Our AI Agents Have the Power; You Have the Talent

Every day, I rely on an assistant specializing in Identity and Access Management (IAM). It continuously monitors standards, frameworks, and best practices (NIST, ISO, ENISA, CISA, OpenID Foundation, FIDO Alliance, etc.), automatically collects new publications, analyzes them, and categorizes them by topic so that the information is immediately actionable. I’ve even given it a name: Norma!

Every morning, I receive a summary of critical publications and alerts about developments that could impact compliance or operations. When I need guidance, I request a summary report tailored for decision-makers: an executive summary, key changes, concrete impacts on IAM programs, and operational recommendations with priority levels.

I also use it for technical tasks: quickly comparing two standards (e.g., NIST vs. ISO), extracting new requirements to incorporate into our IGA/PAM/CIAM policies, and generating technical checklists for integration teams. The deliverables are customizable: a short format for executives, a detailed version for operational teams, or a comparison table when a decision needs to be made.

Project delivery, drafting user stories sourced and enriched through the discovery phase The agent assists me in drafting user stories based on documentary sources (workshop minutes, specifications, business reference frameworks). I also incorporate discovery deliverables to enrich the project context and build requirements that are more robust, better documented, and better aligned with the client’s strategy. The agent draws on this set of sources to produce structured user stories, complete with acceptance criteria and business rules, which I review and refine before adding them to the backlog.

Coaching Product Owners, Structuring the Discovery Process, and Supportive Engineering I support Product Owners (POs) in producing the deliverables that mark key milestones in the discovery process (scoping canvases, workshop facilitation guides, presentation outlines, etc.). The agent serves as a co-creation partner: they help formalize, question, and iterate on these deliverables. They also enabled me to build the support program itself, by optimally allocating the available hours to account for each Product Owner’s training and skill-building needs.

For purely technical issues and code architecture, I turn to Deepseek. Its “Deep Thinking” engine is particularly valuable for analyzing complex call stacks, debugging flaky tests, or understanding the impact of a refactoring on the entire test suite. It helps me get to the root of the problem rather than settling for a superficial fix.

When it comes to writing and optimizing the automation scripts themselves, I use Claude. It’s not just about generating code, but about making it more maintainable and efficient (by applying patterns like the Page Object Model or reducing redundancy). Furthermore, when I submit my user stories or existing test plans to Claude, it acts as a “second pair of eyes”: it highlights edge cases or test scenarios I might have overlooked, thereby strengthening functional coverage.