— 01
Discovery
Scoped conversation. We map your workflows, the repetitive professional-services work eating your margin, and the regulatory constraints the output has to survive. You leave with a written architecture note and an engagement estimate.
Engine · v1.0 · 2026 · Łódź, PL
Custom AI systems. Built for you. Developed with you.
We design, build, and operate bespoke AI agent systems for law firms, clinics, and companies bidding on public tenders. Your data, configurations, and outputs are yours — always portable. The Blackflake engine runs under your retainer and evolves under our hands.
Legal AI · live MedAssist · in build TenderAI · in build GrantAI, TaxAI, ComplyAI · roadmap
What we sell
Every engagement is custom. We don't sell shelf software and we don't sell seats. We build agents, workflows, sources of truth, and verification gates shaped to your firm — then we operate them with you under a monthly retainer. Your data, configurations, and outputs are yours and portable. The Blackflake engine runs under your retainer and evolves under our hands.
— 01
Scoped conversation. We map your workflows, the repetitive professional-services work eating your margin, and the regulatory constraints the output has to survive. You leave with a written architecture note and an engagement estimate.
— 02
Fixed-scope first deployment on one narrow workflow — a single matter type, one bid category, one clinical specialism. You see agent output on real work in weeks, not quarters. Go / no-go before the full build.
— 03
Pilot expands into the full agent suite for your vertical: domain agents, workflows, machine-readable sources of truth, QA gates, and the operator surface your team actually uses. Your data stays in your tenant. Outputs are auditable and exportable.
— 04
Monthly retainer keeps the system sharp: new regulations ingested, agents tuned on live work, new workflows added as your use expands. We develop the engine; you keep your data, configurations, and outputs — portable, always.
Who we build for
— Law firms
Shaped to your firm's practice areas, house positions, and matter taxonomy.
— Clinics & hospitals (PL)
Polish clinical NLP, RODO-compliant processing, fully in-country.
— Public procurement bidders
Shaped to your firm's bid playbook and win rates.
— Grant applicants
Evaluator simulation before submission. Reusable grant library.
— Cross-border tax (UK + PL)
Corporation tax, VAT (MTD + JPK_V7/KSeF), transfer pricing, and R&D credits. HMRC and KAS rubrics resolved in one position — not two disconnected filings.
— EU mandatory-disclosure
Structured against the evolving EU regulatory templates. Living reports, not one-offs.
Start here
A one-paragraph email. A discovery call within three business days. Or a written "this doesn't fit us yet."
enterprise@blackflake.comThe engine underneath
Blackflake is the platform we build every customer deployment on. Multi-agent orchestration, domain-specific reasoning, and a verification layer so outputs can be trusted in regulated work. Not a prompt. Not an add-on. A platform designed for the work that matters.
— Matter MVP-2026-0142
— Composite risk
7.2/10
Medium · Action needed
— Priority findings
"The Indemnifying Party shall indemnify, defend, and hold harmless the Indemnified Party from and against any and all losses, damages, liabilities, claims, actions, and expenses of any kind, without limitation, arising from or related to the performance of this Agreement…"
Uncapped liability — propose §4.2A cap languageProducts
Blackflake is the shared engine. Each vertical is a precise application of that engine against the published rubric of one regulated, margin-rich industry. Same agents, same verification layer, different domain context per product.
Pillar 01
LiveCase analysis, contract risk, and matter classification. Built for solicitors and law firms where every clause is load-bearing.
Pillar 02
In buildPolish clinical NLP. Visit transcription, ICD-10 coding, drug interaction checks. Designed for the Polish healthcare system.
Pillar 03
In buildBid discovery, eligibility matching, and draft generation for public procurement. The highest-margin AI vertical we touch.
Pillar 04
In buildGrant-application drafting against published rubrics. Eligibility scoring, section composition, coherence QA, evaluator-simulation before submission. First POC: our own FENG application.
Pillar 05
PlannedCross-border tax platform for UK + PL companies. Corporation tax, VAT (MTD + JPK_V7/KSeF), transfer pricing, and R&D credits (RDEC/SME + ulga B+R) — against HMRC and KAS rubrics on one engine. Same rubric-reasoning layer as GrantAI, widened to tax codes.
Pillar 06
PlannedRegulatory document drafting against published EU frameworks — CSRD sustainability reports, AI Act model documentation, NIS2 incident records. Ingests operational data, produces the report your board and regulator actually want.
Further verticals under review
Directional only · not committed pillars · enquire: hello@blackflake.com
Built, not promised
A human architect at the centre. Agents orbiting. Skills, sources of truth, models, and external systems forming the substrate. Partial visibility by design — some things don't belong on a public surface.
— Ecosystem schematic · representative topology · 68 nodes shown · 26 labeled · partial visibility by design · hover a node to trace its connections
Autonomous agents
— Each with a defined role and toolset
Workflows
— Invocable as slash commands
Sources of truth
— Machine-readable, auto-propagated
Live surfaces
— Across six brands and three locales
Services, as software
Traditional consulting rents you expertise. Services-as-software sells you the expertise as a permanent asset. You buy the engine once, it runs forever, and every industry, market, or regulated workflow compounds on the same foundation.
Consulting bills by the hour. We build the hour away.
— 01
New industry online in weeks, not quarters.
Each vertical inherits the orchestration, verification, and source-of-truth layer already running. The work is contextualisation, not rebuild.
— 02
New market live in days, not months.
Legal, regulatory, and language adaptation layers are parameterised. Deploying to a new jurisdiction is configuration, not a project.
— 03
Agents verify agents. Scale without headcount.
A dedicated QA layer audits every output, flags confidence decay, and prevents silent drift. Verification becomes infrastructure, not a hiring plan.
— 04
Own the system. Don't rent it.
The agents, the workflows, and the sources of truth belong to you. No per-seat meter on reasoning. No vendor lock-in on the thinking layer.