AI Development Engineer Senior

Gallagher•Rolling Meadows, IL

About The Position

The AI Workflow Architect is responsible for transforming AI agents from demonstrations into production workflows that Gallagher colleagues can use daily. This role involves collaborating with business users to understand their processes, designing and building agents and automations using Microsoft Copilot Studio, Power Platform, and Azure AI services. The position also includes owning the strategy and roadmap for operationalizing AI agents, establishing evaluation and testing procedures, defining and reporting key metrics, and integrating low-code solutions with Azure AI services. Additionally, the role involves evaluating emerging Microsoft agent tooling, embedding governance into workflows, and coaching business teams on agent capabilities. Excellent communication skills are essential for interacting with business users, engineers, and executives.

Requirements

7-10+ years of professional experience in solution architecture, application development, or business process automation, including at least 1 year delivering AI/GenAI or conversational agent solutions into production.
Hands-on, production-level experience with Microsoft Copilot Studio and the Power Platform (Power Automate, Power Apps, Dataverse, custom connectors), including environment strategy and deployment.
Solid working understanding of the Azure ecosystem relevant to AI workloads - Azure OpenAI/AI Foundry, Azure AI Search, Azure Functions, API Management, Entra ID and RBAC - enough to design integrations and hold a credible architecture conversation with the engineering team.
Demonstrated ability to reason about end-to-end systems - business process, agent behavior, integrations, data, and governance - and to anticipate second-order effects of a design rather than optimizing one component in isolation.
Track record working directly with non-technical business users to elicit how work actually gets done and translate that into a working solution they adopt.
Experience defining how a system's quality is measured and proven - test strategy, acceptance criteria, and operational metrics - rather than judging quality by demo.
Exceptional verbal and written communication skills, with the ability to present comfortably to both engineering audiences and senior business leaders.
Bachelor's degree in Computer Science, Information Systems, Engineering, or a related field, or equivalent practical experience.

Nice To Haves

Familiarity with M365 Copilot extensibility, Copilot agents in Teams and SharePoint, Semantic Kernel, LangChain, or Microsoft Agent Framework, and emerging Microsoft agentic-work tooling such as Copilot Cowork and Microsoft Scout.
Practical understanding of how content structure, metadata, chunking, and index design affect retrieval quality and agent grounding.
Insurance, financial services, or another regulated industry, with exposure to model risk management or SOX-type controls.
Experience standing up a Power Platform or automation CoE - governance model, environment and DLP strategy, reusable patterns, and maker enablement.
Working proficiency in Python, C#/.NET, or JavaScript for the cases where low-code isn't enough, plus REST API and JSON fluency.
Experience with Azure DevOps for backlog, pipelines, and solution deployment.
Microsoft certifications such as AB-100, PL-400, AI-103, or AZ-305.

Responsibilities

Design Agent Workflows With Business Users: Sit with colleagues in corporate divisions to map how work actually gets done, identify where an agent adds real value versus where a human decision must stay, and design agent workflows that fit into existing processes and systems of record.
Build and Productionalize Agents on Copilot Studio and Power Platform: Build agents, topics, connectors, and automations in Microsoft Copilot Studio and Power Platform (Power Automate, Power Apps, Dataverse), and take them through the full path from prototype to supported production solution with proper environments and release discipline.
Own the Operationalization Strategy and Roadmap: Define and maintain Gallagher's roadmap for taking citizen developer AI agents to production - which business workflows come next, what has to be true before an agent goes live, what platform capability is needed to get there.
Establish How Agents Get Evaluated and Tested: Create the evaluation and testing approach for agents before and after launch: golden question sets, groundedness and accuracy checks, regression testing on changes, red-team style probing for failure modes, and acceptance criteria business owners can sign off on.
Define and Report the Metrics That Matter: Stand up the measurement and reporting layer for the agent portfolio - adoption, task completion, containment, accuracy, latency, cost, and business outcome - and deliver it as a regular, credible view for the AI Technology leadership team and business sponsors.
Bridge Copilot Studio and the Azure Stack: Determine when a workflow belongs in low-code Copilot Studio and when it needs the engineering team's Azure AI Foundry / Azure OpenAI services, and design the integration points - custom connectors, APIs, Azure Functions, Azure AI Search indexes - that let the two work as one solution.
Evaluate and Introduce Emerging Microsoft Agent Tooling: Assess emerging Microsoft agent and agentic-work capabilities, including Copilot Cowork and Microsoft Scout, run controlled pilots against real Gallagher workflows, and make clear recommendations on where they fit in our environment and where they don't.
Build Agent Governance Into the Workflow: Apply Gallagher's IT policy - Entra ID/RBAC, DLP policies, environment strategy, data residency, audit logging, and human-in-the-loop checkpoints - as part of how agents are designed, not as a review afterward.
Enable and Coach the Business: Create the patterns, templates, and guardrails that let business teams participate in agent building safely, and coach stakeholders on what agents are genuinely good at so expectations are set honestly.
Communicate Across Audiences: Explain agent design decisions, evaluation results, and roadmap trade-offs in terms that business users, the engineering team, and executive sponsors can each act on.