I can’t help but wonder why most of software / tech co-founders I talk to regularly are always asking me the wrong question about AI.
They’re asking and worrying about: “Will AI replace what we are building?”
To me, the real question is: “What part of what we are building not only survives but thrives not if, but when AI gets 10x better?”
I recently came across a framework which broke this dilemma down into five things AI can’t replace, no matter what:
- Trust
- Context
- Distribution
- Taste
- Liability
It’s a useful lens, read more to see why.
What matters to me is how we show up inside real municipal operations, not in theory, not in Silicon Valley, not on Wall Street, but in the field, because that’s where we shine.
1. Trust
In local government, trust isn’t a UX detail.
It’s legal, operational, and even political.
If a mission critical operation is presumed completed when it wasn’t,
that’s not a bug.
That’s liability.
That’s a citizen complaint which leads to a competence discussion.
In the age of AI, software shouldn’t “generate” work. It has to evovle to a point where it helps overloaded teams complete it.
Who should go, where, why and when ?
What is the plan ?
Where are we at on that ?
Why are we being told it wasn't done ?
How did we prioritize this ?
That layer doesn’t get replaced by better AI, it gets improved. It give already great managers and collaborators superpowers.
2. Context
AI without context is just a chatbot.
Municipal operations are nothing but context and nuances:
- specific activities
- distinct streets & locations
- union agreements
- subcontractor agreements
- equipment constraints
- weather conditions
- historical decisions
After 20+ years in the business, there aren't many things I can be sure of, but one of them is that every city runs differently. There is no one-size fits all.
In that essential context, software isn’t valuable because if its features.
It’s valuable because it becomes the operational memory of a city, when implemented properly.
Service routes, infrastructure planning, operational coordination, event management, decision support, analysing historcal operational data. That’s what makes AI useful later.
Without a solid operational foundation to start from, AI is just guessing.
3. Distribution
In GovTech, distribution isn’t sales teams, marketing initiatives, tradeshows, ads, seo or hiring influencers.
It’s:
- education
- relationship building
- demo's
- internal champions
- procurement cycles
- field adoption
- union buy-in
- training crews at 5:30am
You don’t “launch” into a municipality with lots of fanfare, you slowly earn your way in.
And once you’re in, making sure you're staying for the long haul is just as critical:
- Does the new process solve the problem it was supposed to ?
- Is it being used ?
- Is it appreciated ?
- Is it improving iteratively, all the time ?
Answer no to any of these 4 questions, and you’re out. It’s not a question of if, but when.
AI can help, but it won’t solve any of these issues.
4. Taste
(or what I’d call operational judgment)
AI can suggest, but it can’t decide what should be done on a bad moment at 2:30 AM during an epic February winter weather event..
Should we redeploy crews?
Do we prioritize main arteries or residential streets?
Do we send a second pass or conserve salt?
That’s judgment built from experience, powered by AI, but not decided by it.
That layer is human.
5. Liability
This is the one I have yet to see anyone properly understand.
Accountability in the public sector is pivotal. So much so, that it’s the cause for much if the inertia and analysis paralysis everyone criticises governments for doing all the time.
When something goes wrong, managers can’t blame a system. They answer for it.
It’s the very reason municipalities aren’t commonly early adopters of the latest “cool tools”.
They are laggards, cautiously observing, testing and then slowly adopting systems they know they can stand behind.
AI does indeed exist in that reality: Audit trails, Traceability, Defensibility.
“The AI said so” is not a defense.
What comes next
AI is and will continue to make it easier to build software.
But building good software that solves problems, which people enjoy using, and evolves with user needs, that is not where AI thrives.
To be perfectly honesty, and perhaps a little blunt, building the software was never the hard part.
In mission critical public sector operations, that challenge is, and always will be:
- earning trust
- grounding plans and roadmaps in real context
- delivering progress fast and often
- making sure things are adopted in the field
- leveraging and embedding human judgment
- and standing behind it when things go wrong
To me, this is where technology partners in public sector thrive, average ones survive, and those which had a strong pitch and no follow-through disappear, or get absorbed into oblivion.
AI is a powerful tool can and will help people solve problems more efficiently.
