AI & Government January 28, 2026

NYC Spent $600,000 on an AI Chatbot That's "Basically Unusable" — Here's What Went Wrong

New York City Mayor Mamdani puts the Adams administration's AI spending under the microscope, raising critical questions about how governments — and organizations of all sizes — approach artificial intelligence projects.

"He cited the Adams administration's roughly $600,000 spend on an artificial intelligence chatbot he described as 'basically unusable' as the kind of project that will go under the microscope. The mayor said he has to show that city government is committed not only to public services but also to 'public excellence and public efficiency,' and that 'every dollar that's being spent is actually being spent in a worthwhile way.'"

— Mayor Mamdani on the Adams administration's AI spending

The Problem Isn't AI — It's How It's Built

A $600,000 investment in an AI chatbot that doesn't work isn't just a budgetary failure — it's a symptom of a deeper problem. Too many organizations, both public and private, treat AI as a plug-and-play solution. They buy software, deploy it, and expect results. But AI without proper architecture, data strategy, and user-centered design is just expensive software that nobody uses.

Why AI Projects Fail

The pattern is consistent across failed AI initiatives, whether in government or the private sector:

  • 1. No clear problem definition. The project starts with "we need AI" instead of "we need to solve this specific problem." Technology should serve a purpose, not the other way around.
  • 2. Poor data foundations. AI is only as good as the data it's trained on. Without clean, structured, and relevant data pipelines, even the most sophisticated models produce garbage outputs.
  • 3. No iteration loop. The best AI systems are built iteratively — deployed in small scope, tested with real users, refined, and scaled. A single big-bang launch almost always fails.
  • 4. Vendor-driven, not outcome-driven. When procurement is driven by vendor promises rather than measurable outcomes, you get polished demos that fall apart in production.

Public Excellence Requires Technical Excellence

Mayor Mamdani is right to demand that "every dollar that's being spent is actually being spent in a worthwhile way." That standard should apply to every AI initiative — in government and in business. The organizations that succeed with AI share common traits: they start with clear objectives, invest in data infrastructure, hire (or partner with) teams that understand both the technology and the domain, and they measure results ruthlessly.

The Path Forward

AI can deliver enormous value when implemented correctly. The technology itself isn't the problem — it's the approach. Whether you're a city government serving millions of residents or a growth-stage company looking to automate operations, the principles are the same:

Define the outcome first

What specific problem does this solve? What does success look like? If you can't answer these clearly, you're not ready to build.

Build on solid data

Invest in your data infrastructure before your AI models. The foundation determines the ceiling.

Start small, iterate fast

Launch an MVP with real users. Collect feedback. Improve. Scale what works. Kill what doesn't.

Measure everything

Track adoption, accuracy, user satisfaction, and ROI from day one. If it's not measurable, it's not manageable.

The $600,000 lesson from NYC is one that every organization should learn from — not to avoid AI, but to demand better AI. Technology built with clarity, rigor, and accountability delivers results. Everything else is just expensive noise.

Building AI That Actually Works?

At Orion Digital Platforms, we build AI systems grounded in solid architecture, clean data, and measurable outcomes. No hype — just results.