In nearly every first conversation I have with a SLED CIO or IT director, the same line comes up: "We don't really have an AI budget." And technically, they're right — there's no line item in the approved budget labeled "Artificial Intelligence." But when we start looking at p-card statements and expense reports, the picture changes fast.

The budget is already there. It's just invisible, uncontrolled, and working against the organization instead of for it.

$15K
per month in uncontrolled AI spend — the number we found in a single CIO conversation that started with "we don't have an AI budget."

What Shadow AI Actually Looks Like

Shadow AI is not a future risk — it's a present reality in almost every government and education organization we work with. Here's the pattern: individual employees, trying to do their jobs better, start subscribing to AI tools on personal or department credit cards. ChatGPT Plus at $20 a month. Claude Pro at $20. Copilot for Microsoft 365 at $30. A specialized tool for HR at $50.

Multiply that across 80 employees in a mid-size city, county, or school district — many of them on plans that cost $100 to $200 per month — and you are looking at a significant monthly outlay that your IT department has no visibility into, no control over, and no audit trail for.

"The question is not whether your organization is using AI. It already is. The question is whether you know about it."

Why This Is a Problem Beyond the Budget

The cost is the visible part. The risk underneath it is more serious.

  • Data sovereignty. When a staff member pastes constituent data, student records, or personnel information into a public AI tool, that data may be used to train public models. There is no audit trail. There is no way to recover it.
  • Compliance exposure. Individual AI accounts are invisible to FOIA, FERPA, and CJIS discovery requests. When an auditor or attorney asks for records of AI interactions, there is nothing to produce — because the tool was never under IT governance.
  • Dead-end productivity. Personal licenses cannot be connected to your internal data. Staff are getting generic AI outputs, not answers grounded in your actual policies, procedures, and systems. The productivity gain is real but limited — and it does not scale.

The Reframe That Changes the Conversation

This is not a story about spending more money on AI. It is a story about redirecting money that is already being spent — fragmented, ungoverned, and ineffective — into a single platform that your IT team controls.

Google Gemini Enterprise, deployed properly through Google Cloud, changes the equation entirely. Your data never trains public models. Every interaction is logged and discoverable. Access is tied to your existing identity and SSO policies. And instead of 80 individuals doing their best with generic AI tools, your organization has a governed platform that can be grounded in your actual data and connected to your actual systems.

Same budget. Better security. Real governance. A foundation that actually scales.

What to Do With This Information

If you are a SLED technology leader, here is a practical first step: ask your finance team to pull p-card transactions with any of the major AI vendors — OpenAI, Anthropic, Microsoft Copilot, Google One — over the last 90 days. What you find will almost certainly surprise you.

That number is your starting point. It is also your business case.

The path from that number to a governed, centralized Gemini Enterprise environment is shorter than most organizations expect — and it is available through cooperative contract vehicles that require no competitive bid and no council approval cycle.