Earlier this year, executives at Providence needed support for core leaders navigating a workforce transition designed to strengthen our culture by bringing teams back into the office. Within 10 days, an internally developed AI-guided conversation agent called My Conversation Companion was repurposed to align with the initiative, tested, moved through governance and deployed to 1,300 core leaders.
From the outside, that kind of speed looks like luck.
It was far from luck. It was two and a half years of building, testing, letting go and rebuilding, so that when the moment arrived, the capability was patiently waiting and ready.
Here’s what I’ve learned: In a volatile, uncertain world, you don’t get to schedule the moment your idea is needed. You only get to decide how ready you’ll be when the moment shows up.
This is the story of how an idea survives changing environments, and why the version that survives will not, and should not, be the version you started with.
The greenhouse
My idea began in early 2024. When ChatGPT exploded onto the scene, many people focused on productivity and efficiency. As a learning professional, I asked a different question: What could it do to help people grow? What if someone could practice a difficult conversation — such as influencing without authority — with an AI-guided coach, in a safe space, before the real moment?
At the time, Providence had no internal tool that could develop that idea. So it started outside, on my personal ChatGPT account, on my own time. Waiting for the right conditions would have meant not starting.
Think of it as a greenhouse. Seedlings do not survive being planted directly into a harsh climate. They need a protected space to grow into something viable, free from the demands of the enterprise environment. Once something starts to thrive, it can be moved outside and tested to see whether it survives the real climate.
The greenhouse only works with two hard guardrails. First, nothing proprietary ever goes out. No organizational data, no protected health information, no personally identifiable information. None was ever included, and none was ever needed. My experiments were generic patterns of human conversation, such as giving effective feedback. What grew in the greenhouse was a pattern, not a product.
Second, nothing comes in until there is sanctioned soil for it. The pattern waited outside until Providence had an approved tool that could receive it. The greenhouse is not shadow IT. It is a protected space where responsible innovation survives its infancy, and it comes inside through the front door.
Four more environments
When ProvidenceChat, an internal instance of an AI chatbot, became available later that year, the pattern came inside and was rebuilt as an uploadable PDF with the prompt inside it. It was functional. People could have a guided conversation. But it could not scale, because there was no easy way to reach the audience with updates or messaging.
What survived was the pattern and a lesson: The experience was viable; the delivery mechanism was not.
In early 2025, Copilot arrived and ProvidenceChat was sunset. I rebuilt the concept as a Copilot agent. The user experience got better, launching from an agent instead of a PDF, but agents could not yet be deployed at scale, and the governance to enable that was still being formed.
The idea evolved again: validated on an enterprise platform, but not yet ready to scale.
When Copilot Studio became available in late 2025, I rebuilt the concept a third time, and this time it could be deployed at scale through governance. Now the barrier was not technical at all. Getting through governance meant earning executive sponsorship and being prioritized over every other competing use case.
The idea evolved and was ready. The question was no longer, “Can this work?” It was, “Why this, and why now?”
The answer changed the idea more than any platform shift had, and it is a story worth its own telling. In short: A tool I built as learning and performance support turned out to be generating signals executives had no other way to get. That reframe is what earned sponsorship and moved us through governance.
First, though, notice the pattern. Each environment forced the idea to adapt: new delivery mechanisms, new rules about scale, new questions about governance and value. That is the mindset this journey asks of you. Your idea is not a fixed thing you defend. It is a living thing you evolve. If you cannot let go of the version you have, your idea will not survive the pressure test of enterprise scale.
The pause
Then came the part of the journey few people write about — the pause.
The experience was built. The concept was validated. The reframe was in hand. And nothing moved. There was no initiative asking for what we had, and other priorities were ahead of us in line. Most innovation stories have a pause like this, and it is where most ideas quietly die.
What kept this one alive was that the waiting was never idle. We kept testing. We worked out the messaging, the SharePoint site, and learned what did not resonate. We developed additional topics with owners across the enterprise.
None of it looked like progress from the outside. All of it turned out to be the preparation that made those 10 days possible.
The moment arrives
Early this year, the moment showed up. Providence launched the workforce transition, and executives needed a way to support the 1,300 core leaders who would be having those difficult conversations with their teams. The need was real, the timeline was short, and we were standing there with something built, tested and ready to pivot.
Within the 10 days it took to move through governance, we repurposed My Conversation Companion to the new topic, tested it and deployed it. The messaging from the pause went live with it, on a SharePoint landing page focused on showing up in the moments that matter and living into our Promise (Know Me, Care for Me, Ease My Way), rather than on introducing a new AI tool. And we deliberately capped the deployment at those 1,300 core leaders, because no agent with these capabilities had ever run inside the system, and a controlled first deployment was the responsible one.
Among leaders who engaged deeply enough to complete a scenario or explore the tool’s fuller capabilities, My Conversation Companion earned a Net Promoter Score of +54, which lands firmly in excellent territory on a scale running from -100 to +100. The sentiment behind the number mattered more than the number.
Enthusiasm clustered arou
