Paul's Insight: A Real Pain Point
"A good AI tool would be to convert fabrication objects to usable Revit objects."
This is interesting. Paul just identified a specific, tactical workflow problem that he experiences RIGHT NOW. This isn't the grand "replace Revit" vision—this is solving a real pain point WITHIN the existing BIM ecosystem.
This raises a critical strategic question: Do we start with the wedge (solve a specific problem) or go straight for the vision (replace Revit)?
Understanding the Problem
The Fabrication ↔ Revit Gap
In the MEP (mechanical, electrical, plumbing) world, there's a painful disconnect:
- Design phase: Architects/engineers work in Revit (BIM) - higher-level, conceptual objects
- Fabrication phase: Contractors use Autodesk Fabrication software (CADmep, ESTmep, CAMduct) - detailed shop-level objects with exact dimensions, materials, connectors, etc.
- The gap: Converting between these two is manual, tedious, error-prone
Example: Design says "8-inch round duct". Fabrication needs "8-inch spiral duct, 26 gauge galvanized, with flanged connections, specific vendor part numbers." Converting this manually for thousands of objects = nightmare.
Why This Matters
Paul works on data centers - massive MEP coordination. Ductwork, piping, cable trays, electrical conduits. Thousands of objects. If the design team works in Revit but the fabrication shop needs fabrication objects, someone has to manually convert everything.
Why This is an AI Problem
This is pattern matching and semantic understanding - exactly what LLMs and ML are good at:
- Understand the intent of a Revit object (what is this duct trying to be?)
- Map it to appropriate fabrication objects (shop standards, vendor catalogs, specifications)
- Apply rules and constraints (building codes, project specs, material availability)
- Generate the fabrication objects with all the detailed properties
Current solutions are rule-based templates - rigid, brittle, require manual configuration. AI could learn from examples and handle variations intelligently.
Two Possible Strategies
Strategy A: The Wedge
Start with Fabrication ↔ Revit Converter
The Approach
- Build AI tool that converts between fabrication and Revit objects
- Sell as Revit plugin or standalone service
- Solve Paul's actual pain point TODAY
- Build within the ecosystem, not against it
Why This Could Work
- Real, immediate market (MEP contractors, data center designers)
- Paul can validate the tool himself
- Smaller technical problem = faster MVP
- Autodesk is SLOW at solving these workflow gaps
- Easier sales - not disrupting, just helping
The Wedge Logic
Start by making Revit users more productive. Gain trust, learn the domain, build relationships. Then expand: "Oh, by the way, we can also generate these objects from natural language..." Wedge your way into the bigger disruption.
Strategy B: The Vision
Go Straight to AI-First Design Platform
The Approach
- Natural language → construction-ready plans
- Skip Revit entirely
- Target underserved markets (ADUs, small projects)
- Disrupt from the outside
Why This Could Work
- Bigger market opportunity (everyone who needs building plans)
- Not dependent on Revit ecosystem
- True disruption, not incremental improvement
- Matches the "AutoCAD is dead" insight
- More VC fundable (bigger vision)
The Disruption Logic
Don't play in Autodesk's sandbox. Build the future they can't. Target customers who don't want to learn Revit. Create new market, then move upmarket. Classic disruption playbook.
Detailed Comparison
Time to Market
Wedge: Faster
- Smaller, more defined problem
- Existing data (Revit objects, fabrication catalogs)
- Can validate with Paul immediately
- 3-6 months to working prototype
Vision: Slower
- Bigger, less defined problem
- Need to build multiple pieces (parsing, generation, validation)
- Harder to validate without real customers
- 9-12 months to working MVP
Market Size
Wedge: Narrower
- MEP contractors and engineers only
- Mostly larger projects (data centers, commercial)
- Maybe $50-100M TAM?
- Harder to scale beyond MEP
Vision: Massive
- Anyone who needs building plans
- All project types and sizes
- Multi-billion dollar TAM
- Clear expansion path
Technical Difficulty
Wedge: More Tractable
- Input/output formats are known (Revit API, fabrication formats)
- Training data exists (existing conversions)
- Paul can provide domain expertise
- More like your text extraction work
Vision: More Complex
- Need to handle building codes, structural integrity
- Liability concerns from day one
- Multiple output formats required
- Need rendering, costing, validation
Competitive Moat
Wedge: Narrower Moat
- Autodesk could build this themselves
- Or they could acquire you (is that a pro?)
