On July 31, the White House Council on Environmental Quality held its inaugural Permitting Innovators Expo, putting selected technology teams in front of the federal agencies working to modernize environmental review and permitting. It’s the clearest signal yet that permitting modernization has moved from talk to action, the solutions shown there will be compiled into a catalog sent to agencies later this year.

Juume AI was there. We were the only startup invited to the closed-door pitch session with federal agencies, and the only team to bring a live, working demo into that room rather than slides and promises. We left convinced of something we’d like to put to the rest of the field, because we don’t think it’s settled: permitting was never really a paperwork problem, and the more we treat it like one, the longer the timelines stay stuck.

That’s an argument, not a verdict. Here’s how we got to it, and where we’d genuinely like to hear whether practitioners see it differently.

The abundance problem

Start with a number worth sitting with. Boon Sheridan, who spent four years researching permitting technology inside the federal government’s digital service, went looking for how many permitting tools actually exist. He expected scarcity. He found more than 300, and calls that an undercount. His conclusion, in the title of his report, is hard to improve on: we have a tool for everything, but a system for nothing.

The tools aren’t the shortcoming. Most of them already can connect, the majority have APIs. They just don’t, because each one gets built as its own island, tracking its own slice of the process. And yet major infrastructure projects still take four to seven years. If more tools were the answer, permitting would have been solved several times over by now.

So here’s the first question we keep coming back to: if the sector has spent a decade adding tools and the timelines haven’t moved, is the next tool really the thing that’s missing? Look closely at where most of those 300 tools point, and a pattern emerges, they sit downstream, after a project is already deep in the pipeline, tracking documents, deadlines, and evidence. Downstream tracking is useful. It’s also not where the time is lost.

Where permitting actually breaks

This is the part the paperwork framing misses. A permit is not a form. It’s the end of a long chain of judgment that starts well before anyone opens an application: figuring out which permits a project even needs, which agencies and codes apply, which precedents matter, what has to happen in what order, and where the dependencies hide.

Get that beginning wrong, miss a requirement, misread how a rule applies to your specific site, sequence the filings in the wrong order, and no amount of downstream speed recovers the months you lose. In our experience, the expensive failures in permitting are almost always upstream failures, discovered late. You can digitize every form in the process and still be slow, because the delay was baked in before the first form existed.

It’s also why permitting doesn’t behave like the assembly line everyone wants it to be. It isn’t a straight pipeline you fix by speeding up each station. It’s a web of interdependent decisions across multiple agencies, where a determination in one place reshapes the question in another. Straightening a line is an engineering task. Coordinating a web of judgments is a reasoning task. They’re not the same problem, and tools built for the first tend to fail at the second.

What we brought to the table

We built Juume to work at the beginning of that chain, not just the end, and that’s the part we demoed for federal agencies.

Our platform, CapraOne, runs on JUNO: an orchestration engine, not a chatbot. JUNO reads the rulebook, scopes what you’re actually asking for, and routes the work to specialized agents, carrying context between them so nothing gets repeated or dropped. The piece we care most about is what happens before permitting starts. Our Permit Agent takes a project’s details, what you’re building, where, its size, land use, interconnection plans, and maps them against the applicable codes to pinpoint every permit, form, and regulatory hurdle, then organizes the steps, filings, and dependencies into a single permitting matrix. That’s the upstream work most tools skip, and where we think the timeline is won or lost.

From there the same system carries through the rest of the lifecycle: requirements become workflows, obligations become assigned tasks, assessments get drafted from your own policies and evidence, and every milestone and risk flag links back to the source document it came from. That traceability isn’t a nice-to-have. In regulated work, an answer you can’t trace to evidence is a liability, not an efficiency, so JUNO operates inside strict context boundaries, grounded in your actual documents rather than generic guesses.

None of this replaces the expert in the chair. The goal is to spend scarce human judgment on the decisions that actually need it, instead of on reconstructing which permits apply from scratch every time.

Why this hits hardest for lean teams

The teams who feel all of this most are the ones with the least room to absorb it. Rural utilities cover 56% of the U.S. landmass while serving about 12% of the population, and they answer to the same federal, state, and regional rulebook as multi-billion-dollar urban operators, often with single-digit compliance teams. For a team like that, a missed requirement isn’t an inconvenience; it’s a budget-breaking fine, discovered at the worst possible time. If you work on one of those teams, we’d especially like to know whether this matches your reality or misses it.

The open question

The federal government isn’t studying permitting modernization anymore; it’s moving. That creates a fork. In one direction, the sector spends the next few years digitizing forms and adding tool #306 to the pile, faster paperwork, same timelines. In the other, teams fix the part everyone skips: the upstream process where permitting actually gets shaped.

We’ve placed our bet on the second path, but we’re not claiming to have it fully solved, and the cross-agency coordination problem Sheridan describes is bigger than any single platform. That’s exactly why we’re writing this instead of just pitching. If you’re inside an agency, running compliance for an operator, building in this space, or simply think we’ve got the diagnosis wrong, we want to hear it.

Permitting was never a paperwork problem. The path from regulation to revenue runs straight through the part the industry keeps ignoring, the beginning. We think that’s where the conversation should start, too.

Working on permitting or compliance and have a perspective on this? We’d like to talk, reach out to the team at Juume AI. This piece reflects our view as participants in CEQ’s 2026 Permitting Innovators Expo, and references Boon Sheridan’s report “We Have a Tool for Everything, But a System for Nothing” (Environmental Policy Innovation Center, 2026).