“We spent nine months trying to make the right decision, and I got it wrong.”
It’s the kind of admission technology leaders rarely make in public, but Michael Lewis, chief technology officer (CTO) at Management Controls used his experience of walking back on a costly project to make a larger point.
In complex ecosystems where multiple players depend on the same system, a single wrong turn doesn’t only waste resources, it risks adoption altogether.
Speaking at the CIO Association of Canada’s Peer Forum in Ottawa, Lewis shared lessons from nearly two decades of building software for industrial giants and their vast contractor networks. His argument was that features alone don’t determine success. The real test is whether owners, vendors, and front-line users all find a reason to use the technology.
This only happens when CIOs enforce rigorous checkpoints and listen for friction before it becomes failure.
Checkpoints that catch missteps before launch
Lewis stressed the importance of forcing structured feedback, even when it slows things down.
His team relies on a quarterly advisory board that reduces bias and keeps ideas grounded in real-world conditions.
He recalled a mobile timesheet tool that seemed airtight to internal experts but collapsed when tested with customers working in remote locations. The advisory board’s rejection saved the company from rolling out a product that was doomed from the start.
“The product management team is going and refactoring that idea, right? But without that checkpoint, then we would have spent calories building the solution that wouldn’t have worked well in the market,” said Lewis.
That discipline matters, he added, because nearly 40% of his employees once worked as customers themselves. Their insight is useful but dangerous if it replaces external validation.
When failure is the only feedback that works
Not every effort succeeds — a fact that Lewis is very candid about. One costly misstep was the aforementioned nine-month reporting tool evaluation that ended in reversal when users insisted Power BI was better than the system the company had embedded.
He recalled the uncomfortable leadership meeting where he admitted the mistake.
“We’re going to live with this solution for a really long time,” he said. “And so unless I come in and tell you guys this and we switch, then we’re going to be living with this pain for years and years and years.”
The decision to walk back months of work underscored a lesson CIOs know but often avoid: sunk costs don’t justify keeping bad systems. Having an internal services group using the software daily gave Lewis the evidence he needed to change direction.
Where AI earns its keep in ecosystems
Ecosystem design also shaped how his team used generative AI.
Lewis described a chatbot that gave vendors real-time support and let authorizers query contract terms directly, which is critical in environments where procurement files often sat in cabinets, out of reach for the people approving work.
Their tool addressed that gap, making help and contract intelligence available on demand in multiple languages. It also reduced friction from high turnover in contractor roles, where training rarely carried over from one employee to the next.
“At first, when we started talking, I was like, ‘Yeah, everybody’s doing a chat bot,’” he explained. “I thought we were just doing it to do it, but then when you understand the business problems you’re trying to solve, it adds a lot more meat to what you’re delivering.”
In addition to earning a CIO 100 award, it showed how generative AI, when tied to specific business problems, can solve real pain points rather than simply chasing trends.
The real cost of lump sum deals
Lewis and his team found further success with AI in reframing traditional contracting. He described a lump-sum project valued at $600,000 for an estimated 10,000 hours of work. Once the data was checked, only 4,900 hours had been logged, doubling the effective hourly rate.
Procurement teams had little leverage to challenge the numbers without evidence. By using generative AI to reconcile contract language, workforce data, and market benchmarks, Lewis’s team gave leaders transparency they hadn’t had before.
“We’ve gone through and figured out that your blended rate should be $61 an hour, not $122.45, and here’s all the by-line item rates that we’re assuming from the market and the skills that we’re mapping to based off of what this vendor had,” he explained.
That insight not only helped owners push back on inflated rates but also helped vendors show compliance with fatigue rules and crew requirements. Both sides left the table better informed.
Leadership in an ecosystem world
The message is consistent. Lewis showed how ecosystems magnify both risk and reward. Products only succeed when multiple groups buy in, which means CIOs must build processes that surface friction early, admit failure when necessary, and ground new technologies in problems people actually face.
“When you have an ecosystem-type product where it takes multiple user communities,” said Lewis. “The importance of these items gets amplified.”
The lesson for leaders is that discipline is not bureaucracy, it’s protection. Guardrails like checkpoints, previews, and candid reversals may feel slow, but they prevent wasted effort and strengthen trust in the system.
Final shots
- Ecosystems raise the stakes for every decision. A feature that works for one group but fails another isn’t just a miss, it’s a system breakdown. The margin for error is narrower when adoption depends on multiple communities.
- Admitting a mistake is cheaper than forcing adoption of the wrong tool. Leaders who can change course early save their teams years of frustration and build credibility by showing that evidence matters more than ego.
- Guardrails like checkpoints and previews can feel like neurotics, but they mean survival. They give leaders space to catch blind spots before they spread, creating room for better decisions in environments where the cost of failure multiplies quickly.
Digital Journal is the national media partner for the CIO Association of Canada.