- Other point solutions already exist
- Harder to defend long-term
Vision: Wider Moat
- Network effects (more designs = better AI)
- Outside Autodesk's ecosystem
- Brand advantage if you go first
- Data advantage builds over time
Funding & Growth
Wedge: Bootstrap or Small Seed
- Narrower vision = less VC interest
- Could bootstrap or raise $500k-1M
- Slower growth trajectory
- But: profitable sooner
Vision: VC Fundable
- Big vision = VC interest
- Could raise $2-5M seed
- Higher growth expectations
- But: burn more, profitability later
The Hybrid Strategy: Wedge to Vision
What if you don't have to choose? What if the wedge LEADS to the vision?
Phase 1: Wedge Product (Months 0-12)
Build:
- AI tool that converts Revit MEP objects to fabrication objects and vice versa
- Train on Paul's data center projects + public fabrication catalogs
- Revit plugin or API-based service
- Focus on ductwork first (Paul's domain), expand to piping, electrical
Sell to:
- MEP contractors (like Paul's firm)
- Data center designers
- Large commercial MEP projects
- Pricing: $100-500/project or $2k-5k/year per seat
Why this works:
- Solves Paul's immediate pain point
- Revenue from month 3-6
- Learn BIM data structures deeply
- Build relationships with contractors, architects
- Understand how real projects work
Phase 2: Expand the Wedge (Months 12-24)
The Shift:
- "Instead of converting, what if we could just GENERATE the fabrication objects you need?"
- "Describe what you want in plain English: 'Supply duct, 8-inch, from AHU-1 to Zone A, 150 CFM'"
- AI generates both the Revit object AND the fabrication object
This is the bridge:
- You're still working within Revit ecosystem (safe)
- But you're adding generative capability (disruptive)
- Users see: "Wow, I barely need to use Revit anymore"
Phase 3: The Vision (Months 24-36)
Launch Standalone Platform:
- Natural language → complete building designs
- Target markets that DON'T use Revit: homeowners, small builders, ADU market
- No Revit required - direct to IFC/DXF output
- Use all the learnings from Phase 1-2
Leverage Wedge Success:
- You already understand BIM data structures
- You have relationships with architects/contractors
- You have revenue and credibility
- You have a brand
Attack from Two Sides:
- Bottom-up: New market that never used Revit (ADUs, small projects)
- Top-down: Existing customers from Phase 1 who realize they don't need Revit anymore
Why The Hybrid Works
Phase 1 solves a real problem and generates revenue immediately. You learn the domain, build credibility, establish relationships.
Phase 2 adds generative capability while staying "safe" inside the ecosystem. Users get addicted to natural language design.
Phase 3 is the full disruption, but you approach it with experience, customers, and revenue. Much less risky than starting with the vision.
Analogy: Slack didn't start by saying "we'll replace email." They started as a gaming company's internal tool. Then a chat tool for companies. Then it became obvious email was dying. Wedge → Vision.
The Recommendation
Start with the wedge. Use Paul's insight.
Build the fabrication ↔ Revit AI converter first. It's a real problem Paul faces every day. You can validate it immediately. You'll learn the BIM world deeply. You'll generate revenue within 6 months. You'll build relationships with the exact customers who'll want the bigger vision later.
But - and this is critical - build it with the vision in mind. The architecture should be modular. The AI pipeline should support generation, not just conversion. The brand (Plai) should work for both.
Phase 1: "Plai converts fabrication objects to Revit - AI powered, actually works."
Phase 2: "Plai generates MEP objects from descriptions - faster than modeling."
Phase 3: "Plai designs buildings from plain English - no BIM software required."
Same company. Same vision. Different entry points. Paul's insight just gave you the wedge.
Next Steps
- Validate with Paul: How much would he pay for this tool today? How often does this problem occur? How many hours does it waste?
- Map the data: Get examples of Revit objects and their fabrication equivalents. Understand the transformation.
- Check competition: Are there existing tools? How do they work? Where do they fail?
- Prototype quickly: Can you build a demo in 2 weeks that converts 10 objects? Show Paul.
- Talk to 5 MEP contractors: Validate that this is a widespread problem, not just Paul's firm.
If Paul's insight resonates with other MEP contractors, you have your MVP. If not, pivot to the vision. But start by testing the wedge.