Article
Inside our AI transformation: What we’d do again—and differently
An honest look at the moments that shaped our AI journey so far—what worked, what we’d change, and what we’re still learning.
Table of contents
Constant change isn’t a phase—it’s the new operating environment
Nearly every AI transformation story implies there’s a destination. Finish your roadmap. Mature governance. Scale adoption.
But AI isn’t moving toward a discrete finish line. It’s changing the nature of work in real time, forcing leaders to continuously rethink not only the tools they deploy, but how people, teams, and entire organizations create value.
In this environment, progress depends less on reaching an endpoint than on building the capacity to adapt.
“If you aren’t in motion, you aren’t building a new muscle,” says Ali Minnick, senior managing director of AI transformation at Slalom.
Our own transformation continues to reinforce that lesson. Every step forward reveals the next challenge: expanding access creates new enablement needs, experimentation demands new operating disciplines, and scaling adoption requires rethinking governance.
“What was written in pen is now written in pencil,” Minnick says. “The challenge isn’t avoiding reaction. It’s reacting quickly and intentionally.”
Even with decades of experience helping organizations adopt new technology, we recognized that the AI landscape demanded something different. The technology is evolving too quickly for anyone to claim they have mastered it all. As Minnick puts it, “If someone tells you they've been doing this for 20 years, run.”
What follows isn’t a playbook for every organization. It’s the lessons we’re taking with us as we continue to navigate the messy middle.
Access is the catalyst
Early access was an opportunity to learn before we had to lead. It allowed us to test our assumptions about how AI would change work, learn what actually drove adoption, and bring those lessons to clients from firsthand experience instead of theory. We wanted to do the hard learning ourselves so our customers wouldn't have to.
“We put an entire team’s learning and development budget into AI tools and just started experimenting,” Minnick says.
As access and experimentation expanded, we turned to our longstanding relationships with technology partners.
“Microsoft Copilot is what really kicked us off,” says Michele Bleser, VP of business transformation at Slalom. “It was our partnership and the early access to what they were developing that really got us moving.”
Our unique positioning as both a partner and a customer gave us a 360-degree view of what Microsoft was working on, allowing us to learn alongside the product teams instead of waiting for the technology to mature.
This early access gave us a front-row seat into the realities of incorporating generative AI (GenAI) into the enterprise. We tested different approaches through six employee cohorts, pairing early access with increasing levels of training, office hours, and support. Each cohort revealed new questions, not only because the technology kept changing, but because deeper training led people to experiment more and uncover new ways AI could reshape their work.
It also taught us our first lesson: providing training wasn’t enough.
"We realized early on that this wouldn’t just be a technical implementation," Bleser says. "We were entering an age of a mindset shift. And mindset shifts require coaches."
What this means for you:
Give executives direct exposure before asking them to set the strategy. Demos and presentations build awareness, but firsthand use creates conviction. They don’t need to become technical experts, but they do need enough experience to understand how AI could change the work they're asking others to transform.
Your people are your strongest scaling tool
In early 2022, Christopher Richardson was working as an organizational effectiveness consultant at Slalom. By the end of that year, he was one of our first AI enablement coaches, working directly with leaders to help them understand not just how to use AI, but where it fit into their day-to-day work and what it could help accelerate.
"It's about how you transform your work, not about how you use the tools," Richardson says. "If we don't change the culture, we're just going to scale silos at breakneck speed."
Driving adoption companywide meant changing habits, workflows, and mindsets—not simply delivering training sessions. As our Copilot rollout reached 90% adoption, we saw the value of coaching in helping people rethink their work. The challenge was figuring out how to scale it across a nearly 10,000-person organization.
"A central office can set the conditions, but it can't be everywhere the work actually happens," says Josh Ritter, a program manager in our AI Transformation Office (AITO). Instead, we identified leaders already driving AI within each capability and helped them build networks of champions and subject matter experts who could coach their peers where work was actually shaped, sold, and delivered. "That's what let us scale, and it's where accountability really took hold."
Deborah Walkoczy joined the effort knowing little about AI herself. Her willingness to experiment ultimately became one of her greatest strengths, and approaching AI from the perspective of a new user gave her a unique lens on the experience many employees were about to have.
As an adoption and enablement lead, she quickly found herself tackling a challenge few organizations had solved.
