AI in city managent

 

The Job Has Outgrown the Tools: Why Council Oversight Needs AI

Watching this council term, it's tempting to look for a villain. A councillor who isn't trying hard enough. A staffer who dropped the ball. A process someone designed badly on purpose. It's a satisfying story because it has a fix: replace the person, and the problem goes away.

It's also mostly wrong.

What I actually see is three separate, unglamorous problems stacking on top of each other — and none of them get solved by electing a more virtuous individual to do the same job with the same tools.

Three failures, not one

Staff capacity is uneven — because it's always uneven. Any organization the size of the City of Ottawa has strong performers and weak ones. That's not a scandal. It's a normal distribution. The real problem is that there's no systematic way to catch where capacity gaps intersect with the decisions that actually matter — the ones with nine-figure price tags and multi-year timelines, like Ādisōke, the new Civic Hospital campus, or Lansdowne 2.0. A gap in a low-stakes file is a rounding error. The same gap in a high-stakes file becomes a public inquiry two years later.

Councillors don't control the information they act on. They get what staff decides to put in front of them, framed the way staff frames it, on the timeline staff sets. A councillor with every intention of providing real oversight still can't act on what they never see. This isn't a character flaw. It's what happens when your only lens on a file is the report someone else chose to write.

Individual capacity is real, and it isn't a moral failing. Councillors are generalists asked to render expert-level judgment across planning, finance, transit, policing, environmental policy, and procurement — with a small staff budget and a punishing schedule. Expecting every councillor to personally out-analyze a professional staff report, on every file, every week, isn't a standard. It's a fantasy that lets us avoid asking why the system depends on it.

Put those three together and you get a council that looks paralyzed, staff who look inconsistent, and a public that assumes bad faith — when the actual diagnosis is bandwidth. Nobody in this system, staff or council, has the capacity to see everything at the scale the city now operates at. ($5.2 billion cad!)

What I already found by hand

I didn't come to this argument in the abstract. Earlier this year I ran an analysis of the city's 311 open data looking for patterns in how service requests get handled. What I found wasn't a ward-level story. It was a data integrity story: over 123,000 tickets bulk-closed in a six-day window. Not resolved — closed. That's the kind of anomaly that's invisible in a spreadsheet review and completely obvious the moment you run the numbers systematically.

Nobody assigned me to find that. I found it because I looked, and looking at data of that volume by hand is exactly the kind of work a human being should not have to do to catch a problem that size.

That's the argument in miniature. The oversight gap on this council isn't a motivation problem. It's that the tools available to councillors and their small staff teams are built for a city half this size, moving half this fast.

What AI-assisted oversight actually looks like

Not a chatbot bolted onto the city website. Not a generic "innovation" line in a platform. Specific, boring, auditable applications:

• Automated variance tracking on capital projects — flagging budget and timeline drift the moment it happens, not in a post-mortem report two years after the fact.
• Pattern detection across service data — catching data integrity problems like the 311 bulk-closure before they become a footnote nobody notices.
• Plain-language summarization of staff reports — so councillors, and the public, can actually parse what they're voting on instead of relying on the executive summary staff chose to write.
• Systematic cross-referencing of votes against public statements over time — the kind of accountability check I've been doing manually, file by file, that should be routine infrastructure instead of one person's spare-time project.

None of this replaces judgment. It replaces the manual grunt work that currently stands between a councillor and having enough information to exercise judgment in the first place.

The obvious objection — and why it doesn't undercut the case

AI in government can absolutely become another opaque process, bolted on without transparency, that nobody outside city hall can question. That risk is real, and I'm not interested in trading one black box for another.

The answer isn't to avoid the tool. It's to insist that it serves the open-data and accountability framework rather than replacing it — every flag it raises has to be traceable, every method has to be published, and the output has to be something a member of the public can check, not just trust. If it isn't auditable, it isn't accountability. It's just a new place to hide.

The point

This council isn't struggling because the people in it are uniquely unqualified or uniquely lazy. It's struggling because the job of overseeing a city this size, with this much capital spending and this much data, has outgrown the manual tools we still expect councillors to use. Fix the tools, and you fix a problem that no amount of finger-pointing at individual councillors or individual staffers will ever touch.

That's the platform. Not "trust me more than the other guy." Build the infrastructure that makes trust unnecessary, because the numbers are checking themselves.