top of page
Search

AI Tools Procurement: Good Practice Guide for Legal Departments

  • Writer: Marc May
    Marc May
  • 2 days ago
  • 3 min read

The procurement of Artificial Intelligence (AI) tools is no longer just a technical demonstration (and it never should have been but this is another topic).

In 2026, for legal departments, it’s a high-wire act that must balance technological agility, compliance with the AI Act, or event budgetary discipline (non-exhaustive list). To succeed, the project must align with professional software procurement standards.


I. The Audit Phase: People, Process, Technology


1. People: Identifying Needs and Countering Shadow IT (or shadow AI)


One of the primary drivers of procurement is often security. Without an official tool, lawyers may use free versions of AI, exposing the company to data leaks (and then you know: data breach, etc.).


The method: 

  • Map out the “pain points.”

  • Organize a workshop where each team member lists their tasks where AI could be helpful.

  • Prioritize those with high impact and low added value from the lawyer (e.g., reviewing 200 NDAs per month, drafting emails, or even translating content).


2. Process: Don’t Automate Chaos


Effective procurement requires not “automating disorder”, even if it’s with AI.

If your contract templates aren’t standardized, AI will produce inconsistent results.

Simplify/build up your workflows before introducing technology: reducing upstream clause variations drastically increases AI accuracy (and you get efficient processes, isn’t it wonderful?).


3. Technology: Identify your TCO


Beyond functionality, the buyer must assess the TCO (Total Cost of Ownership). The license price is just the tip of the iceberg.


Include in your calculation (e.g.):


  • Implementation and configuration fees.

  • Consumption costs (tokens) if usage explodes.

  • Human time dedicated to setup and maintenance.

  • Time spent on training and hypercare.


You will notice that the cost is not “just” the license fees.


II. The Pillars of Software Procurement


1. Vendor Viability


The AI market is volatile. A good practice is to audit the supplier’s financial health:


  • Is it a startup that might disappear in 18 months?

  • What is its roadmap?


Procurement must ensure the tool is scalable and can integrate via API with existing tools (like your CLM, ERP, or Microsoft 365).


2. The “AI-Native” Contractual Framework


  • Opt-out from Training: Your data should never (in our view) be used to train the provider’s global model. Demand strict segregation through technical certification (and ensure the contract allows you for either audit rights or an annual certification, for example).

  • IP & Ownership: You must retain exclusive ownership of the inputs (prompts) and outputs (results).

  • AI Act & Compliance: The provider must guarantee sufficient and documented transparency on its datasets and risk classification according to European regulations (if the AI Act applies, but since it could become a GDPR-like standard, it’s best to anticipate).


3. Reversibility and “Vendor Lock-in”


To avoid being locked into a solution, add to the contract free export of data (incl. for instance answers) and prompts in standard formats (JSON/CSV) as well as migration assistance (free or paid, depending on your negotiation skills and leverage).


III. Governance and Legal-IT Co-Leadership


Successful procurement finally also relies on a Legal Department / IT Department partnership. Legal defines the need, while IT validates cybersecurity (SOC 2, ISO 27001) and interoperability.


1. From PoC to Industrial Pilot


It goes without saying, but it’s worth repeating: don’t sign a multi-year contract after a simple demo, however “big” or “small” you are.


Proper procurement requires a PoC (Proof of Concept) of 4 to 8 weeks (for instance) with verification elements regarding the tool’s compatibility with your needs.


For this, establish 5 to 10 test scenarios (Test Cases) representative of your daily work.

  • Example: “Extract the termination clause from 10 complex bilingual contracts” or “Identify deviations from our liability policy in a supplier contract.”

  • Success Criteria: For each test, determine in advance what constitutes success (e.g., 90% correct extraction, output format respected). The supplier selection should be based on these factual criteria, not on a visual impression (despite this could also be another criteria you can add to your requirements).


2. Measuring ROI and Adoption


Three roles are essential for managing the investment:


  • Sponsor (General Counsel): Validates the budget and vision.

  • Legal Ops / IT: Technical management and KPIs.

  • Ambassadors: “Champion” lawyers who test and train their peers.


Key KPIs:


  • Adoption: e.g., > 60% of the team active after 6 months (check logs).

  • Satisfaction: e.g., Score > 7/10 (send questionnaires during the hypercare phase).

  • Savings: e.g., Contract cycle time reduced by 15% to 30% (quantify the gain, as the C-suite speaks in numbers, not employee well-being/fun).


Conclusion


By adopting a rigorous procurement methodology focused on TCO, vendor viability, strong contractual guarantees and clear governance, you ensure that AI becomes a strategic internal asset rather than just another tool destined for the graveyard of failed projects.


Quentin Peltier

Legal Tech Consultant


Comments

bottom of page