Frequently asked questions.

Blackflake is a proprietary multi-agent orchestration platform — the engine beneath regulated-industry AI products such as Legal AI, MedAssist, and TenderAI.

It is not a wrapper around a foundation model, not a chatbot builder, and not a no-code tool. It hosts domain-specific autonomous agents, workflows, sources of truth, and a verification layer that together produce outputs you can trust in regulated work.

Those are consumer-facing conversational surfaces over a single foundation model. Blackflake is the infrastructure beneath such surfaces: many agents with defined roles and toolsets, a source-of-truth layer, a QA gate that verifies outputs, and a documented methodology for composing new verticals.

A GPT you talk to. Blackflake is what you deploy into production when the output has to be trusted, audited, and owned.

Those are frameworks or general platforms — you assemble your system from primitives. Blackflake is a vertically-built system: the agents, workflows, sources of truth, and verification layer ship already composed for specific regulated industries (law, healthcare, procurement).

You do not write Blackflake agents from scratch. You deploy the vertical and configure it for your firm, matter taxonomy, and jurisdiction.

Traditional professional-services firms bill by the hour. A law firm bills by partner time. A consulting firm bills by associate day. A BPO bills by the seat. Revenue scales with headcount.

Services-as-software packages the same outcome as a permanent, owned system. You license the deployed agents, workflows, and sources of truth. You run them forever. Every new industry, jurisdiction, or matter type compounds on the same engine. Margin shifts from hours-to-revenue to software multiples.

Think: the shift Workday forced on enterprise HR, or Palantir on government analytics — now arriving for regulated professional services.

Short version: you own your stuff; we own the platform. Matches the commercial model of every serious platform firm — Palantir, Veeva, Dataiku. It's honest, audit-proof, and removes "what if the founder disappears" risk.

You own (permanently): your data; your configurations (rubrics, playbooks, house positions, matter taxonomies); the outputs your deployment produces; the documented specification of everything we built for you. These are yours, always, and exportable in standard formats.

Blackflake owns (permanently): the engine, the agent architecture patterns (HACD methodology), cross-customer infrastructure, platform code. Your deployment accesses the engine under your retainer.

If the retainer ends: your data, configurations, and outputs export cleanly. You keep what you've built. Access to the engine ceases; the built-for-you spec remains with you for a future rebuild on any platform.

Customer data never trains cross-customer models. A written data-processing agreement governs every engagement, with a documented processor chain and per-matter isolation.

Built and running: ~55 autonomous agents, 64 workflows, 31 machine-readable sources of truth, 225+ production surfaces across six brands and three locales. Counts observed 2026-04-06.

In production: Legal AI.

In active build: MedAssist (Polish clinical NLP, visit transcription, ICD-10 coding), TenderAI (public-procurement bid drafting), GrantAI (grant-application drafting against published rubrics).

Planned, Q3–Q4 2026: TaxAI (UK RDEC/SME + Polish B+R tax credits — shares rubric-reasoning architecture with GrantAI), ComplyAI (EU regulatory drafting for CSRD, AI Act, NIS2, GDPR ROPA).

Under review, not committed: DueDiligenceAI, ImmigrationAI, PlanningAI, ExportAI, PatentAI.

GrantAI's first POC was a complete FENG Ścieżka SMART application drafted through the platform — we built the full application, ran it through our own rubric-scoring agents, and ultimately chose not to file. The system is what became the product. See the write-up.

Development is self-funded.

Blackflake operates as a data processor under EU GDPR and UK GDPR with:

• A written data-processing agreement governing every deployment.
• A documented processor chain — all sub-processors declared and versioned.
• A Data Protection Impact Assessment (DPIA) for each regulated vertical.
• Per-matter isolation preserving lawyer–client privilege in Legal AI deployments.
• Polish NFZ and RODO-compliant handling for MedAssist clinical data.

Full posture: Privacy Policy · Legal Notice · Security.

Engagements are bespoke and scale with deployment size:

• Scoped pilot — a defined 6–12 week engagement on a single vertical, with an agreed success criterion and an optional conversion to a licence.
• Per-vertical licence — unlimited use within one vertical (Legal AI, MedAssist, or TenderAI) at one deployment.
• Group / enterprise licence — multi-vertical, multi-jurisdiction, group-wide deployments.

Pricing reflects the ownership model — you pay for a deployed system, not for hours. Discuss: enterprise@blackflake.com.

Blackflake is operated by Bennovate sp. z o.o. (KRS 0000597272, NIP PL7272799328, REGON 363700466), a Polish limited-liability company headquartered in Łódź.

The platform was built by a solo-founder architect using the Human-AI Collaborative Development (HACD) methodology — a human architect defines the structure, AI agents build the components, and a QA layer verifies. Development and team scaling are self-funded and sequenced against commercial traction, not against grant milestones.

Email enterprise@blackflake.com with a one-paragraph description of your problem and industry. We respond within three business days with a discovery call or a written "this doesn't fit us yet". No drip sequences. No sales pipelines. No "circling back" from a BDR.

If your question is technical: hello@blackflake.com. If it is media: hello@blackflake.com. Vulnerability disclosure: security@blackflake.com.