“How do you make learning materials for a tool that could drop new features tomorrow?” she says. “We’re really on an evolving enablement journey. There is no ‘done.’”
Enablement breaks down when learning can't keep up with AI, she says. Instead of teaching people one tool at a time, the goal was to help them build adaptable skills that carried across new technologies.
"We've created a foundation so that when new tools come in, it's not a shock to the system," she says.
Enabling coaches, capability leads, and employee communities solved one problem: how to help people learn and adopt AI. It also exposed another. If the technology was constantly changing, enablement couldn't rely on a fixed playbook.
The playbook itself had to become an experiment.
What this means for you:
Build for continuous learning, not one-time proficiency. AI skills will expire faster than most traditional learning programs can be updated. The most durable investment is an ecosystem of coaches, subject matter experts, and leaders that help people continuously adapt.
Experimentation only creates value when learning becomes repeatable
We started by experimenting because we had to. We learned to experiment with intention because we couldn’t afford not to.
“The first chapter was a bit of a ‘shiny fish lure’ situation,” says Minnick. “We got really excited about what was ‘possible,’ and lacked consistency in durable value at scale.’”
At first, experimentation felt a bit like the Wild West. Through technical hackathons and exploratory pilots, teams were trying new tools, testing new features, and discovering what GenAI could do.
Creating structure around experimentation allowed us to channel that excitement into something intentional. Hackathons expanded beyond technical teams, and business users were increasingly expected to experiment within their own roles. Instead of chasing every new feature, teams started prioritizing use cases with the most business potential, redesigning role-specific workflows, and noting successful approaches.
“While early experimentation should generate curiosity, mature experimentation should generate decisions,” says AITO senior research manager, Pam Harris.
Over time, experimentation became a disciplined process for deciding what was worth scaling, what needed refinement, and what should be left behind.
Even our customer solutions reflect that evolution. What began as an AI value calculator for individual use cases grew into our AI Value Platform as we realized organizations needed more than a business case—they needed a way to prioritize investments, assess readiness, and continuously adapt their AI strategy. Like the journey itself, the platform continues to evolve.
That kind of experimentation doesn’t happen accidentally. It requires intentional investment and for leaders to establish clear priorities, decision criteria, and dedicated space to learn.
“You need to get rock solid on the 20 percent that you think is going to drive 80 percent of the value,” says Minnick.
What this means for you:
Define what AI will mean for your business. Instead of jumping straight into use cases, be clear about what you're trying to prove and where you want to create value. Then let that vision guide what you test, scale, and stop.
Governance should accelerate innovation, not slow it down
At this point in our transformation, two things became clear: speed without coordination wouldn’t scale, and AI was forcing decisions that no single function could make on its own.
We saw that firsthand as early AI projects made their way through governance. In one case, an internally facing AI application spent months in review before launching. Looking back, leaders agreed the review was more extensive than the situation required. But the experience helped refine our risk posture, creating a more practical approach for future AI initiatives.
“We realized early on that we had to bring people together from across the organization to move quickly,” says Bleser. “That operating model—with go-to-market leaders, executive sponsors, IT, legal, and security working together—became the foundation of our AI Office.”
Rather than owning AI adoption, the AI Office became a source of truth, equipping capability leads, coaches, champions, and business leaders to extend governance and enablement across the organization. That shared visibility also helped identify teams solving similar problems, reducing duplicate effort while bringing more perspectives to each solution.
This cross-functional approach transformed governance from a control mechanism into an ongoing operating rhythm that evolved alongside the technology, allowing the organization to adapt as new models, risks, and opportunities emerged. As Bleser puts it, “This is not a one-time technical implementation.”
Neither was the governance that supported it.
What this means for you:
Establish an AI Office early, but as an enablement tool—not a review board. The most effective AI Offices create alignment across business, IT, legal, security, HR, and enablement so experimentation can happen safely and consistently as conditions change. Governance should accelerate learning, not slow it down.
Your operating model will become the bottleneck before technology does
Our people were ready. Our governance was evolving. But as AI adoption accelerated, we discovered that the organization itself wasn't designed to support what came next.
Teams were wanting to build, deploy, and share AI-powered solutions in ways that existing infrastructure couldn't accommodate. Usage dashboards, token limits, deployment pathways, and internal support models all became new operating challenges that hadn't surfaced during early pilots.
