LegalFab
Proposal · Confidential
Prepared for GSK Stockmann

The AgenticLaw Firm.

Die Kanzlei der Zukunft
Your expertise is the firm's greatest asset. We propose to make it work for you.
GSK Stockmann  ·  250+ professionals  ·  Germany · Luxembourg · UK
Scroll
Executive summary

Turn the firm's expertise into
a compounding asset.

GSK Stockmann's greatest asset is the judgment of its 250 professionals — today trapped in siloed systems and the heads of the partners who built it. LegalFab proposes to build the data foundation that makes that expertise searchable, reusable and firm-wide — in 25 weeks, without migrating a single byte — starting with the use case that pays back fastest: business development.

The problem
Client data split across STP, NetDocs, HighQ and email — no single source of truth, and expertise that walks out when a partner does.
The approach
A metadata-driven Knowledge Fabric connects your systems where they stand — no migration — resolves every client to one golden record, and answers questions in plain language.
The outcome
One knowledge graph, natural-language access for every lawyer, and a Studio where your teams turn their own methods into governed, auditable agents.
The commitment
A phased 25-week engagement — every stage signed off against acceptance criteria, at the level of human oversight you choose.
25 wks
from contract to an agentic firm — phased & gated
8 stages
each with a signed-off deliverable and acceptance test
0
bytes of source data migrated — it stays in place
About LegalFab

The operating system for
the agentic law firm.

LegalFab is an AI-native platform that unifies data, workflows and decision-making across the firm — so legal work is carried out by intelligent, adaptive agents rather than fragmented manual processes. It spans the entire legal lifecycle:

01
Market analysis
02
Business development
03
Onboarding
04
Risk & compliance
05
Matter execution
06
Billing
07
Post-engagement intelligence

This engagement lays the foundation and proves it on one stage of that lifecycle — business development — with every other stage available to follow, on the same platform.

01 — The asset you can't yet see

Every evening, your best asset
walks out the door.

A law firm's real value was never in its software. It lives in the judgment of its people — three decades of structuring funds, closing real estate, navigating capital markets — and in the matter files that hold how it was actually done.

But that knowledge is scattered. When a partner retires, their method leaves with them. When a matter closes, its lessons disperse into folders no one opens twice. The very thing that makes GSK Stockmann GSK Stockmann is the one thing the firm cannot yet search.

02 — The hidden tax

The most expensive hour is
the one spent looking.

A single client appears in STP under one name, in Outlook under another, in a HighQ workspace under a third. To answer one simple question — what is our full exposure to this client? — someone opens four systems and reconciles them by hand.

Now multiply that by every conflict check, every pitch, every matter. It isn't a tooling gap. It's a tax the firm pays every day — in billable hours, in avoidable error, in the cross-sell no one saw.

4+
systems to open for one complete client view
0
shared source of truth across the firm today
Hours
to answer questions that should take seconds
STP"Müller ImmobilienGmbH" NetDocsdocuments,mid-migration HighQ"MÜLLERIMMOBILIEN" Outlook"Mueller Immo." PowerBIa fourthversion again ONE CLIENT · FOUR SYSTEMS · FOUR NAMES · RECONCILED BY HAND
Each system excellent on its own — and an island. The lawyer is the only bridge between them.
03 — The real diagnosis

It isn't the missing CRM.
It's the plumbing.

It's tempting to blame the absent CRM, or the NetDocs migration still underway. Those are symptoms.

The root cause sits beneath them: GSK does not lack information — it cannot connect the information it already owns. STP, NetDocs, HighQ, SharePoint, PowerBI — each excellent, each an island. Until the plumbing joins them, no dashboard, no new hire, and certainly no AI can give the firm a single, trustworthy view.

Fix the foundation first, and everything above it becomes possible. Skip it, and every tool you add just automates the mess faster.

AI the firm can actually trust Dialog · Studio · Agents Resolved knowledge graph — one source of truth The Knowledge Fabric — the data foundationTHE LAYER EVERYONE SKIPS STPNetDocsHighQMicrosoftregistries EVERYTHING ABOVE RESTS ON THE LAYER BELOW
Fix the foundation, and every layer above it becomes possible.
Agentic AI is only as good as the data beneath it. Most firms are about to automate on sand.
The foundation is the strategy
04 — The turn

Now imagine the firm
knew what it knows.

