Good morning, Madam Chair and members of the committee. Thank you for the opportunity to appear.
My name is Michael Lee. I am the chief strategy officer of UrbanLogiq. UrbanLogiq is a company headquartered in Vancouver with Canadian ownership. We have deployed operational AI decision systems in municipalities, provincial and state ministries, emergency management agencies, and transportation and transit agencies across North America. I appear before this committee to share what a decade of operational deployments for governments by UrbanLogiq has taught us about the conditions under which AI generally improves government decisions and outcomes.
This committee has devoted significant time to the risks of frontier AI with autonomous agents. That scrutiny is certainly warranted, but it risks creating a second, quieter gap: treating all AI as equivalent and allowing legitimate concern about frontier systems to crowd out investment in the responsible use of AI, which is already delivering measurable public benefits. This use is machine and deep learning on structured datasets, with crash records, traffic counts, land use records and emergency incident histories, for example, in order to surface analysis for human decision-makers, who retain full authority over every consequential action.
Three things are true in every deployment by UrbanLogiq of solutions for government agencies.
First, the most valuable work is integrating the data governments already hold, not building AI models. The first question in any AI procurement should not be “What will it predict?”, but rather “What will it take to get our data ready?”
Second, successful deployments are those where staff can trace an insight into their inputs and where officials can defend a decision in a public meeting. Where deployments have stalled, the cause has rarely been the AI. More often, the government was not positioned to act, or citizens held fears that did not reflect what the system was doing. An audit trail is the mechanism through which AI becomes publicly sustainable and accountable.
Third, every deployment has demonstrated augmentation and results. For example, in B.C., land use review across federal, provincial, indigenous and municipal jurisdictions on one auditable platform compresses timelines from months to minutes. In Edmonton, planners move from static reports to dynamic scenario analyses. In San Jose, infrastructure investments are evaluated across traffic, equity, emissions and economic access before decisions are made, replacing months of consultant studies in minutes. In Minnesota, machine learning on incident data, building characteristics, weather and demographics has shifted the fire department from reactive response to predictive action.
These outcomes are not incidental. Decisions about land use, emergency response and infrastructure investments are decisions that define whether communities trust their governments. Accountability with a documented, traceable record is what makes them defensible.
These tools change what is possible. Our recommendations to this committee reflect what field experience has demonstrated as necessary in order to enable responsible, effective AI adoption in government.
Let me highlight four of them. One, invest in data integration before AI model investment. Two, mandatory auditability standards with immutable audit trails, model output lineage and explainability requirements should be procurement criteria. Three, require workforce plans with phased rollouts as a condition of government AI contracts. Four, mandate structural AI governance, including risk-calibrated controls, as a condition of government AI contracts.
The evidence is already in the field: a fire chief acting on risk data before a fire happens, a planner evaluating a decade of infrastructure decisions in an afternoon and a land administrator assessing approvals across jurisdictions in minutes rather than months. That is not what AI might do. It is what responsible AI is doing right now.
We ask this committee to give equal attention to the cost of moving too slowly. Canada needs a compliance framework that allows it to move confidently with AI that is proven, governed and ready, while maintaining appropriate caution around systems that are not. That distinction is the most important one the committee can draw.
Thank you.