Common questions.

Foundation models are thinking tools. We are infrastructure. A salesperson brainstorming objections with Claude is doing something fundamentally different from a customer-facing answer that has to come from your authoritative documents — under permissions, with citations, in a system your compliance team has signed off on. The first is iteration. The second is operations. Your team should keep using the public model for ideation; the moment an answer needs to come from your proprietary data, be auditable, be embedded in a product, or be served at scale — that needs a different system underneath. Not a competitor. A complement. Read: Thinking Tools and Infrastructure →

Enterprise plans solve one problem: your prompts won’t be used to train the vendor’s next model. Necessary, not sufficient. They don’t give you a private corpus index across millions of documents with your permissions and access controls intact. They don’t produce per-page citations every reviewer can verify. They don’t run inside your VPC. And they don’t let you swap models, route different tasks to different models, or keep your data and workflows when the next model leader emerges. An Enterprise plan is a subscription. A private AI platform is infrastructure.

Yes — and several of our customers tried first. The architecture is well-understood; the engineering between “demo” and “system that survives a real engagement” is not. Document ingestion across messy formats, OCR for scanned pages, page-level chunking and re-ranking, permissioning, audit logging, single-tenant isolation, evaluation harnesses, fine-tuning to your firm's voice — each is weeks of work that ships no demo applause and decides whether the system can be trusted on a real deal, brief, or audit. We use the same foundation models you would. We've done the unglamorous engineering tier so you don't have to staff it. Read: Why Enterprises Need a Private AI →

The opposite. Foundation models churn — the leader at one task in January is rarely the leader at the same task in June. A platform built directly against one model is a platform whose roadmap is hostage to a single vendor’s release calendar. We treat the model as a swappable component: route different tasks to different models, swap when something better arrives, and keep your retrieval layer, your data, your access controls, your evaluation harness, and your fine-tuning exactly where they were. The model is the most replaceable part of the stack — by design.

Most engagements move from kickoff to first production analysis in 4–6 weeks, depending on the depth of fine-tuning to your firm’s voice and the integrations required.

A discovery call to scope use cases, sample documents to fine-tune on (prior memos, reports, or reviewer notes), and a designated technical sponsor. Implementation is hands-on from our side.

Yes. Most firms start with one team — diligence, audit, disclosure review, corp-dev — and expand from there. We size engagements to match.

In your environment. Single-tenant deployments in your AWS, Azure, or GCP VPC, or fully on-prem. See Security for the full controls list.

No — never. Every customer runs in a dedicated siloed data room. Nothing is shared, federated, or learned across firms.

Yes — SOC 2 Type II audited annually. ISO 27001, GDPR, and CCPA aligned. Request the audit report.

Yes. Native integrations with leading data room providers, plus secure API ingestion for any source. Documents can also be uploaded directly.

Yes. Fine-tuning on your prior memos, reports, and reviewer notes is included in every engagement. See how it works on the Platform page.

Every answer cites its source at the page level. You’re never asked to trust an answer without being able to verify it. See the citation pattern.

We price by engagement scope rather than seat count, since deployments vary substantially in fine-tuning, integrations, and team size. Talk to sales for a quote tailored to your needs.

Implementation, fine-tuning to your firm’s voice, integration work, and ongoing model updates as your data and frameworks evolve. Pricing covers the platform and the work to make it yours.

Yes. Most engagements start with a paid pilot scoped to a specific project, team, or workflow. Pilots typically run 30–60 days.