what an alert from RMM software means. It knows the difference between a P1 and a P3. It knows which clients have SLAs that require 15-minute response times.
A generic AI tool will never know that. But I can build something that does.
Why Engineers Have an Unfair Advantage
Here's the insight I keep coming back to: the verticals with the biggest opportunity are the ones that are "boring."
Not boring in a dismissive way. Boring in the sense of: unsexy industries that run the real world. HVAC contractors. Dental practices. Construction companies. Commercial real estate. Title insurance. Managed IT services.
These industries have three things in common:
- High labor costs and labor constraints — they desperately want leverage
- Existing software they're locked into — there's infrastructure to integrate with
- Low AI sophistication — they haven't been targeted by Silicon Valley yet
The founders going after these markets don't come from the industry. They come from adjacent places. Engineers who used to build enterprise software. Consultants who worked with these companies. People who just got curious.
But engineers specifically have an edge: they can go from validated idea to working prototype faster than anyone. The cost of an experiment is a weekend and some API credits.
The "Unfair Advantage" Framework
When I evaluate vertical AI opportunities, I use a simple filter:
Who has the right to win this market?
For the MSP AI market: someone who understands IT operations, knows the PSA software landscape, and has enough relationships to get customer discovery calls. That's a relatively small group — and I'm working to be in it.
For dental practice AI: someone who understands dental workflows, speaks the language of dental software, and can earn trust in a relationship-driven industry.
For construction project management AI: someone who knows what a general contractor's daily work looks like and can integrate with Procore and Buildertrend.
The moat isn't the AI. The models are commoditizing fast. The moat is domain knowledge + integrations + distribution.
What I'm Building
I'm applying this thesis directly. The AI Operations Manager for MSPs is my bet on this market.
The hypothesis: MSP owners will pay $X/technician/month to automate Tier 1 ticket resolution. The value proposition is clear — if we save each technician 2 hours per day, that's 40 hours per month, at a fully-loaded cost of $50-80/hour. The ROI math works before we've written a line of code.
That's the power of vertical AI. The value is so clear that customers can calculate it on a napkin.
What This Means For You
If you're a developer thinking about where to go next, here's my honest take:
-
Pick an industry you have some connection to. You don't need to be a domain expert. You need to be curious enough to do 20 customer discovery calls.
-
Find the most painful, repetitive workflow. AI thrives on repetitive structured work. Find the thing that a smart 22-year-old could do after reading a manual — and build AI that does it.
-
Map the existing software ecosystem first. Your product probably needs to be an extension of what's already there, not a replacement. Find the APIs.
-
Start with one very specific use case. The temptation is to build the "AI platform for X." Don't. Build the "AI tool that solves Y specific problem for X." Expand later.
The horizontal AI market is being fought by giants. The vertical markets are waiting for people who care enough to understand them.
That's where I'm going. Come with me.
If you're building vertical AI, or you work in an industry where you see this opportunity, I'd love to talk. Reach out →