One place to ask anything — and receive an answer drawn from every system, with its source attached.

The same client resolved to one golden record across every office. A junior who can question thirty years of precedent in plain language. A partner's method for structuring a fund, captured once and reused by the whole practice.

This is the agentic law firm: not lawyers in service of the software — software in service of the lawyers.

05 — How it works

Three moves. Not ten features.

1

Unify — without moving a thing.

The Knowledge Fabric connects to your systems exactly where they stand. No migration. No duplication. Source data never leaves its home. What used to take months of bespoke integration takes days — and the "invisible data" trapped by legacy systems and growth becomes an instant, usable asset.

Knowledge Fabric METADATA-DRIVEN · NO MIGRATION STP⌂ stays in place NetDocs⌂ stays in place HighQ⌂ stays in place SharePoint⌂ stays in place PowerBI⌂ stays in place Registriesvia OSINT
Connect where they stand. The fabric reads through metadata — the source data never moves.
2

One client. One truth. Ask anything.

Entity resolution stitches the same client, matter and counterparty across every source into a single golden record — matching partial names, reconciling conflicting details. Then Dialog lets anyone ask, in German or English, in plain language — and every answer comes back with full provenance you can defend.

STP"Müller Immobilien GmbH" Outlook"Mueller Immo." HighQ"MÜLLER IMMOBILIEN" Müller Immobilien GmbH GOLDEN RECORD · one client · full provenance
3

Turn expertise into assets.

In Studio, your experts capture their own methods as agents — reusable, governed, auditable, and bound to the firm's schema so they never invent. The secret sauce stops walking out the door at 7pm. It compounds. And what a firm can package, a firm can eventually license — a new line of value from expertise you already own.

06 — Under the hood

Four parts. One coherent platform.

Enough of the architecture to know it's real — without the spec sheet.

Knowledge Fabric
The foundation
  • Persistent knowledge graph — schema-bounded extraction with full source provenance
  • Entity resolution with golden-record management across every system
  • MCP connectors — federated queries across databases, SaaS, legal systems and registries
  • Data observability — quality monitoring, freshness tracking, automated alerts
Dialog
The surface people talk to
  • Playbook-based — tenant-scoped, version-controlled, with guardrails and persona
  • Context-aware every turn — user, role, active matter, client, conversation history
  • Answers with traceable lineage across the fabric — nothing unsourced
  • Behaviour gated by the operational mode you've set
Studio
Where expertise becomes assets
  • Build Data-Driven Agents, widgets, datasets and utilities
  • Chain of agents — multi-step workflows with human-in-the-loop review gates
  • Business-domain discovery — extract schemas straight from your documents
  • Natural-language to working agent, via AI hybrid planning
AI & LLM Layer
Governed intelligence
  • Provider-agnostic gateway — no model lock-in, swap as the field moves
  • Schema-validated output — the structural guard against hallucination
  • Model provenance — every answer's model version tracked
  • Feedback loops, accuracy monitoring and reasoning-chain transparency

Knowledge Fabric

  • Persistent knowledge graph
  • Entity resolution & golden records
  • MCP connectors — federated
  • Two-way data flow — read & write-back
  • Search sessions — accumulated context
  • Data observability

Studio

  • Data-Driven Agents (DDAs)
  • Widget types — SYSTEM & OUTPUT
  • Datasets — schema-enforced
  • Utilities — APIs + agents
  • Chain of agents
  • Business-domain discovery
  • MCP integrations
  • Asset search
  • Text-to-Pipeline
  • AI hybrid planning

Dialog

  • Playbook-based communication
  • Context-awareness
  • Tool-plan execution
  • Widget rendering — dialog & canvas
  • System self-awareness
  • Memory policy — tenant-isolated
  • Operational-mode binding

AI & LLM Layer

  • Provider-agnostic gateway
  • Output consistency — schema-validated
  • Model provenance
  • Quality assurance
  • Reasoning-chain transparency
07 — Human oversight, by design

The firm is never automated
faster than it's comfortable.

Five operational modes, set per workflow and changed as trust grows. You are always in control of how much the agents decide.

