Director of AI

Director, Artificial Intelligence (AI Leader)

Client is standing up an enterprise AI capability on a data foundation it has just finished building. Microsoft Fabric is live, Copilot is deployed to an initial seat population, Claude Enterprise is scoped, and an AI Center of Excellence charter is drafted but not yet launched. What does not exist yet is production AI: no governed intake process, no operational use cases beyond individual productivity, and no ML engineering capacity.

The AI Leader is hired to close that gap. This role owns Client's AI strategy and its execution — standing up and running the AI Center of Excellence, scaling assistant adoption from pilot to enterprise, shipping the first production use cases in estimating, RFP response, and document intelligence, and building the governance framework that makes AI safe to deploy in a business where bad output has cost and safety consequences.

Success is measured in adopted, governed AI that moves named workflows, not in demos. The role is a builder's role with a clear runway: a funded three-year AI investment plan, an executive mandate, and a target of meaningful productivity uplift in named workflows within three years. It works in close partnership with the data function — data quality is the explicit gate on this roadmap — and with Applications, Integration, and Cybersecurity.

• Take the drafted AI CoE charter to launch — operating model, membership, decision rights, funding model, and reporting cadence to IT and executive leadership.
• Build and run the AI intake process: how ideas are submitted, scored for value and risk, prioritized, funded, piloted, and either scaled or killed.
• Establish the reusable delivery pattern for AI work — reference architecture, prompt and model standards, evaluation methodology, testing, and production support expectations.
• Embed and enable business AI champions in each operating company so demand is generated and qualified locally rather than routed through IT alone.
• Own the AI roadmap and portfolio view, sequencing productivity, workflow, and agentic use cases against capability and data readiness.

• Ship the first wave of operational use cases to production — estimating assist, RFP and specification parsing, submittal and contract-clause extraction, and document summarization.
• Progress the portfolio into workflow automation: three-way PO and invoice matching, change-order triage, dispatch and route optimization, project risk scoring, and financial close assistance.
• Build toward agentic use cases with appropriate human approval gates — bid response, field ticket coding, supply chain, tier-one customer service, and continuous project controls reforecasting.
• Define the value case and measurement approach for each use case before build, and report realized benefit against it after deployment.
• Partner with the data function to ensure each use case has governed, quality-gated data with documented lineage before it goes to production.

• Own the enterprise assistant program — scaling Microsoft Copilot from its current pilot population to broad white-collar deployment, and deploying Claude Enterprise from scoped to live at scale.
• Drive real adoption rather than seat counts: role-based training, workflow-specific playbooks, champion networks, office hours, and measured usage by function.
• Design and deploy Copilot Studio and comparable low-code agents for targeted workflows such as HR and onboarding policy Q&A, parts lookup, and field technician guidance.
• Manage license allocation and seat economics against the approved investment plan, reclaiming and redeploying unused capacity.

• Build Client's AI governance framework — acceptable use policy, model and tool approval process, data handling rules, human-in-the-loop requirements, and audit trail.
• Establish model risk management: evaluation and accuracy thresholds, bias and fairness review, hallucination controls, drift monitoring, and defined rollback criteria.
• Use Microsoft Purview and equivalent controls to govern prompts, model access, and data exposure, in partnership with the…