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 around the human core of the experience: One leader called the coaching feedback “a real eye opener,” and others described new confidence to navigate conversations they would otherwise have faced unprepared. The detractors were not rejecting the concept. They were reacting to friction, describing a tool that felt too slow or asking for the ability to speak to it rather than just type. Several of the most engaged users pushed not for less of the product but for more of it, asking us to make it available to the rest of their teams.
Read together, the score and the sentiment told a consistent story: When the experience works, it works powerfully, and the path to an even higher score runs through refining the delivery rather than reinventing the design. That is exactly the signal you want before asking for more resources.
The prototype is step one of five
Right now, many talented people in learning and talent organizations are building agents. Here is the hard-won news: The first working prototype is only step one of a five-step journey, and each step is a different discipline.
- Idea and functional build: Grow it in a greenhouse if the enterprise has no soil for it yet, with the guardrails detailed above in place.
- Pressure-testing with real people: Build enough to share, then test, refine and repeat. Follow the energy and resonance of people who use it.
- Additional executive alignment: Find where the idea meets additional business needs from various executives, and be ready for the answers to transform the idea itself.
- Governance as a priority: Sponsorship is what moves a use case ahead of others. Governance will ask questions the prototype never had to answer: Who supports this? How is it documented and updated? What happens when it breaks?
- Adoption and sustainability: Deployment is not adoption. The idea now needs messaging, change management, impact measurement, continuing feedback and evolution.
If you are the executive rather than the builder, you have a role at every step: Protect the greenhouse without demanding immediate returns, be the leader whose frame a builder can step into, sponsor the idea before it is completely safe, and clear the governance path once the case is proven. Builders create the capability. Leaders create the conditions.
One more insight on that last step. My first messaging focused on the technology, and it did not move anyone. What never worked was “AI-guided conversations.” What did work was to put the person first: “Show up in the moments that matter, and be empowered to live our Mission and Promise.” That work was done in the pause, which is why, when the 10-day window opened, it could go live so quickly. Preparation hides in every step of this roadmap.
The view from the sponsor’s chair
Contributed by Sandra Murray
As an executive sponsor, I do not start by asking whether a technology is interesting. I start by asking whether it solves a meaningful business problem and whether it helps our leaders show up more effectively in the moments that matter. My Conversation Companion did both.
What made it different was that it addressed a challenge we have struggled with for years: How do we create meaningful skill practice for leaders around sensitive, high-stakes conversations? Role plays, simulations and facilitated practice during in-person events are valuable, but they are limited by time, scheduling and scale. My Conversation Companion brings that practice into the flow of work. It meets leaders where they are, when they need it.
When the workforce transition emerged, members of my own team were affected, and I used the tool myself to prepare for conversations I knew would be difficult. It gave me a safe place to move from reaction to thoughtful action, to anticipate emotions, consider perspectives and practice navigating those moments with compassion. That experience reinforced what I already believed: The risk of doing nothing was greater than the risk of trying something new. Leaders were being asked to have conversations that could significantly affect caregivers’ lives. The technology was innovative, but the business need was urgent.
I often say crisis creates opportunity. The real work is being prepared when the opportunity appears. That readiness is why this idea was able to move so quickly when the need emerged.
Creating your own luck
You cannot control when the environment shifts or when the moment of need arrives. You can make sure that when it does, you are standing there with something built, tested and ready to pivot. From the outside, that looks like luck. From the inside, it is a strategic practice of creating your own luck, and anyone willing to take the whole journey can build it.
What we did not expect was where the journey would lead. The tool we designed to help people practice turned out to be an instrument executives could navigate by — one that provided real-time signals no other tool provided. That is the next part of this story, and we will share it soon.