“Scaling exposed the parts of the organization that pilots don’t test,” Bleser says. “We now had teams wanting to host and deploy things that are not IT or engineering teams. It created an infrastructure bottleneck.”
As more employees began building with AI, we needed new ways to share knowledge, support citizen developers, and make information accessible across teams. AI became exponentially more valuable when it could tap into our collective intelligence, but that depended on more than connecting a knowledge graph. It required an operating model that encouraged people to contribute knowledge instead of protecting it.
“Our data guardians need to know that their culture is going to shift to be data stewards,” says Richardson. That requires building a trusted, governed source of truth—one with the security, access controls, and governance needed to support employees, traditional analytics, AI tools, and autonomous agents alike.
What this means for you:
Prepare for employees to become builders, not just users. Once adoption grows, people will create their own agents, workflows, and solutions, whether the operating model is ready or not. Invest early in deployment pathways, reusable architecture, support models, and guardrails so successful pilots don’t become isolated products or sources of risk.
Three principles we’re carrying forward:
1. When certainty isn’t available, intentionality has to be.
If there's one lesson our AI transformation reinforced, it’s that we can’t wait for stability before we act. New models, workflows, and expectations will continue to reshape how work gets done. The goal isn’t to predict every change—it’s to learn and unlearn in faster cycles so we can respond intentionally as the technology evolves. "Measuring AI is messy, but leaders still have to lead," says Minnick. "It’s about making the next best decision instead of the final one."
2. Use what you can measure today to fund what you want to measure tomorrow.
Rather than treating uncertainty as a reason to slow down, we’ve treated it as a reason to measure and mature differently. We’re focused on capturing the signals we can measure today, from workflow improvements and employee adoption to qualitative feedback and emerging patterns of value. Those insights help where we invest next, while tools like our AI Value Platform help ground ROI conversations.
3. Let your customers help shape your roadmap, not just validate it.
Don’t separate internal innovation from customer transformation. The challenges your customers face will pressure-test your assumptions, surface needs you hadn’t anticipated, and reveal which capabilities create lasting value. At the same time, your own experimentation gives you the credibility to share practical lessons from the work you're doing.
Contributors
What would it mean to manufacturers if they could increase their efficiency rates by 5%, 10%, or even 20%? That would put enormous value back into the business. The big question is, what investments will get you there? That’s where we can help manufacturers navigate their options and take steps that will lead to better, faster, more informed decisions about the business.
We are about to enter a period of significant business transformation, primarily driven by emerging technologies that will create step changes in how we work, in business operating models, and in overall performance improvement.
What would it mean to manufacturers if they could increase their efficiency rates by 5%, 10%, or even 20%? That would put enormous value back into the business. The big question is, what investments will get you there? That’s where we can help manufacturers navigate their options and take steps that will lead to better, faster, more informed decisions about the business.”
True advantage in healthcare begins with a clear purpose and grows with intelligence. When organizations learn to reinvent—not just react—they earn the trust that moves people and systems forward.
This year marks an inflection point. Scaling how we work now won’t get us where we need to go. Instead, we need to reimagine how to walk and run. This means reinvesting in data and innovation for predictive pathways, partnering in fluid and creative ways, and upskilling teams for an AI-augmented future.
The disruption government experienced in 2025 is likely to continue. Government leaders can use this moment to simplify, clarify their mission, and connect people, policy, and technology, and they will come out of this period stronger.
AI has moved beyond the proof of concept—it’s now the proof of imagination. The next phase isn’t about smarter models; it’s about smarter organizations that can evolve as fast as the technology does.
AI has shifted from a small, centralized team or a business unit-level effort to a CEO- and board-level imperative. It’s moving from internal-facing applications to external use cases, resulting in significantly improved customer experience and business outcomes. Where technology focused on cost reduction has often been seen as a barrier to a positive customer experience, AI is proving to be an enabler and accelerator in these areas.
If you rewind 30 years, banking systems have always been static. You’d put in a core, and it’d be your core for decades. You’d put in infrastructure, train your people, and they'd use those same policy manuals for years at a time. Now all of that is changing. Fast. And it keeps changing fast. Because it needs to. Open, modern platforms with adaptivity are the key.