0
Traditional
No agents. Manual investigation and analysis.
1
AI-assisted
Agents suggest only. Every decision needs human approval.
2
Routine automation
Agents handle the routine. Humans focus on judgement.
3
Autonomous + escalation
Full automation, with escalation on the exceptions.
4
Automated + audit
End-to-end, with post-hoc human audit.
0 1 2 3 4 MORE HUMAN CONTROL MORE AUTOMATION
You set the point on this line — per workflow, and move it only as trust is earned.
08 — Secure by architecture

Built for a firm that lives on
confidentiality.

Data stays in place
Metadata-only architecture. Source data is never migrated, copied or moved out of your systems.
Agents can't invent
Schema-bounded extraction: every answer is constrained to your defined structures and traces to its source.
Encryption everywhere
AES-256 at rest, TLS 1.3 in transit — across every component.
Audit everything
Tamper-evident logging with full provenance on every retrieval and action.
Least privilege, zero trust
Every request authenticated and continuously verified, regardless of source.
Built for legal
Matter confidentiality, conflict checking and privilege protection are first-class, not add-ons.
Compliant by design
GDPR, SOC 2 and the money-laundering and industry standards your practice already meets.
You hold the keys
Your credentials, your identity provider, your encryption keys — in every deployment mode.
Deploy your way
From managed cloud to fully in-boundary — the isolation level is your decision, not ours.
Designed to support
GDPR
CCPA / CPRA
Money Laundering Regulations 2017
SOC 2
FCA requirements
Matter confidentiality
Conflict checking
Privilege protection
YOUR BOUNDARY — DATA STAYS HERE Your systems Documents Matter files KnowledgeFabric + Graph Dialogin-boundary LegalFab services control · updates only metadata /governed only ⊘ raw datanever crosses
Source data never leaves your boundary. Only governed, metadata-level queries cross the line.
09 — We've done the homework

This is built around your firm —
not a template.

Your landscape, as we mapped it
STPSystem of record — time, matters, client contacts
NetDocsDocument management migration in flight
HighQLegal collaboration & matter workspaces
MicrosoftOutlook, SharePoint, PowerBI
PerBitHR & talent
Where it hurts
No CRMNo dedicated system for client relationships
No 360°Client detail split across STP, email, folders
Cross-sellOpportunities invisible across departments
KnowledgePrecedent & method locked in people and files
OnboardingEvery new matter re-solves what the firm knew

We would rather validate against your real systems from day one than against synthetic tests. That principle shapes the entire engagement below.

10 — Where we start

The first use case:
business development.

Given the gaps above — no CRM, no 360° client view, cross-sell invisible — business development is where the foundation pays back fastest. So it is the first use case we scope, and the one the first five agents serve.

The 360° client view
One brief per client, assembled across STP, NetDocs, HighQ, email and registries — resolved to a single record.
Cross-sell, made visible
Which clients rely on one practice and have never been introduced to another.
Warm paths in
Who inside the firm already holds a relationship with a target company or contact.
Every pitch remembered
What was proposed, to whom, and how it landed — so the firm stops re-learning its own history.
Prospect intelligence
OSINT on a target — corporate registries and public sources — folded into a ready BD brief.
Relationship health
Where activity is quietly declining — surfaced before the client notices first.

This use case shapes the engagement: the schema is defined around business-development entities, Dialog's playbooks answer these questions, and the five agentic assets built in Phase 2 are these tools.

11 — The journey

From contract to an agentic firm
in 25 weeks.

A phased path with a tangible win early — natural-language answers over your connected data long before the finish line.

Phase 0

Wk 1

Analysis & alignment

We install and localize the platform, and sit with your domain experts to agree the schema — the contract that binds every downstream step to your reality.

Phase 1 · Foundation

Wk 2–20

The Knowledge Fabric, live

Schema definition → MCP connectors to your sources → OSINT enrichment → automated discovery & cataloguing → the knowledge graph with entity resolution → Dialog for natural-language access → Studio, so your teams can build. Each stage ships a signed-off deliverable; the firm sees value accrue, not a black box.

Phase 2 · Agents

Wk 20–25

Five agents in production

We capture a business domain with your experts and stand up five operational agentic assets — the first of the firm's expertise, made reusable, governed and auditable.

PH 0PHASE 1 · FOUNDATIONPHASE 2 · AGENTS AnalysisInstall & configSchemaConnectorsOSINTDiscoveryGraph & ERDialogStudioAgents 1223134225 wk 0510152025 weeks ▲ Phase 1 accepted Go-live ▲
Sequential, gated, and transparent — with a working, queryable foundation well before the finish line.
12 — The engagement, step by step

Every stage ships a signed-off deliverable.

