Two AI programs: one governs the company, one runs it
A corporate AI policy and enablement program that made adoption safe at scale, and an agentic service platform that became the company's single front door — IT, Facilities, Marketing, and People on one AI-powered desk, with a staffing strategy built on automation rather than headcount.
Adoption was going to happen. The only question was who was steering.
At a company building hardware for AI, employees reach for AI tools faster than any policy cycle can keep up. Demand for Claude seats outran the available pool. The realistic choice was never whether the company would adopt AI — it was whether IT would be the organization that enabled it or the one everybody routed around.
Left alone, that becomes real exposure. An org-wide audit surfaced 34 users paying for duplicate AI subscriptions, and unsanctioned tools meant company data moving through accounts nobody controlled. Aaron named it for what it was — a material business continuity and data exfiltration risk — and took a formal risk-treatment decision to the steering committee rather than letting it accumulate quietly.
The second problem was internal, and bigger than IT. Employees needing something — a laptop, a desk, a brand asset, an HR answer — had no single place to ask, while the company grew toward 490 and then 650 people. Staffing four service functions linearly was never going to work. So the answer to both problems was built as two deliberate programs: one that governs how the company uses AI, and one that uses AI to run the company's service operations end to end.
What the work actually was.
Program one — corporate AI policy & governance
Drafted Lightmatter's first AI Governance Policy, ran an org-wide AI tools audit, prohibited personal-account usage, and took AI tool sprawl to the IT & Security Steering Committee as a documented risk — with a recommended mitigation of SLA-backed enterprise contracts plus CASB blocking of unsanctioned tools, and an explicit risk-acceptance alternative for leadership sign-off.
Program one — enterprise AI at scale
Grew Claude from zero to 224 seats in six months with demand still exceeding the pool, signed a Cursor enterprise agreement covering 126 engineering users to stabilize cost, and enabled 350 Gemini users with roughly 170 active daily — 194 paid AI users under active management, with a dashboard giving leadership visibility into AI spend and adoption.
Program two — one front door for the whole company
Built and scaled an agentic service platform from nothing on an AI-native service management platform, then extended it beyond IT into Facilities, Marketing, and People — a genuine one-stop shop where employees ask once and the right function answers, instead of learning four separate intake processes. Live end of Q4 2025; 1,713 tickets through it in H1 2026.
Program two — agents doing the work
Deployed the AI agent layer across every queue: instant first response, automated first-level security incident handling with industry-standard categorization and SLAs, and automated application access approvals replacing manual email chains. The agents now resolve 30–50% of employee requests across the four queues, with a further meaningful share of remaining IT tickets closed by AI in queue — a combined footprint approaching half of all IT requests.
Program two — the knowledge that feeds the agents
Mined closed-ticket patterns into a 60-article knowledge base — 25 end-user, 35 technician runbooks — reaching 85%+ coverage of recurring patterns, with AI now generating articles directly. Eighteen were added in a single sprint. Better knowledge is what raises the deflection rate.
The thesis both programs serve
Set IT staffing strategy explicitly to beat industry benchmarks through AI and tooling rather than linear headcount growth: from a 3.8% IT-to-employee baseline to a 3.2% target at 490 employees in 2026, and 2.9% at 650 in 2027.
What it delivered.
“One program made AI safe for the company. The other made it a one-stop shop — IT, Facilities, Marketing, and People behind a single door — with a large share of requests closing without a human. That is how a small team supports a company this size.”
Four ways it mattered.
IT positioned as the accelerator of AI adoption rather than its bottleneck.
Sprawl named as a formal risk and taken to executives with a treatment plan.
One AI front door across IT, Facilities, Marketing, and People.
AI used to bend the IT staffing curve, not just to answer tickets faster.
