How CNCC is testing AI-powered research for France's statutory auditors
- Industry
- Financial Services
- Company Size
- 1001+
- Department
- Finance
Key Highlights
- 85-87% satisfaction rate among pilot testers, remarkable for a highly rigorous profession
- 40 external auditors testing the solution (double the initial target)
- Transparent reasoning built into every AI response for compliance traceability
This story was written by Wesype, a Dust Platinum Partner specializing in regulated professional services. It shows how CNCC, France's regulatory body for statutory auditors, is deploying AI-powered document research to transform how auditors access standard regulation information. With Wesype's support, CNCC is building the foundation for profession-wide AI adoption while maintaining the trust and rigor that define the auditing field.
About CNCC
When a statutory auditor needs to verify compliance on a complex transaction, they have minutes, not hours, to find the right doctrine buried in thousands of pages of standards and regulations. Until now, keyword search was their only option. This method is about to change for over 10,000 auditors across France.
CNCC (Compagnie Nationale des Commissaires aux Comptes) is the French regulatory body responsible for the statutory audit profession. Unlike a commercial enterprise, CNCC's mission is institutional: providing the regulatory framework, professional standards, and trusted resources that enable statutory auditors to certify financial statements with confidence.
At the heart of this mission sits Sidoni, CNCC's trusted document database that auditors rely on daily to ensure alignment with doctrine, accounting standards, and legal requirements. With approximately 2,000 connections per day, Sidoni is the profession's key source of truth, making it the natural starting point for CNCC's AI transformation.
What makes this story unique: CNCC isn't just adopting AI for internal efficiency. They're building the foundation for an entire profession to work smarter while maintaining the unwavering standards of trust that define statutory auditing.
💡 Interested in learning more about how Dust can help your team? Reach out to our sales team.
Challenge: When keyword search fails a profession built on precision
Statutory auditing faces a paradox: the profession demands absolute precision, yet the tools for finding authoritative information were built for basic research. For CNCC, this challenge crystallized around one critical question: how do you help auditors find exactly what they need when they don't know the exact keywords to search?
The limits of keyword search
CNCC's Sidoni database had already undergone two major evolutions, moving from a basic system to a modernized Elasticsearch implementation. Usage was strong, 2,000 daily connections demonstrated clear demand. But something was missing.
Auditors weren't looking for documents containing specific words. They needed answers to complex questions: "What are the reporting requirements for this type of transaction?" "How should I handle this unusual audit finding?" "What does the doctrine say about this specific scenario?"
Keyword search could find documents. It couldn't interpret questions.
"Users expected a tool capable of interpreting and understanding complex questions, not simply matching keywords. AI is very relevant for this." — Abel Jabol, CIO CNCC
The compliance imperative
In statutory auditing, getting it wrong isn't just inefficient—it's professionally dangerous. Auditors serve as trusted authorities certifying financial statements. Every conclusion must be defensible, every source verifiable, every reasoning traceable.
This created a unique set of requirements that generic AI solutions couldn't address:
- Confidentiality: Client data must remain protected
- Traceability: Every AI response must show its reasoning
- Source attribution: Conclusions must link to authoritative documents
- Professional judgment: AI must support—never replace—human expertise
"The role of the statutory auditor is to be a trusted authority to certify accounts. We cannot rely on an approximate or random approach to data analysis. Auditors must be convinced according to their training and ethics." — Abel Jabol, CIO CNCC
Solution: building AI that auditors can trust
CNCC faced a classic build-versus-buy decision. Initially, they explored developing an internal solution. The reasoning was sound: maximum control over data, complete customization, no external dependencies.
The reality proved different.
Why internal development wasn't the answer
CNCC quickly recognized the fundamental challenge: their core competency was audit regulation, not software development. Building internally would require:
- Significant ongoing investment in AI expertise
- Constant updates as models evolved
- Infrastructure that would become obsolete within months
- Distraction from their institutional mission
"Our core business is auditing, not software development. The effort required would have been far too significant for a product that would clearly have been obsolete within a few months." — Abel Jabol, CIO CNCC
Why Dust won the evaluation
CNCC selected Dust based on critical factors aligned with their institutional requirements:
- Model agnosticism: Flexibility to leverage different AI models without vendor lock-in
- Focus on use cases: Ability to concentrate on professional applications rather than technical infrastructure
- Ease of administration: Simple management without heavy IT investment
- Legal data protection: Contractual safeguards complementing technical security
The decision reflected a strategic shift: rather than controlling data through technical isolation, CNCC would protect it through robust legal frameworks while gaining access to continuously improving AI capabilities.
Four-milestone transformation approach
CNCC structured their deployment with institutional rigor:
Milestone 1: Concept validation Explored multiple use cases, validated core feasibility, and identified Sidoni as the flagship application.