No black box. Each workstream has a concrete deliverable and an acceptance test you approve before we move on.

Phase 1 · Foundation — weeks 2–20
Schema definition
2 weeks
Working sessions with your domain experts. Using Business-Domain Discovery we extract schemas from your documents and define the core entities, attributes and relationships — the contract that binds all downstream extraction.
DeliverableApproved domain schema registered in the Schema Registry.
AcceptanceDomain experts sign off coverage of the use case; validated against sample documents.
MCP connector integration
3 weeks
We connect the selected sources from the connector library, build the sync flows, calibrate sync rates per source, and test schema-bounded extractors — all using your credentials via the vault. Source data stays in place.
DeliverableSelected sources connected with configured sync flows and validated extractors.
AcceptanceEach connector authenticates and syncs at the agreed rate; extractors produce schema-conformant output.
OSINT configuration
1 week
Configure the external-intelligence processes that enrich internal data: schemas for external entities and the search domains relevant to your practice — corporate registries, regulatory registers and public sources.
DeliverableConfigured OSINT schemas and search domains.
AcceptanceOSINT queries return schema-conformant results from the agreed domains for a set of test entities.
Discovery
3 weeks
The Discovery service runs across all connected sources, automatically identifying and cataloguing assets. Active metadata profiling classifies content against the domain schema.
DeliverableFull data catalogue — every source mapped to schema, with quality and freshness metrics.
AcceptanceAll sources appear in the catalogue; mapping coverage and quality metrics reviewed and accepted.
Knowledge graph & entity resolution
4 weeks
Construction of the central knowledge graph. The resolution engine — blocking, matching, clustering — is calibrated to your data, so the same client, matter or counterparty is recognised across every system.
DeliverablePopulated knowledge graph with operating entity resolution and golden records.
AcceptanceEntities resolve at agreed precision; multi-hop queries (client → matter → counterparty) return correct, provenance-backed answers.
Dialog
2 weeks
Connect the conversational layer. We prepare the playbooks for your use case and configure routing, so questions flow to the right destination — graph retrieval, agentic flows or OSINT.
DeliverableDialog live with use-case playbooks and query routing configured.
AcceptanceAn agreed set of representative business questions is answered correctly, with correct routing across paths.
Studio integration
2 weeks
Connect the agentic layer to Studio so your teams can build their own agents, widgets and workflows. Operational modes (0–4) are set to your oversight level, and your team is enabled on the agent-creation flow.
DeliverableStudio connected to the Knowledge Fabric; your team enabled to build agents.
AcceptanceYour team independently builds and runs a working agent against the graph in the agreed mode.
Phase 2 · Agents — weeks 20–25
Agentic assets creation
5 weeks
We capture a business domain with your experts and build up to five operational agentic assets — agents, widgets and chains of agents — the business-development tools above, and the first of the firm's expertise made reusable and governed.
DeliverableA captured business domain and up to five operational agentic assets.
AcceptanceUp to five agentic assets configured and working in the system.
Indicative delivery plan

Deliverable, value and time frame.

