Palantir for Small Business

Palantir's core product is an ontology: a model of the objects a business runs on and how they connect, so software can reason over the business instead of looking strings up. Enterprises pay for that layer because their truth is scattered across systems that never agreed on what a "customer" is.

A solo contractor has the same problem with worse odds. The objects exist, but they live as texts, forwarded threads, a job-site photo, a handwritten estimate on a clipboard, a quote half-remembered from last Tuesday. The enterprise at least built records to reconcile; the contractor never built any. There is nothing to integrate with, only the raw work to model. Before AI can answer "who hasn't paid," it has to construct the object model from the mess.

I have been building toward this. For local service work the objects hold across trades: customers, properties, jobs, leads, quotes, invoices, payments, reminders, reviews. A contractor texts a number "finished the attic, $1,800, Venmo" and the system resolves the customer, opens the job, writes the paid invoice, and ties the three together. Defining the schema is the easy half. Parsing that one line into it without inventing what it cannot read is the work.

The model has to act, not just file. It draws a PDF invoice off a job, OCRs a crumpled estimate against the right record, and queues the payment chase when an invoice goes out. Keep the reasoning and the state apart: the LLM reads the mess, the objects stay durable and inspectable, because a business you query next month cannot live in whatever fit in the last prompt.

One trap remains. An ontology designed up front grows fields nobody fills and statuses nobody sets. Let the objects earn their place by appearing in what contractors already text, and let the rest wait. More on what I am building soon.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

;)