Investors
The preparation layer.
A configurable workflow engine that prepares repetitive knowledge work, deployed on hardware dedicated to each client, with inference running locally.
The thesis
Models are commoditizing. What businesses actually run on is not a model — it is the workflow around it.
The durable asset is not access to intelligence. It is the accumulated, tested knowledge of how specific business work is really performed: the triggers, the approved sources, the validation rules, the exceptions, the human checkpoints, and the integration patterns. Every deployment adds to that library, and every addition makes the next similar deployment faster.
Our second position is that a meaningful segment of the market cannot use hosted AI at all — not because of preference, but because of what their data is. Running inference on hardware dedicated to a single client removes the objection that stops those businesses from automating anything.
Where we actually are
Pre-revenue. Not pre-product.
Several deployments are running today across distinct domains — claims work, document intake, receipt handling, reconciliation, provisioning and monitoring — each on its own dedicated hardware. All of them are pilots. None of them are paying yet. We are finalizing the workflows and the business processes around them before converting.
We will not present projections as traction, and there are no customer logos, retention curves or revenue figures on this page, because there are none to present. What there is: a working system, running workflows end to end, that we can demonstrate on request, and references who will speak to their experience.
What is worth diligencing
- Whether the workflow engine genuinely reduces the cost of the next deployment, or whether each one is bespoke work wearing a platform costume.
- Whether local inference is a real buying trigger for the target segment, or a feature buyers admire and do not pay for.
- The unit economics of dedicated hardware per client, and what that implies for the floor on deal size.
- Whether pilots convert, and at what price.
Those are the four questions we ask ourselves. We would rather discuss them directly than be asked them later.
Get in touch