Removing Work Is the Real AI Leverage
Like the Internet, the AI revolution touches everything. At first it looks smaller than the hype and slower than the promises. That’s normal.
The Internet spent decades bumping into existing systems before it meaningfully changed how work actually happened. Integration takes time.
AI feels different. Not because it’s smarter, but because it doesn’t need permission to rebuild everything. It can wrap itself around what already exists. It can slip into workflows, habits, and interfaces without demanding a full platform shift.
For a long time, I thought the leverage in GIS was giving analysts better tools. Fancier geoprocessing. Smarter UI. More automation in front of experts so they could do their jobs faster and better.
And that works, to a point. But it’s still additive. More tools. More capability. More surface area.
What I’ve started to realize is that the bigger shift isn’t about making analysts faster. It’s about removing work from them entirely.
Most GIS tickets aren’t hard. They’re repetitive. They’re well understood. They exist because clients don’t have safe, self-serve ways to answer their own questions. So they submit a ticket, an analyst runs a familiar workflow, exports a result, and sends it back.
There’s nothing inherently wrong with that process. But it ties skilled people to routine translation work that doesn’t really benefit from their expertise.
The real opportunity is upstream.
Instead of waiting for the ticket, generate a small, constrained, purpose-built app that lets the client solve their own problem. On demand. Safely. With guardrails.
That includes controlled access to enterprise data when appropriate.
GIS teams already act as data shepherds. They set boundaries, enforce trust, and protect meaning. Turning that stewardship into encoded constraints inside self-serve systems is where AI actually helps.
No ticket. No back-and-forth. No analyst time spent on something they’ve done a hundred times.
This is the vision behind ArcGISPro CLI. It started as a way to compress UI into commands. But the deeper idea is compressing intent into action. Turning “I need to know X” into a system that can answer X without human mediation.
I don’t know what the future of work looks like. But I’m relieved that AI doesn’t have to mean yet another addition to the toolbox. Another new technology to learn. Another interface layered on top of an already crowded stack.
Deployed well, AI doesn’t add another layer. It lets you prune entire branches of the enterprise legacy software tree instead of continuing to grow it taller, wider, and more complex just to support the same questions.
When that works, analysts stop being button-pushers and start being system designers. They define constraints. They design automations. They handle the weird cases, the ambiguous questions, and the genuinely hard stuff. The work they actually enjoy.
This also shouldn’t be marketed as adding AI to GIS. The win is removing software from view. Fewer screens. Fewer clicks. Fewer tools. Less dependency on specialized intermediaries for routine questions.
The pitch isn’t “AI-powered GIS.” It’s simpler than that.
The next wave of AI in GIS won’t be about making analysts faster. It will be about making routine work disappear.
What’s left is the work that actually benefits from expertise.