Director of AI

ExtensisHR•Houston, TX

2d•Remote

About The Position

We have started building the foundation with an AI Council, Practitioner Academy, 100+ internal skills, 25+ active Claude CoWork/Code/Design users, governance in flight. But the path forward is still being shaped, and we need a leader who can help shape it. The right person brings their own perspective on where AI is going in our space, challenges our assumptions, and partners with the CEO & CSO and entire ELT to set the vision, and then runs the practice that delivers against it. This is a builder role for someone who wants to define what AI looks like at a 25-year services firm transitioning into something new. You will lead a small, focused team, coach our consultants and engineers, and turn our deep manufacturing and distribution expertise into productized AI offerings that customers pay for. You will grow into a larger role as the practice scales.

Requirements

5+ years of relevant experience.
Hands-on AI builder.
Direct experience with Claude/Anthropic, agent frameworks, RAG, and evaluation systems.
You can architect and ship, not just delegate.
Track record of leading small, high-performing teams across technical and functional disciplines.
Strong commercial instincts.
Comfortable with margin profiles, pricing conversations, and customer-facing engagements.
Experience in SMB or mid-market services, ideally in manufacturing, distribution, ERP, or adjacent industries.
Ship production agents, not demos.
Hands-on with agent frameworks, RAG and retrieval, and evaluation tooling, and you build in the reliability basics every agent needs: bounded loops, eval criteria, monitoring, and human checkpoints.
You can read and write code and make the architecture calls, not just delegate them.
Comfortable working in a modern cloud and wiring agents into ERP, CRM, and operational systems through clean APIs and trustworthy data.
Strong handle on AI cost.
Token economics, model tiering and routing, caching, and consumption tied to a real budget.

Nice To Haves

Exposure to Acumatica, Sage Intacct, Infor, or comparable mid-market ERP ecosystems.
Background that blends startup builder energy with services-firm discipline.
Has stood up an AI program or center of excellence, including the operating model, governance, and the path from proof of concept to production.
Experience driving AI adoption across delivery teams with tools like Claude Code, GitHub Copilot, or Cursor.

Responsibilities

Partner with the CEO, CSO and ELT to shape and continuously refine the WM Synergy AI strategy.
Create and continuously communicate a shared vision of where the AI practice is going.
Translate market signals, customer needs, and competitive moves into a roadmap the team can execute against.
Surface market and customer signals back to the ELT.
Lead a small, focused AI team and hire and develop 1 to 2 Business Analysts, agent engineers, and/or AI experts in year one.
Add fractional or part-time specialists as needed (FinOps, evaluations, data engineering).
Coach and support our consultants, functional and technical, and software engineers as they learn to build agents, ERP customizations, reports, and productized IP based on their industry expertise.
Run the AI Practitioner Academy.
Grow the Skill Library.
Run the customer feedback loop.
Sit in on flagship reference customer engagements, listen for what customers actually need, and feed it back into the roadmap.
Deliver the productized agent roadmap.
Build 3 to 5 industry-specific agents in year one tied to manufacturing and distribution workflows.
Partner with Sales and Delivery to support customer-facing AI engagements, from scoping through deployment and optimization.
Own all internal AI spend: Anthropic, open-source infrastructure, and so on.
Build cost dashboards, model-routing policies, and consumption governance.
Establish token discipline as a team practice: bounded-loop design, model tiering, caching, prompt optimization, evaluations.
Build AI FinOps as an external service line, assessment, governance, model selection, value tracking, turning the discipline we build internally into recurring customer revenue.
Run the Idea-to-Production Factory in partnership with the AI Council.
Intake, prototype, evaluate, productize, retire.
Maintain the AI Project Submission Form, value-per-token forecasts, and reliability standards (bounded loops, evaluation criteria, human-checkpoint design).
Own the decision framework for when we use Claude/Cowork, when we use Velocity Suite, when we use Acumatica's native AI, and when we use open-weight models.
Maintain the architecture standards across the team.
Make reliability an economic discipline.