Milestone 2: Dual-track pilot (current phase)
- Internal track: ~20 staff members using Dust for research, meeting minutes, and content creation
- External track: 40 statutory auditors testing Sidoni AI (originally planned for 20—demand doubled the scope)
Milestone 3: Production launch (targeted June 2025) Full deployment of Sidoni AI to all authenticated statutory auditors, with monitoring and iteration.
Milestone 4: Profession-wide expansion Potential platform for smaller audit firms, prompt libraries, and community-driven AI resources.
The Sidoni AI agent: intelligent document search
The deployment transforms how auditors interact with CNCC's trusted knowledge base:
Semantic understanding: Auditors can ask natural questions rather than constructing keyword queries. "What are the documentation requirements for related-party transactions?" returns relevant doctrine, not just documents containing those words.
Reasoning transparency: Unlike black-box AI, Sidoni shows its analytical steps, filtered to essential phases rather than overwhelming technical detail. Auditors can follow the logic and verify conclusions.
Mandatory source citation: Every response links to authoritative documents. No unsourced claims, no AI-generated assertions without foundation.
Hallucination prevention: Agents are configured to acknowledge limitations rather than fabricate answers. When the knowledge base doesn't contain relevant information, the AI says so.
Beyond Sidoni: internal productivity gains first for CNCC team
While Sidoni serves the broader profession, CNCC's internal team discovered immediate value in complementary use cases:
Meeting minutes automation: Drag-and-drop transcription processing generates structured summaries in minutes rather than hours.
Technical research acceleration: Experts writing doctrine articles can conduct comprehensive research across multiple sources instantly.
Training content creation: Trainers using tools like FrameMaker saw meaningful acceleration in material development.
Results: rapid adoption from a rigorous profession
The pilot phase demonstrated rapid adoption: non-technical professionals embraced AI quickly, with both satisfaction and demand surpassing initial expectations.
Measured satisfaction in a demanding environment
CNCC leveraged Dust's internal analytics to measure pilot performance:
- 85-87% satisfaction rate among testers—remarkable given the profession's natural rigor and skepticism
- Demand exceeded expectations: External pilot expanded from 20 to 40 auditors due to interest
- Minimal friction: SSO deployment and chrome extension enabled seamless access
"I'm surprised by the level of adoption, knowing these aren't really tech-savvy people. But in the browser, just making meeting minutes, drag and drop the transcript, and there you go. Adoption was rather quick." — Abel Jabol, CIO CNCC
The adoption formula
Several factors drove successful uptake:
SSO integration: Single sign-on eliminated authentication friction, making Dust as accessible as existing tools.
Chrome extension: Natural integration into daily workflows rather than requiring users to navigate to a separate platform.
Copilot comparison: CNCC recognized that without seamless integration, any AI tool would struggle against Microsoft Copilot's native positioning. The chrome extension leveled the playing field.
Use case focus: Rather than mandating adoption, CNCC made AI useful for specific tasks. Once users experienced time savings, organic expansion followed.
"We don't want to force AI usage for the sake of it. We want AI to be useful in the user's practice. When you see the time you can save on certain tasks, you wonder why take four times longer." — Abel Jabol, CIO CNCC
The "positive problem" indicator
Perhaps the most telling result: CNCC anticipates managing high demand rather than driving adoption.
"My feeling is that we'll probably have more problems controlling usage than getting people to use it, which is a good problem to have. I think it will be expected and successful." — Abel Jabol, CNCC
Key learnings: deploying AI in regulated professions
Based on CNCC's transformation journey, several principles emerged for organizations in trust-dependent, regulated industries:
Start with your highest-impact knowledge base
Sidoni's 2,000 daily connections made it the obvious starting point. The use case with the clearest demand and measurable impact builds confidence for broader deployment.
Make reasoning visible
In professions where conclusions must be defensible, black-box AI creates unacceptable risk. CNCC's emphasis on reasoning transparency and source citation addresses this fundamental requirement.
Integration beats innovation
Browser plugins, SSO, and seamless workflow integration drove adoption more effectively than sophisticated features. Users shouldn't notice they're using a new tool—they should notice their work getting easier.
Leverage AI Deployment Strategist Partner for acceleration
CNCC's collaboration with Wesype accelerated deployment by bringing external expertise in use case definition and agent development.
"Wesype helped us launch, secure the approach, and better define our agents. Not just talk, but actually go for it, test, and see. I think Wesype is a useful partner to accelerate AI adoption." — Abel Jabol, CIO CNCC
The trust equation solved
CNCC's transformation answers a question many regulated professions are asking: can AI be trusted in environments where trust is everything?
The answer, it turns out, is yes—but only with the right approach. Not AI that replaces professional judgment, but AI that enhances it. Not black-box answers, but transparent reasoning. Not generic capabilities, but specialized agents built for specific professional needs.
For France's 10,000+ statutory auditors, the future of compliance research is about to change. And for other regulated professions watching closely, CNCC is writing the playbook.
Interested in learning more about how Dust can help your team? Visit our solutions page or reach out to our sales team.
