We handle privileged data. How do you access our code?

Under NDA, with read-only repository access. We don't clone to personal devices, we don't train models on your code and we don't subcontract. If you need stricter access (VPN, secure room), we'll set it up.

Does the audit help in front of a professional bar or data protection authority?

The report documents findings, severity, what was fixed and what remains, useful evidence of due diligence. It doesn't replace a legal audit if your bar or regulator requires one specifically, but the two complement each other.

Our AI assistant uses privileged material as context. Is that a risk?

Depends how. If documents flow into the prompt and the conversation stays in platform logs, yes. We audit what reaches the logs, what's retained by the model provider and what deletion policy you have contractually.

Do you cover integrations with specific legal/professional systems?

Yes. We audit the connectors your team built, what permissions they asked for, what they can modify, how they handle errors. If your integration isn't standard, we read it the same way we read any other code.

What is the end-to-end process, and who closes the audit?

At intake we agree scope, environments and owners. We combine static review with dynamic tests in staging using synthetic matters across permissions, connectors and logs; the report prioritises findings and assigns each remediation to an owner on your team. We then retest corrected controls, and the designated partner or security owner accepts closure against the agreed criteria.

Service · AI code audit · Professional services

AI code audit for firms and consultancies handling client data.

Your team built a client portal, a contract-review assistant or an intake bot with Cursor, v0 or Lovable. We audit it before one client reads another client's code.

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

What looks like ‘internal productivity’ is also privileged information.

Law firms, consultancies, accounting practices and marketing agencies have leaned on AI tools to build internal assistants: contract review, billing automation, client portals, intake bots, draft generation. The speed is attractive. The problem shows up when two clients share the same codebase and a sloppy ownership filter gives client A’s documents to client B’s team.

We audit what your team built with the sector lens on: professional secrecy, audit trails for regulators, custody of privileged material, and the spots where the AI accepted as ‘good enough’ a layer of security that a professional bar wouldn’t accept on review.

Implementation contract

How it runs in production

Workflow and actors

With engineering, security, privacy and matter owners, we trace the portal or assistant from identity and client through document, prompt, draft and billing, reproduce findings, and agree remediation and retest.

Systems and data

We review code, APIs, authentication, data queries and policies, storage, prompts, logs, secrets and CRM, DMS or billing connectors handling matters and privileged information.

Exceptions and risks

Tests use synthetic or authorised matters and avoid extracting another client's content. A cross-client leak, public key or unintended transfer to a provider is escalated for immediate containment.

Human review

The firm confirms conflict walls, permissions and professional duties; privacy or security validates handling and engineering accepts the fix. A finding closes after an evidenced retest.

Implementation pattern

We model client, matter and role, combining manual review, authorisation and RLS tests, prompt and log tracing, and connector-scope analysis. Every recommendation includes a verifiable acceptance criterion.

Relevant integration

We audit the connectors the tool actually uses with its DMS, CRM, iManage, NetDocuments, Clio, QuickBooks, Xero or other in-scope systems; access and coverage depend on the authorised environment.

What we build for this sector

Use cases that ship to production.

See full catalogue →

Isolation

Per-client isolation in internal tools

Ownership filters and access rules per matter / engagement / account. We verify the database query carries the right WHERE, not just the UI.

Modelled acceptance target: no cross-client leaks

Audit trail

Trail of every AI assistant action

Logs of which document was opened, which query ran, which draft was generated and who approved it. Ready for a bar review or a data protection audit.

Modelled acceptance target: all actions traceable

Confidentiality

Privileged material in prompts and logs

We look for sensitive context being sent to the model, retention in platform logs, and names / IDs / matter numbers in console.log statements that survived the build.

Modelled acceptance target: no PII in platform logs

Integrations

Connectors to legal CRM, DMS and billing

Read/write permissions of the AI agent on iManage, NetDocuments, Clio, QuickBooks, Xero and similar, what it can touch, what it must never touch, with explicit deny rules.

Modelled acceptance target: permissions audited piece by piece

Illustrative composite scenario · modelled figures and targets, not client results

A 35-lawyer firm.

Hypothetical corporate law firm with an AI assistant built on Cursor and Supabase for contract review and historical-matter search. Illustrative composite scenario modelling an audit before widening access; it does not describe a client or actual findings.

Modelled audit hypothesis

Modelled risk to test: the UI might filter matters by user while SQL queries omit that filter, allowing an unassigned matter to be requested through the API. Debug logs could also send client names and matter IDs to the platform's log viewer.

Modelled remediation target

Modelled illustrative outcome: an audit could identify 12 findings and prioritise 4 criticals before widening access: an ownership filter in SQL, PII removed from logs, RLS policies on matter tables and rotation of keys exposed in the client bundle. The remainder would be documented with a fix order and timeline.

Modelled target: 4 criticals remediated before extending access to 35 lawyers

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
We handle privileged data. How do you access our code?
Under NDA, with read-only repository access. We don't clone to personal devices, we don't train models on your code and we don't subcontract. If you need stricter access (VPN, secure room), we'll set it up.
02
Does the audit help in front of a professional bar or data protection authority?
The report documents findings, severity, what was fixed and what remains, useful evidence of due diligence. It doesn't replace a legal audit if your bar or regulator requires one specifically, but the two complement each other.
03
Our AI assistant uses privileged material as context. Is that a risk?
Depends how. If documents flow into the prompt and the conversation stays in platform logs, yes. We audit what reaches the logs, what's retained by the model provider and what deletion policy you have contractually.
04
Do you cover integrations with specific legal/professional systems?
Yes. We audit the connectors your team built, what permissions they asked for, what they can modify, how they handle errors. If your integration isn't standard, we read it the same way we read any other code.
05
What is the end-to-end process, and who closes the audit?
At intake we agree scope, environments and owners. We combine static review with dynamic tests in staging using synthetic matters across permissions, connectors and logs; the report prioritises findings and assigns each remediation to an owner on your team. We then retest corrected controls, and the designated partner or security owner accepts closure against the agreed criteria.

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

Start the conversation →