Consultant Agentic AI & Prompt Engineering – Insurance Document Processes

Project overview

Context and Mission Details

Mission Context

Within a major insurance player's Data, AI, and Innovation department, the mission consists of supporting the deployment of agentic AI solutions applied to business document processes.

The mission is part of a transformation initiative aiming to:

The consultant will work in close collaboration with the insurance business units, AI teams, product teams, and IT teams.

The mission is based in the West region, with regular interactions with various business and IT teams.

Main Missions

Gathering and Scoping Business Needs

Designing Agentic Solutions

Configuration in an Internal AI Platform

Quality Improvement and Industrialization

Business Support

· Validate use cases with users.

· Adapt agents based on field feedback.

· Disseminate best practices for using agentic AI.

· Create a best practices guide for the AI platform.

Profile Sought

Technical Skills

Generative & Agentic AI

· Mastery of LLMs and agentic architectures.

· Experience in document processing: Document AI, RAG, NLP.

· Good understanding of orchestration tools: LangChain, Semantic Kernel, or equivalents.

Advanced Prompt Engineering — Strong Requirement

Mastery of agentic prompting patterns, including:

Demonstrated expertise in these patterns is essential to ensure the performance, reliability, and industrialization of the agents developed within the AI platform.

Data and Documents

Functional Skills

Expected Experience

Work Environment

The consultant will work within an internal AI framework aiming to industrialize the use of artificial intelligence within the organization.

They will collaborate with:

Estimated Start Date: 07/01/2026

• improve the exploitation of document corpora;
• automate the production of business documents;
• deploy business AI agents via an internal AI chat solution based on an open-source solution;
• contribute to the implementation of the governance associated with the tool: agent evaluation principles, sharing modalities, and best practices.
• Lead workshops with insurance business units: underwriting, claims, legal, compliance, etc.
• Analyze existing document processes: research, analysis, document production.
• Identify and prioritize use cases with high business and operational value.
• Translate business needs into AI agent architecture.
• Design specialized agents for several uses:
• document analysis and understanding;
• information retrieval via RAG;
• generation of standardized documents.
• Define advanced prompting strategies adapted to business use cases.
• Configure and set up agents in an internal AI platform based on an Open Source solution.
• Implement prompts, workflows, and business rules.
• Connect document sources: DMS, internal databases, document repositories, etc.
• Test, adjust, and optimize agent performance.
• Structure prompt and template libraries.
• Ensure the robustness, traceability, and reproducibility of the results produced.
• Document the implemented solutions.
• Participate in defining the governance, evaluation, and sharing principles for agents.
• ReAct — Reasoning + Acting;
• prompt chaining;
• plan-and-execute;
• agentic RAG;
• self-refinement / iterative critique;
• multi-agent orchestration.
• Ability to design robust, traceable, and industrializable prompts.
• Experience in handling structured and unstructured documents.
• Knowledge of document standardization issues.
• Design of automated document models.
• Good knowledge of the insurance sector, highly appreciated.
• Ability to understand and formalize complex business needs.
• Sense of business value and operational impact.
• Significant experience in generative AI.
• Experience in document or knowledge management projects.
• Experience in implementing agentic solutions, highly appreciated.
• insurance business teams;
• Data / AI teams;
• product teams;
• IT teams.

Project details

• Start: ASAP
• Duration: 12 mois (renouvelables)
• Location: 72000 LE MANS, FR
• Languages: French