This fall, state governments made it official: artificial intelligence is no longer an experiment.
The Center for Digital Government's 2026 Digital States Survey put it plainly. Governors are signing executive orders on it. States are naming Chief AI Officers, opening AI Centers of Excellence, and writing frameworks that every agency below them is expected to follow.
At the CAD console of a county 911 center, a dispatcher is working in a system that was bought years before any of those plans existed. The distance between those two places is the story that isn't making the headlines.
The Vision Is Moving Fast
The pace at the top is real. In 2024, 21 percent of states had rolled out AI productivity tools across the enterprise. In 2026, 76 percent have. AI-assisted application development went from 16 percent implemented to 53 percent, with only 9 percent of states still evaluating it. AI-assisted contract review went from 2 percent to 22 percent.
The direction is being set in writing, too. Missouri's governor issued an executive order directing agencies to build a statewide framework for AI in government operations. Other states have AI councils, review boards, and approved-tool lists. Two years ago, most of this was still a question. Now it's policy.
The Ground Is Moving Slower
The view from the other end of the pipeline looks different. In Euna Solutions' 2026 survey of public procurement professionals, 46 percent said they still rely on spreadsheets, email, and paper. Only 17 percent were very or extremely confident their organization is ready for AI, and 41 percent doubted their agency would adopt it responsibly.
CivStart's State of GovTech 2026 report heard the same gap from local leaders. Agencies have appointed AI leads and run pilots, but are stuck on how to scale them without splintering across departments. Regional innovation leads described their tool deployments as "a drop in the bucket" next to the enterprise vision.
None of that contradicts the progress at the state level. It describes the other end of the same plan. A vision written at the top still has to travel through budgets, contracts, and systems that were set up long before it, and most of it hasn't arrived yet.
Where Public Safety Sits in State AI Plans
Look at the flagship AI examples in this year's survey: unemployment claims, benefits renewals, procurement chatbots, legislative tracking, contract review. Public safety appears, but mostly in records. Michigan State Police used text recognition to eliminate more than 100,000 manual reviews in its firearms records unit. That's valuable work, and it's squarely back office.
Live dispatch, where a call comes in and units move, barely appears at all. Yet the expectations land there anyway. Transcription, real-time translation, and AI triage of non-emergency calls are already on every public safety conference agenda. And the AI guardrails agencies are writing elsewhere, with humans making every decision and no output going out unreviewed, will apply to dispatch as much as to any office in the building.
Even State-Led Modernization Arrives One Center at a Time
The survey's own public safety examples show how this works in practice. Ohio is replacing fragmented emergency communications with a statewide Next-Generation 911 platform, and nearly a quarter of its counties have migrated so far. Arizona now has all 78 of its 911 centers on Next-Generation 911, connected to a single network.
Both are real achievements, and neither is a criticism of anyone. Migrating a live 911 center is careful work, and it should be. But they show that even a funded, state-led plan reaches the console one center at a time. The distance between a strategy and a working dispatch floor is measured in years, not announcements.
The Mismatch Between AI Strategy and Dispatch Systems
An AI strategy written at the state or county level carries a quiet assumption: that the systems underneath it can take part. Most CAD in service today was bought under a different budget, for a different era, with a contract written before anyone was asking about AI. Nothing in the strategy connects to it, and nothing in the system was built to receive the strategy.
That isn't an agency shortcoming. The people in communications centers are doing exactly what's asked of them, with the systems they were given. The gap is between plans and systems, and it widens every time a new directive is written for an environment that can't act on it.
An AI strategy assumes the systems underneath can take part. Most CAD in service today was never built to.
What Makes a CAD System Ready to Take Part
The question for every agency isn't whether it has an AI vision. It's whether its mission-critical systems can take part in one. AI readiness in 911 dispatch comes down to three things: whether the CAD system shares data in real time, whether it can be updated without an upgrade project, and whether it can connect to whatever the plan calls for next.
10-8 Systems was built for that. It's cloud-native on AWS GovCloud and continuously updated with no downtime, so new capability reaches the console as part of the platform, not as a future upgrade project. As we wrote in Your Legacy CAD Can't Do What's Next, readiness is a foundation, not a feature.
Before the next directive arrives, ask whether the system on your dispatch floor is part of the plan, or just underneath it.
Next in this series: the outside deadlines landing on CAD environments, and why the clock is speeding up.
Questions Worth Asking
Is AI being used in 911 dispatch?
A: In early forms, yes. Transcription, real-time translation, and AI triage of non-emergency calls are starting to appear in 911 centers. Most state AI programs so far, though, focus on back-office work like benefits processing and contract review rather than live dispatch.
How do state AI strategies affect 911 centers?
A: They set expectations and guardrails that 911 centers will be held to, often without addressing dispatch directly. Ask your state or county AI lead where 911 fits, and what's expected of the systems you run.
What makes a CAD system ready for AI?
A: It shares data in real time, updates without a major upgrade project, and connects to other systems. You don't need to adopt AI to prepare, but you do need to know whether your current system could take part when guidance arrives.
What guardrails should apply to AI in dispatch?
A: People make every decision, outputs are reviewed, and vendors disclose how AI uses agency data. Agencies are already writing these rules for AI elsewhere, and any CAD vendor should be able to explain how its platform supports them.
Is your dispatch system part of the plan, or just underneath it? Let's look at it together.