Built from inside the verticals, not from a deck.
This is how we enter new industries, build the software in step with how the business actually runs and deploy AI inside our portfolio companies. It is written from the perspective of an operator first and an investor second, because that is the sequence in which the conclusions were earned.
The thesis, in brief.
The durable moat in Vertical AI is the operator knowledge captured inside the customer's organization — structured, controlled, and made searchable — not the model, the software that connects it, or how deeply it's wired into the workflow.
Everything else commoditizes on a twelve-month cycle. Captured knowledge is the one thing that compounds.
The operating principles distilled from our seven-section position paper, that we test every vertical, every module, and every dollar of capital against.
Operator first. Software second.
No engineering team, regardless of pedigree, can produce industry fluency from observation. It has to be earned from participation. We build software for our own teams before we build it for the market.
Unstructured inbound and judgment-intensive workflows.
SmartInbox™ — our AI-native triage layer — is the clearest expression of the thesis in production. The same module now handles private charter requests, window installations and equipment RFQs across the portfolio.
Captured context is the durable asset.
Before deploying a single AI agent, we capture how the company actually operates — its workflows, decisions, and hard-won judgment — in a structured, access-controlled repository organized the way the company is: team by team, function by function. That captured context, not the model, is what compounds.
From operator-driven to fully agentic.
We position every portfolio company on a single arc. Era 1: tribal knowledge. Era 2: platform plus AI-assisted workflows. Era 3: agentic systems executing end-to-end within defined guardrails.
The model is not the moat.
Frontier AI models are converging — the leading ones are increasingly interchangeable. The orchestration layer that wires them into workflows gets commoditized too. What actually makes two companies running the same AI stack produce radically different results is the depth and structure of the captured context each one can draw on.
The platform multiplies the asset.
A single-vertical AI company builds one vertical's worth of context. We bring five. The cost savings from a shared platform are real — but the bigger edge is that every vertical's learning makes the others smarter.
Three eras, one trajectory. We position every portfolio company on the same line.
Operator-driven systems.
Core workflows executed manually by experienced operators. Critical business logic lives in people, not systems. Spreadsheets for reconciliation, QA, and edge cases. High dependency on tribal knowledge. The state every underserved industry is still in.
Platform plus AI-assisted workflows.
NXTOrder™ centralizes CMS, CRM, workflows, pricing, and communications. SmartInbox™ structures and acts on inbound demand. Eight AI features sit in production across the portfolio. The proprietary data being generated in Era 2 is what makes Era 3 possible.
Agentic systems.
AI evolves from assisting workflows to executing them end-to-end within defined guardrails. Agents ingest demand, interpret context, make decisions, and execute actions. Humans shift to exception handling and approval. First build underway: the four-agent Agentic Mesh inside Levo.