AI adoption starts with solving real problems, not hype\” />

Louis Savard, CIO of the City of Cornwall, (left) and Robbie Beyer, director of data science and AI at RSM, at the CIO Association of Canada’s Peer Forum in Ottawa. – Photo by Jennifer Friesen, Digital Journal
Twelve years. That’s how long one service request sat unresolved in Cornwall’s system. The discovery prompted city leaders to rethink their entire approach to case management, leading to a new platform built to support AI from the start.

“We really realized the power of the tools we already had that we didn’t even know we could use,” said Louis Savard, CIO of the City of Cornwall. 

Savard shared the story during a conversation with Robbie Beyer, director of data science and AI at RSM, during the CIO Association of Canada’s Peer Forum in Ottawa. Their discussion focused on the leadership choices that turn pilot projects into lasting change.

Across Canada, municipalities are under pressure to modernize citizen services while operating with lean teams and tight budgets. Cornwall’s experience shows that AI adoption doesn’t have to start with expensive, futuristic projects. AI adoption can begin with the tools an organization already has. The lesson is to pinpoint the pain points, start small, and build a culture of trust that allows technology to scale.

Cornwall’s answer to its service-request backlog was Cornwall Connect, a centralized case-management platform built on Microsoft Dynamics. Staff can now intake and assign cases, and the city is preparing to embed AI for both citizen-facing and internal use. 

The aim is straightforward. If someone reports a fallen tree branch, an agent can immediately flag that there are already multiple requests and that it is being handled. On the back end, staff want Microsoft Copilot to accelerate knowledge transfer, such as surfacing how often a problem is reported or whether it is a repeat report.

A first chatbot, named CiCi, launched in June with a narrow focus on housing, Ontario Works and childcare, which are the three topics that generate the most citizen inquiries. 

The project has been in development for about 10 months in partnership with RSM. Savard shared that while it could have been launched years ago, he said the city deliberately aligned it with demonstrated demand.

By starting small, Cornwall can test how residents interact with the chatbot before expanding its capabilities. Early versions will answer common questions and direct people to relevant services, but will not yet help complete applications. 

Savard said this phased approach will allow his team to measure uptake and refine the service before adding other areas such as economic development, bylaw enforcement and recreation. 

“We’re following what the need was, and we want to see what the uptake is,” he said. “If the uptake is there, then we’ll expand to other departments.”

The shift is internal as well as external. 

“We’re looking at what we’re building on the outside and rebuilding it on the inside for our workforce,” Savard said. 

With only two people answering phones for more than 40,000 residents, the city is exploring internal agents to handle routine questions instantly and reduce voicemail bottlenecks.

Start with real pain points

Beyer said successful AI programs usually begin with quick, provable wins, often in personal productivity. 

“Going from those low hanging fruit, those quick wins to demonstrate value, to moving up a chain that’s a little bit more complex, but there’s really some high ROI to be able to provide for that,” Beyer said.

Minutes saved per person each day add up across teams and help justify broader investment.

One early surprise in Cornwall was who championed Copilot. 

One of the strongest advocates for Copilot in Cornwall turned out to be the city’s CFO, who had long preferred paper-based processes. 

After seeing the tool summarize a complex file in minutes, she began using it regularly for document summaries and email drafting, saving hours each week and urging colleagues to adopt it as well. Savard said her shift in perspective has been a powerful example of how quickly AI can change day-to-day work when it addresses a real need.

Listening, not prescribing, guided the next steps. 

A conversation with HR revealed that incomplete onboarding details were causing delays in payroll and benefits. 

Savard’s team built a simple Power App to trigger notifications as soon as forms were submitted, which quickly created demand for more automation. He added that his team follows what he calls the “rule of six.” 

The approach is to ask “why” up to six times when investigating a problem, drilling past surface complaints and temporary frustrations until the root cause is revealed. Fixing that root, he said, prevents the issue from returning, whereas quick fixes to early symptoms will only offer short-term relief.

Beyer stressed the importance of defining the use case and the outcomes to track from the very start. 

“What are the expected outcomes we expect to track and achieve with this,” Beyer said. “It’s really important to make sure that you’re following that along the way.”

For leaders, that means treating AI projects like any other strategic initiative. Document the challenge you are addressing, set clear metrics for success and commit to measuring against them as the project unfolds. By doing so, teams can quickly pivot if the results aren’t matching expectations and avoid wasting resources on the wrong approach.

Beyer added that this discipline also pays off when it comes time to secure funding. Having hard numbers to demonstrate time saved, error rates reduced, or service quality improved makes it easier to justify further investment. 

Without that evidence, leaders risk what he called a “scatter-shot” approach to AI, a series of disconnected experiments that create excitement but fail to build momentum.

Build trust, then scale

Fear of job loss is real, and Savard did not shy away from that point. 

“It’s a reality,” he said, though his emphasis is on moving people from repetitive tasks into roles that steward data and improve systems. 

He reduces resistance with familiar examples, like asking people to unlock their phones and pointing out that the thumbprint recognition they just used is a form of AI already in daily life.

Governance and privacy underpin adoption, Savard said. Cornwall keeps all AI data in its own secure Microsoft “tenant” — essentially a private, walled-off environment for the city’s systems and files. 

Only staff with the right role-based permissions can access specific data, and every new tool goes through a privacy impact assessment before launch. Public chatbots are limited to pre-approved, non-sensitive sources such as FAQs and website content.

Beyer described a similar approach he recommends to clients: set up a private copy of a generative AI model using Microsoft’s Azure AI Foundry, Microsoft’s platform for creating and managing custom AI models in a secure environment, train it only on the organization’s data, keep it inside that same protected environment, and secure the connection points for anything the public can access. 

He said designing with these safeguards from the outset makes it easier to meet security requirements while still delivering value.

Beyer emphasized that security and governance need to be designed in from the start. Building safeguards early makes it easier to meet requirements and still deliver value. 

More importantly, it establishes the trust that allows AI projects to expand. Without clear rules, permissions and protections, even the most innovative tools risk rejection.

From efficiency to engagement

Moving service intake online has improved accessibility. 

Residents are no longer bound by nine-to-five phone hours, and public works staff can respond in real time. Early feedback has been striking. 

“I was blown away that 93% was positive,” Savard said of post-case surveys that focus on platform experience. Community groups have since asked to participate in the next iteration so niche needs are reflected in future features. 

“We’re going to bring them to the table to say we want to hear from you while we’re going through our development cycle.”

For both leaders, the throughline is to start with a specific problem, prove the result, measure what matters and bring people along. Do that, and the apprehension that often surrounds AI fades as users see their own pain points solved.

It’s the same lesson Savard drew in Cornwall. The real power of AI often comes from using tools already within reach, applied to the problems that matter most.

Final shots

Solve a real, high-value problem first, and measure the results so you can prove the case for scaling.

Turn early adopters into champions who influence peers and shift culture.

Build security and governance into AI from the start to maintain trust inside and outside the organization.

Digital Journal is the national media partner for the CIO Association of Canada.