Workflow and actors
The knowledge team selects the corpus, taxonomy and gold set; models classify, extract and retrieve content. Lawyers, consultants and matter owners review outputs before using them in professional work.
Service · Custom AI · Professional services
Custom AI for professional services that learns from your firm.
Document classifiers, clause extractors, private RAG over your know-how and case-scoring models, trained on your matters, not on a generic internet dataset.
The workflows, examples and figures on this page are illustrative composites and modelled targets, not measured client results. In a real project, we define the baseline, thresholds and human review with your data before rollout.
What we solve
Generic ChatGPT doesn’t understand your kind of contract.
A law firm, a consultancy or an engineering practice has its own taxonomy: clause types, proposal templates, client-specific jargon and success criteria that don’t live in any off-the-shelf model. Asking a generic LLM to classify your opinions or estimate project cost from a brief ends in answers that look right but aren’t reliable, and weekly manual corrections.
We build models trained on your history: document classification by type, critical-clause extraction, case- and project-success scoring, and a private RAG that answers from your work product, not from general knowledge. Evaluated against a gold set signed off by partners before going live.
Implementation contract
How it runs in production
Systems and data
SharePoint, Microsoft 365 or the DMS supplies documents and permissions; the matter manager supplies client, engagement, hours and metadata needed for context, evaluation and traceability.
Exceptions and risks
New document types, ambiguous clauses, low confidence, permission conflicts, corpus drift and bias by client type or jurisdiction are blocked or escalated; client indexes remain separated.
Human review
The responsible professional approves interpretations, proposals and sensitive predictions; partners validate the gold set and thresholds, and security resolves any access conflict.
Implementation pattern
Private permission-filtered retrieval, versioned models, recurring evaluation against the signed-off gold set, and source, confidence, version and approval logging for every output.
Relevant integration
SharePoint or the DMS connects to Microsoft 365, the matter manager and billing; the assistant inherits existing identities and permissions on every query.
What we build for this sector
Use cases that ship to production.
Classification
Document classifier by type and matter
Contracts, opinions, proposals, technical reports, court filings, classified using your internal taxonomy. Integrates with SharePoint, M365 or whichever DMS you use and routes each document to the right matter file.
Modelled target: F1 > 0.9 on document classification, depending on corpus
Extraction
Critical-clause extractor
Reads contracts and opinions and extracts limitation-of-liability, non-compete, jurisdiction, deadline and penalty clauses, with field-level confidence and a back-reference to the source paragraph, ready for review.
Modelled target: Confidence per clause and page
Knowledge
Private RAG over the firm's know-how
Search assistant over your memos, precedents, opinions and reports, designed to inherit practice-area and client permissions, cite its source and block cross-matter retrieval.
Modelled target: DMS permissions inherited
Scoring
Case-success probability scoring
Model trained on your history of closed matters: estimates success probability, expected timeline and cost for a new case. Built to support the accept/decline decision, not to replace it.
Modelled target: Explainable per feature
Predictive
Project hours and cost predictor
Estimates hours per role and total cost for a project from the client brief and similar closed matters. Cuts under-priced proposals and gets quotes out within 24 hours.
Modelled target: MAE −40% vs manual estimate, on historic data
Illustrative composite scenario · modelled figures and targets, not client results
A 45-lawyer firm.
Boutique commercial-law firm with 12 years of history digitised in SharePoint and a custom DMS. Illustrative composite scenario; the figures below are a modelled baseline and targets, not measured client results.
Modelled baseline
Baseline assumption: Junior staff spend 6–9 hours a week classifying incoming documents and locating specific clauses in long contracts. Proposals are drafted from a template, with no reuse of similar prior work. Project-hour estimates rely on the responsible partner’s judgement, with average deviations of 35%.
Modelled target
Modelled operating target: The classifier auto-assigns 92% of incoming documents to the correct matter with high confidence, the remaining 8% goes to manual review. The extractor pulls critical clauses in seconds and leaves the contract annotated for review. The hours predictor brings estimate deviation down to 12% and shortens proposal turnaround by 60%.
Modelled target: −68% junior hours spent on document classification and lookup
We connect to your stack
Integrations that matter in this sector
CRM
HubSpot
Mid-market CRM with broad APIs, a natural fit for sales agents and lead enrichment.
CRM
Salesforce
Enterprise CRM with fine-grained permissions, AI workflows that respect the data model.
CRM
Zoho CRM
CRM with strong adoption among Spanish SMBs, automations and agents at a contained cost.
CRM
Microsoft Dynamics 365
Enterprise CRM/ERP suite in the Microsoft ecosystem, native fit with 365 and Power Platform.
ERP
Holded
Spanish cloud ERP widely adopted by SMBs, invoicing, expenses and reconciliation automation.
ERP
Odoo
Modular open-source ERP, AI agents and workflows on top of sales, inventory and project modules.
COMMS
Microsoft 365 / Outlook
Email, calendar and SharePoint as channel and context, triage, drafting and RAG over your inbox and files.
Frequently asked
What clients ask us
- 01
How do you guarantee attorney-client confidentiality?
Models are trained and run in private infrastructure, your Azure or AWS tenant or your own servers, without data leaving for third-party APIs. Embeddings and model weights stay inside your perimeter and logs are auditable. We sign an NDA and, where required, a data-processing agreement under GDPR.
- 02
How do you control bias in sensitive decisions?
The case-success scoring is a decision-support tool, not a decision-maker. We document the training set, the features the model uses and its performance by sub-segment (client type, matter, jurisdiction). We audit known biases before going live and provide a panel for partners to review predictions whenever they diverge from expert judgement.
- 03
What happens to the model when case law or your way of working changes?
A model in production isn’t a finished model. We agree a retraining cadence, typically quarterly, using the new history, and monitor drift in real time. If quality drops below the agreed threshold, the model is retrained or, if that isn’t enough, the dataset and architecture are revisited. It’s in the contract, not a surprise.
- 04
Do we need to change DMS or document manager?
No. We work on top of SharePoint, M365, Google Workspace, Notion and the most common legal DMS via API or intermediate connector. The rule is: the model adapts to your stack, not the other way round. If the system has no public API we look at webhooks or scheduled exports.
Same sector, other services
Same service, other sectors
Trust
Safe, traceable AI,
enterprise-ready.
We design for privacy from the start, human control, traceability, usage limits, permissioning and documentation. For sensitive processes, we help assess risk and applicable obligations under GDPR and the EU AI Act.
- 01We never train models on your data without explicit authorization.
- 02Human review built-in for processes where risk demands it.
- 03Traceability: prompts, sources, permissions, errors and metrics, all documented.
- 04Privacy, security and control integrated from day one.
- 05Solutions engineered to be maintained, audited and improved over time.
GDPREU AI ActAEPDISO 27001 readyEU data residency
Personal diagnosis
We work with
few clients.
Every engagement is led personally by one of the partners. If there's a fit, you get a personal first read of your case within one business day, not a canned demo.
How we work
- 01Tell us which process eats your time
- 02Personal reply within one business day
- 0320-minute call, no demo, no pitch