Step
Deliverable
Value
Time Frame
1
Schemas
A stable, extensible knowledge model that aligns the entire firm and constrains all downstream AI behaviour.Schemas give GSK Stockmann a shared, authoritative vocabulary. They eliminate semantic inconsistency across systems and create the foundation for all intelligence, automation, and agentic capabilities.
  • Establish schemas that define the firm’s knowledge model
  • Creates a canonical vocabulary across systems
  • Ensures precision and bounded extraction, preventing drift
  • Enables schema versioning, allowing evolution without reprocessing
2 – 4 Weeks
2
Connectors / MCPs
A unified, continuously updated view of the firm’s data landscape without duplication or migration.Connectors give the firm real-time, secure access to all relevant data without moving it, enabling unified intelligence. Provide deterministic, auditable access
  • Enable incremental syncs
  • Allow source-level control
8 Weeks
3
Discovery
A complete, navigable map of the firm’s data assets; the prerequisite for intelligence, automation, and agents.Discovery creates a complete, structured catalogue of all data across systems, showing what exists, where it lives, and how it connects.
  • Automated discovery of entity types, relationships, attributes
  • Produces a baseline schema proposal grounded in real data
  • Creates a clear understanding of what data the firm has and where it lives
  • Catalogues data across key operational systems
2 – 3 Weeks
4
Knowledge Graph
A unified, auditable, intelligence-ready representation of the firm’s entire knowledge base.The PKG transforms raw data into structured intelligence, enabling search, reasoning and agentic workflows.
  • Extracts entities, relationships, keywords
  • Provides granular provenance
  • Performs entity resolution
  • Enables contextual analysis
4 Weeks (depends on data volumes)
5
Dialogue System
A secure, context-aware conversational interface that allows users to interact with the firm’s knowledge, data, and workflows through natural language.The Dialogue System transforms complex data and processes into intuitive, conversational interactions, enabling rapid insight discovery and operational efficiency.
  • Enables natural-language querying across structured and unstructured data
  • Interprets user intent and retrieves contextually relevant information
  • Provides auditable conversational history for compliance and review
  • Supports multi-turn reasoning, allowing deeper exploration of issues
  • Reduces friction in accessing firm knowledge, improving adoption and productivity
2 Weeks
6
Agentic Studio
A no-code/low-code environment for designing, orchestrating, and deploying intelligent agents and workflows.Agentic Studio empowers the firm to build automation and intelligence without engineering overhead, accelerating innovation and reducing operational bottlenecks.
  • Allows rapid configuration of agents, workflows, and decision logic
  • Provides visual tools for mapping processes, triggers, and data flows
  • Ensures governance through versioning, auditability, and role-based access
  • Enables experimentation and iteration during the POC without code changes
  • Bridges business teams and technical teams, reducing dependency on developers
2 Weeks
7
Use Case specific agents / Utility
Targeted agents designed to solve the firm’s highest-value problems using its own data, workflows, and domain knowledge.These agents deliver measurable business impact by automating tasks, enhancing decision-making, and reducing manual effort.
  • Automate repetitive or labour-intensive workflows (e.g., document review, case triage, compliance checks)
  • Provide domain-specific reasoning aligned with the firm’s policies and expertise
  • Deliver consistent outputs, reducing human error and variability
  • Integrate with firm systems to trigger actions, generate reports, or update records
  • Demonstrate real-world value early in the POC, validating ROI and scalability
TBD but will run in parallel to some steps
What we'll need from you

We validate against your real data —
from day one.

The systems in scope
A real example of every system we integrate with:
  • System specification and version
  • Integration documentation (APIs, files)
  • Business requirements
  • Technical considerations & limitations
  • A test environment with credentials
  • Data sets in scope, plus sample data sets
Your domain experts
Time with the people whose judgment defines the practice:
  • Schema-definition working sessions
  • Business-development domain capture
  • Sign-off at each acceptance gate
Your environment
So deployment fits how GSK already works:
  • Cloud strategy
  • Data strategy
  • IT-environment assessment (OS versions, etc.)
13 — What you'll have

At the end, the firm owns
a compounding asset.

A complete data catalogueEvery connected source mapped to your schema, with quality and freshness you can trust.
One knowledge graphCross-source entity resolution — the same client, matter and counterparty, resolved once.
Natural-language accessEvery lawyer able to ask complex, multi-hop questions and get provenance-backed answers.
A studio of your ownYour teams building their own agents on a trusted layer — no longer dependent on a vendor for every change.
Governed by designMatter confidentiality, conflict checking, privilege protection and full audit — the standards your practice already lives by.
A foundation that lastsModels change; frameworks change. Your resolved knowledge and your agents carry across all of it.
14 — Commercials

Priced as what it is —
an asset, not a licence.

A phased build and a monthly platform subscription, scaled to the sources and scope agreed at design. Full commercial terms are provided under separate cover, alongside the outcome of scoping.

15 — Investment

Build, then platform.

Item
Basis
Amount
Build
Over the duration of the 6-month project
€20,000 / month
Monthly licence fee
Platform subscription
€20,833 / month
What’s included
  • Unlimited users
  • Knowledge Fabric — full resolve-in-place graph
  • Agentic Studio — build firm-owned agents
  • Security, audit & governance
  • Standard support
Why now

The firms that build the foundation now will compound.

AI will not wait, and neither will the talent that expects to work this way. Lay the foundation this year and every future advance becomes leverage. GSK Stockmann has the expertise worth capturing. The only question is whether it stays trapped — or becomes the thing that sets the firm apart.

Let's build the agentic law firm — together