We care more about solving the problem than using the most powerful technology.
Key Responsibilities
Own the platform decision for each enterprise AI use case — Copilot Studio, Microsoft Foundry or the Power Platform — working from the business outcome rather than the product, choosing the least powerful option that meets the requirement, and defending that trade-off in design review
Say so when the answer is not an agent at all: structured, predictable work belongs in deterministic automation, and static retrieval in classic search. Recommending that is part of the job
Use agile and design-thinking methods to uncover, challenge and prioritise valuable AI and automation use cases
Size and scope agent use cases against a published classification framework; own the estimate and sign off the scope
Design the per-agent security model: agent identity in Microsoft Entra, least-privilege permission scopes, Conditional Access treatment, and placement of human approval gates on write-actions
Define the data-protection checklist for each agent — sensitivity labels and DLP behaviour across grounding sources and action paths — and route findings into the appropriate remediation track
Architect complex and autonomous builds end to end: solution design on Copilot Studio, Microsoft Foundry, custom connectors and API integration, entitlement design, environment and subscription structure, and cost model
Run AI-specific threat modelling for high-autonomy agents, validating the existing threat framework against MITRE ATLAS and the OWASP generative-AI risk lists, and plan adversarial testing
Define the evaluation method for every agent — grounding-quality question sets, functional test cases, acceptance thresholds — and approve results before go-live
Design and advise governance guardrails within the customer's Power Platform — managed environments, DLP connector policies, the custom connector approval route, capacity and license consumption, and the Center of Excellence tooling that keeps the estate visible
Own the environment strategy and the deployment process across development, test and production
Lead effort calibration and change-control conversations with customers as delivery data accumulates
Mentor agent engineers, own their skilling plan, and contribute reference architectures, patterns and playbooks and other reusable assets