AI Operations Lead

About Karbon Karbon is the global leader in AI-powered practice management software for accounting firms. We provide an award-winning cloud platform that helps tens of thousands of accounting professionals work more efficiently and collaboratively every day. With customers in 40 countries, we have grown into a globally distributed team across the US, Australia, New Zealand, Canada, the United Kingdom, and the Philippines. We are well-funded, ranked #1 on G2, growing rapidly, and have a people-first culture that is recognized with Great Place To Work® certification and on Fortune magazine's Best Small Workplaces™ List. Karbon is further ahead on AI adoption than most companies our size. Every department is running AI daily. Hundreds of skills and workflows have been built by our own team. Fourteen department champions drive adoption and share knowledge across the business. The foundation is real and it works. The next phase is scale. We need one person whose entire job is extending what we have built, measuring what it returns, and adding the next layer of capability. Not a strategy deck. A working system, the governance that makes it trustworthy, and the numbers that prove it. You will report into the CEO and work across every department. You will have the access and sponsorship to make changes stick. About the Role Build the Tooling • Ship internal automations and agents that remove repeated manual work across Sales, CS, Implementation, Support, and Operations. Connect AI to the systems we already run, through APIs and MCP servers. • The first build is the data gateway. Karbon runs its own MCP server. You extend it into a routing proxy that sits between Claude and every internal data source teams need. Authentication via Okta. RBAC enforcement by department. Request logging for spend attribution and audit. Once the gateway is approved by Security, new connectors plug into a config entry instead of triggering a full review each time. That is your first thirty days. • After that: CS gets a renewal risk signal. Sales gets a proposal draft automation. Implementation gets a SOW generator with scope-creep flags. Support gets a ticket triage router. • Prototype fast. Harden what people use. Delete what they ignore. Hand every tool to an owner in the team that uses it. You build it. They run it. Own Governance and Visibility • Manage the Claude access and provisioning workflow — seat requests, model and tool access, connector approvals — and the Slack workflows that support it. Run the connector security review pipeline in partnership with Security. • Own the spend dashboard across all AI tools, not just Claude. Cursor, Codex, Linear AI, and anything else the org is running sit alongside Claude in one view. Department leaders see their own number weekly without asking you to pull it. Bad trends surface before they become invoice surprises. • Extend that visibility into an ROI framework: adoption rate, hours saved, cost per seat, project-level impact. Partner with Finance on budget accountability. • Run the AI champions program. Fourteen department champions, organized into pods, own peer enablement and skill submissions within their teams. You run the cadences, maintain the hub pages, and hold champions accountable for engagement. When people leave Karbon, their projects and skills do not leave with them. You own the offboarding flow that captures and reassigns that work. • Build and maintain the AI wiki — per-department documentation of tools, active use cases, and skills in use. Keep it current as tools evolve. Train Teams and Set the Standard • Work with each department lead to agree how their team uses AI, then hold that standard. • Run practical training. Short, specific, and repeatable. Teach people the efficient way to get the same answer. Most cost problems are habit problems. After month one, you know what each team's workflows look like because you built them. That makes you a better trainer than anyone who has not. • Build and maintain a shared library of prompts, skills, and patterns so knowledge stays when a person leaves. Own the pipeline that moves skills from intake through review to the shared library — and automate the parts of it that should not require a human. Measure and Report the Return • Define the small set of metrics that show whether AI is working: cost, adoption, delivery, and quality. • Build the data pipelines behind those metrics. Expect the source data to be messy. Publish a regular report to the executive team. Same format every time. Bad news included. • Delete any metric that has not changed a decision in a quarter. About You! • You shipped production software in the last two years, as the person writing the code. Python or TypeScript, SQL, and REST APIs. • You built something real on top of a large language model. Not a demo. Something with users, error handling, and a cost you had to control. • You worked directly with non-technical teams as the tec