The Frontier Firm Playbook: What Separates AI Leaders from Laggards

By Michelle Lancaster, General Manager and Partner, Global Leader for Business Strategy, AI Business Solutions at Microsoft

Most organizations are asking the wrong question about agentic AI. They want to know which agents to build. The successful ones are asking a different question: where should agents fit into our existing workflows?

This distinction matters. More than you might think.

Start With Apps, Not Agents

Here’s what we’re seeing with customers like Lumen and others who are getting real results. They’re not starting with agents. They’re starting with apps.

The future is agents. We know that. What apps currently do will eventually be done by agents. But right now, apps are the right starting place.

Why? Because when you start with apps or the Power Platform, you’re already working with workflows. You have people who understand low-code environments. You have connections and processes already mapped out. You have inherited security and governance structures.

When customers start with agents natively, they miss all of that. They haven’t identified the right users. They haven’t mapped the security paths. They haven’t identified where agent reasoning adds value across the workflow.

Most of that work is already defined inside your existing app stack. Power Apps gives teams a natural language way to reason through it. And when you attach agents to apps, you hit the go-fast button. But that speed depends on having the right foundation in place first.

Vision Without a Roadmap Is Just a Dream

We work with customers who want to be deliberate about every step of their agentic journey. They invest early in governance and controls and are eager to realize ROI as the supporting infrastructure comes together.

Being long on vision without a tactical roadmap is problematic. It creates friction instead of momentum.

Then there are customers who nail an impressive proof of concept. They build an agent. They stand it up and see it work. And when they try to roll that success out across a department, progress stalls.

The successful ones are the ones who break through by knowing what business outcome they want to drive. They start with their existing systems. They’re determining, very deliberately, which parts of those workflows are best handled by agents, and which are not.

Let’s take CRM systems as an example. The best way forward isn’t to rip them out and replace them with agents. It’s to be precise about where the work is best suited to agents, where supervision still matters, and where that work happens.

We’ve seen this firsthand at Microsoft as customer zero. The breakthrough comes from starting small and intentional. We started with deploying a sales-qualification agent, not attempting to reinvent the entire CRM system overnight. By aligning each agent to the right business task and security posture, momentum builds. Once one agent is working well, extending that lifecycle across the organization becomes far easier.

The Starting Point Problem

Customers tell us the future is AI and agents, so they start trying to identify which agents to build. That’s backwards.

Power Apps are already reshaping workflows, made accessible through low-code development. Apps remain the primary interface for getting work done, while agents are introduced within that structure, extending, not replacing it.

The companies that figure out how to integrate agents into existing structures will be ahead when agents increasingly shape the app experience.

Beyond Technology: The Culture of Adoption

There’s a tendency to assume that if you build an agent, adoption will automatically follow. That’s not what we’re seeing.

People need clarity on what agents are and what agents do. More importantly, they need to see how agents drive real outcomes in their day-to-day work.

That’s why Prompt Coach has emerged as one of the most widely used agents today. Better questions lead to better actions from your agents, just as thoughtful app design leads to better results.

When people go in cold, building an agent without understanding its constraints, it can lead to disappointment. When pre-built agents are difficult to find or poorly explained, they get applied in ways they weren’t designed for.

Take something like “Sales Lead” in Copilot. It’s easy to assume it will replace their CRM. Really, it’s a conversational interface over sales data. That gap in understanding is an opportunity to keep frustration from creeping in. It’s why guidance matters.

Using agents effectively is a new skill and one that will develop over time. It’s a muscle that needs to be built. Success requires practice, reinforcement, and support, not being treated like a one-time rollout.

We’ve seen this play out in large deployments. Organizations that reached broad adoption didn’t get there because people used Copilot for everything. They got there because enough people found clear value—using it for research or analysis—and realized they could turn work that once took hours into something they could do in minutes. Even something as high-stakes as a board presentation.

Security and Governance Can’t Wait

Years of successful low-code and Power Platform deployments have made strong security the norm, but not the exception. But there’s still inherent tension between empowering creators and IT people protecting the enterprise.

We have customers who have hardened every other part of their organization. But if an employee who has access to research agents or analysts gets phished, that single compromised identity can unintentionally backdoor access to company information that’s otherwise well protected.

This risk, managing access, intent, and oversight as agents become more capable, is the most significant challenge customers are working through today.

That’s why our deployments start with ideation of agents, clarifying where they can add the most value. Then we focus on building the infrastructure required to scale responsibly. We call this an Agentic Center of Excellence.

Once you have that foundation in place and the muscle of adoption is treated as a shared capability, not a one-time event, progress accelerates quickly. Teams can move faster because guardrails are already established.

Change Management That Actually Works

One of the strongest patterns we’re seeing is managers becoming power users themselves.

In sales or service scenarios, managers who use agents to prepare for meetings and understand how their sellers are performing can provide more precise coaching. They have a better window into what’s happening, and that context makes their guidance more effective.

That behavior drives positive changes by creating momentum. When your manager uses an agent and sees success, you’re more likely to use it too.

Another successful approach is making learning social. Getting assigned a new agent to use feels like work. But when teams introduce incentives, run prompt-a-thons or hackathons with leaderboards, or use light gamification with leaderboards, engagement changes dramatically. Adoption can take off.

Organizations that turn technical documentation into group activities see real results. It drives adoption. It provides managers with visibility into which agents people use regularly. And it helps the organization understand where to invest and double down.

What Frontier Firms Do Differently

They understand the time-to-value equation. Power Platform and Power Apps enable teams to build and deploy faster than custom development, with adoption curves that reflect that speed.

They put tools in individuals’ hands first, then scale deliberately to teams and departments. They get the fundamentals of Power Apps down and then layer in agents once their foundation is in place.

They are realistic about their journey.  They understand no one becomes a frontier firm overnight, and they’re not trying to boil the ocean. They’ve seen what happens with high activity and agent sprawl, but low productivity and poor governance.

They have a plan for the full governance and adoption curve. They know agents require the same level of rigor as people with clear permissions, oversight and accountability always top of mind. They’re not expecting to build everything themselves.

They understand this is about humans assisted by agents, not humans replaced by agents. It’s about people being empowered to do better work, with agents alongside them.

The Platform Foundation Question

We are often asked about how RPA, Power Platform, and agentic AI all fit together. Here’s the straightforward answer: the best way to sustain success with agents is to start with a solid app platform.

The investment you make in the application layer doesn’t get left behind; it compounds. That foundation gives agents context, structure, and guardrails, allowing them to deliver more value over time. That’s the strategic decision.

Apps aren’t dead. SaaS isn’t going away. They’re just changing. And the companies that understand how apps, automation, and agents work together are the ones that will lead what comes next.

Michelle Lancaster is General Manager and Partner, Global Leader for Business Strategy, AI Business Solutions at Microsoft, where she helps organizations navigate digital transformation and AI adoption across industries.

For more information please visit https://www.microsoft.com/en-us/power-platform