AI in Permitting: Two Models for AI in Government, and Why the Choice Matters
April 13th, 2026
Two product launches in the past month show where enterprise AI in government is heading — and the approaches couldn't be more different.
cBrain, a Danish govtech company, launched its F2 platform with agentic AI capabilities for government case management in the US market. F2 is purpose-built for government workflows: it handles document routing, case decisions, and inter-agency coordination with AI integrated into the process layer. The system is designed so that AI actions are auditable, role-scoped, and embedded in existing approval chains. Government staff don't interact with a chatbot. They interact with their case management system, and the AI operates within it.
Microsoft took the other path. Copilot is now available in government cloud environments (GCC and GCC High), bringing the same general-purpose AI assistant that works across Word, Outlook, and Teams into agency workflows. The pitch is familiar: summarize documents, draft emails, answer questions from your files. It's a horizontal layer that sits on top of whatever you're already doing.
Both approaches will find buyers. But for permitting, the distinction matters.
Environmental review is a structured process with defined roles, sequential dependencies, and legal accountability at each step. A reviewer doesn't just need help writing — they need the right documents routed to the right specialist at the right time, with a record of who reviewed what and when. That's a workflow problem, not a writing problem.
Purpose-built tools like cBrain's F2 (or CEQ's CE Works, which launched its BLM pilot this month) embed AI into the process structure. They know that step three can't start until step two is signed off. They know that a biological assessment needs to route to a specific resource specialist. They know that the decision record needs to include who made the call and under what authority.
General-purpose assistants don't know any of that. They'll happily summarize a biological assessment or draft a finding of no significant impact. But they can't enforce the sequence, track the dependencies, or ensure the right person reviewed the right thing. That's not a criticism — it's a description of what they're designed to do.
The risk for agencies is adopting the tool that's easiest to buy (Copilot is already in most government Microsoft licenses) rather than the tool that fits the work. Document summarization is useful. But if the problem is coordination, routing, and decision tracking — and in permitting, it usually is — then a general-purpose assistant addresses the symptom, not the cause.
There's a pragmatic middle path. Some agencies will use general-purpose tools for the parts of permitting that genuinely are unstructured (drafting public notices, summarizing comment letters, answering applicant questions) while building or buying structured tools for the review process itself. That's a reasonable strategy, as long as the two layers actually connect. If they don't, you've added a tool without reducing the coordination burden.
The question for any agency evaluating AI right now isn't "should we adopt AI?" It's "which tasks are structured enough to need purpose-built tools, and which are unstructured enough that a general assistant will do?" Getting that boundary wrong is expensive, and most agencies won't get a second chance at the budget line.
What to watch: cBrain's US government traction, Microsoft's GCC High adoption numbers, and whether any federal agency publicly draws the line between process-layer AI and assistant-layer AI in a procurement document. That document, when it appears, will shape the market.
cBrain F2 Launch · Microsoft Copilot GCC · CE Works Pilot
AI in Permitting is a recurring column on Permitting Tech covering how artificial intelligence is entering permitting workflows. Written by Boon Sheridan.
