Frequently asked questions

You invest in two things: hardware you buy once and own, and a delivery that is billed in gated phases. A phase opens only when the one before it produced working evidence. There are no per-token charges and no per-seat subscriptions for the system we build. If you want reference points on cloud pricing versus owned hardware, the Insights article on capacity and control lays out the published numbers with their dates.

No. The delivery is built so your existing team operates it. Non-technical staff drive the process automation, and every sizing decision is documented and handed to your team. After go-live we provide technical maintenance: model and software upgrades that improve the stack's efficiency, plus new features from our continuously developed products. The system's results are measured on a fixed cadence.

Assessment comes first and sets the plan. The first working version is then built and validated on real hardware before larger capital commits. The exact schedule is part of the roadmap you approve: each phase carries the dates it must hit, and a phase that misses its gate is retired, not quietly extended.

The system runs inside your network, on hardware you own, and keeps running without us. The open-source components, custom backends, and models are yours under commercially friendly licenses, and every decision is documented and handed over. Our agent platform Hefty is deployed on your infrastructure and a source-available license is available, so technology-centric clients can take control of the code or build their own product on it. What ends with us is the maintenance relationship - model and software upgrades and new product features - not the system itself.

For document review, triage, and retrieval over your own holdings, that question is answered by measurement, not opinion. The validation phase runs your real workload and scores accuracy against your own ground truth, in your language and terminology, before you commit capital. You fund the build only if the numbers hold, and quality keeps being measured after go-live.

For regulated workloads we recommend infrastructure you own: dedicated capacity, answers that stay inside your network, and costs that do not reset every month. We are pragmatic about the rest. The stack we build also runs cloud models where your organization needs them - fully local, cloud, or mixed, switchable at any time - and the gateway records every call either way. The sizing recommendation comes from measurements on your own workload, in our GPU lab, before you buy.

Rollout is part of the delivery. Staff Workshops put your teams in the room where the plan is written. Process Transformation rebuilds the surrounding workflows so the system lands where it returns value. After go-live, adoption is one of the four metrics we track - alongside accuracy, cycle time, and avoided cost - so spread is measured, not assumed.

Yes. The platforms we deploy run local models by default and connect to cloud providers where you choose to - fully local, cloud, or mixed, switchable at any time. If your data cannot leave, the same controls keep everything inside your network. If a cloud model is the right tool for a task, it is one configuration away - and every call it carries is recorded.