At the Environmental Council of the States (ECOS) Spring Meeting earlier this year, I participated in a panel on AI, and it was one of the most hotly discussed topics among the state agency leaders in attendance. The interest was undeniable, but so was the caution: any AI an agency adopts must prioritize accuracy, data security, while keeping public trust front and center.
What I heard—and what we’re seeing in our own work—is that AI has already arrived at environmental agencies, both through formal pilots and, just as often, through staff quietly experimenting with tools on their own. The question is no longer whether to engage with AI, but rather how to apply it in practical, controlled ways that align with the agency mission and rules.
Start with one well-defined problem
We recently put AI to work for a client on a narrow, practical problem: too many permit applications were being returned for corrections, not because of regulatory complexity, but because required information was missing, inconsistent, or buried in attachments.
Permit applications routinely require attachments containing specific kinds of content, such as engineering plans, maps showing environmental impacts, and sampling results. Most attachments include their own detailed content requirements set by the agency, such as Surface Water Pollution Prevention Plans (SWPPP) or asbestos demolition project inspection reports. However, historically, the only way to confirm that an attachment meets these requirements is for a reviewer to open it and read it.
So, a Windsor team worked with this client to build AI into their application intake workflow to do that “reading” automatically; each attachment was checked against content requirements, flagging any deficiencies before the application even reached a staff reviewer. This work was well suited to AI: reading, interpreting, and cross-referencing information in both structured and unstructured documents against clear, consistent criteria configured in the electronic form itself. Early results point to a substantial drop in applications returned for corrections and, consequentially, faster permit issuance.
The next step is to move that automated review even further upstream. We’re currently building the ability for agencies to configure specific online forms so the AI checks the content on the applicant’s side, before submission. This catches gaps while the applicant can still fix them, rather than after the application is in the queue.
The success we saw with this client project reflects a broader trend I heard at the ECOS conference: agencies are finding success by focusing on small opportunities, such as a single workflow or use case, rather than attempting broad implementation from the start.
_____________________
_____________________
Identify and leverage informal AI use that’s already happening
In many cases, small opportunities are already emerging through informal use, highlighting where staff are spending time and where AI can realistically support their work. Agency staff are already using or experimenting with AI tools like Microsoft Copilot and Google Gemini in their everyday work. This points to a practical reality: agencies cannot wait on AI. Instead, they need to understand how it’s being currently used and identify where it provides value.
Encourage staff to share how they are using AI
The next opportunity is to bring program staff’s informal AI use into the open.
Creating simple opportunities for staff to share how they are using AI, such as through informal discussions or team roundtables, can surface practical use cases that might otherwise go unnoticed. It also helps identify what is working, where there are gaps, and what guidance is needed. The people doing the work are the ones who know where AI can help. Creating space for them to share these insights ensures adoption is grounded in real needs, not just technology decisions.
From there, the focus shifts to putting the right controls in place.
Establish guardrails for AI use early
As interest in AI grows, so does the emphasis on governance, privacy, and responsible use. Organizations like the National Association of State Chief Information Officers (NASCIO) have observed that agencies are approaching AI cautiously, and with good reason. Agencies operate under strict requirements around data protection, accessibility, and compliance. Any new capabilities must be implemented in accordance with official standards such as SOC 2 and the Americans with Disabilities Act (ADA).
For that reason, governance and policy need to be part of agencies’ approach from the beginning, so that there is clarity around:
- What data can and cannot be used with AI tools
- When AI outputs need to be reviewed or validated by a human
- Where human oversight is required within a workflow
Focus on data readiness—something that is easy to overlook
While much of the interest in AI focuses on capabilities, the limiting factor is often much more familiar: disparate data. Environmental agencies have spent decades building systems and collecting records, but that information is often distributed across multiple systems and formats.
AI alone does not solve this problem; it depends on reliable data. If data is incomplete, inconsistent, or difficult to access, the outputs will reflect those limitations. When an agency connects its datasets, improves data quality, and standardizes information, it creates a strong foundation for effective AI use.
_____________________
Moving toward a unified data management platform will substantially improve agencies’ ability to leverage AI effectively.
_____________________
Moving forward, deliberately
AI is already integrated into environmental agency operations to varying degrees—the question is whether that adoption is deliberate or accidental. The agencies “getting it right” aren’t the ones with the biggest deployments; they’re the ones focusing on a single, real problem, listening to staff already using AI tools, setting guardrails before scaling, and investing in the data underneath it all.
Submitted by Will Rensmith, Chief Innovation Officer
About Windsor Solutions
Windsor Solutions partners with environmental regulatory agencies across the United States to design and implement software that supports the full regulatory lifecycle. For more than 27 years, we have worked alongside agencies to streamline workflows, improve data quality, and adapt to evolving requirements through configurable, purpose-built technology—including the thoughtful application of AI where it supports program work.
