The problem
The value of a data platform is not in the analysis, it is in the ontology. That is the lesson of Palantir: its strength is not computation, it is turning raw tables into business objects (an asset, an engagement, a supplier, a site) linked together, which teams and agents act on directly.
Today this layer is reserved for very large organisations. General-purpose platforms can build ontologies, but rarely ship a ready-to-use domain model: it is up to the client to build it. An expert firm has neither the budget of a Foundry nor the team to model its business from scratch.
What ArborSpace is
ArborSpace is the memory of the firm: an operational ontology for companies that sell expertise (due diligence, technical consulting, commodity trading).
It is neither a chatbot over your documents nor a better search. ArborSpace captures what your experts know, in a form the system can apply and your teams can review.
How it works
The unit of value is the rule: a piece of expert knowledge written down explicitly, which the system applies and anyone can review.
- Define. The firm declares its vocabulary in the ontology: asset classes, engagements, stakeholders and their relationships.
- The system works. Rules run against the data, and a similarity engine matches each case with the firm’s previous ones.
- People decide. Every conclusion arrives as a finding in a review queue. The expert approves or rejects it, and that decision feeds the memory.
Example: a vibration above 8.2 mm/s on a class 7F gas turbine generates a finding. That finding is linked to the asset, the site, the engagement and the lender concerned, along with the rule that produced it.
What sets it apart
- A domain model, not a blank page. ArborSpace ships with a first model of the domain (energy assets, due diligence engagements) that the firm adapts rather than builds.
- Readable rules. Rules are declarative and inspectable: every conclusion traces back to a named rule and an explicit path of relationships.
- On your premises. Deployed in the client’s environment, on site or in its own cloud, with code the client can audit.
- Sized for a firm. A pilot is measured in weeks, not in a transformation programme.
Next steps
We propose a six-week pilot on a narrow scope of the firm. The first two weeks are used to write the first fifteen rules, starting from three questions put to your experts. The following four apply them to your own engagements.
The question to answer at the end of the pilot is simple: is the captured memory worth extending to the rest of the firm?