Rex vs. Microsoft Copilot: A guide to choosing your AI stack

By Dan Duke—If you’re a business leader evaluating AI platforms, you might be thinking:

“We already pay for Microsoft 365—shouldn’t we just use Copilot?” It offers a powerful, multi-layer platform with huge distribution and lots of prebuilt integrations.

But businesses can get more from AI than a long list of features and applications to speed up workflow. AI can help you develop a new operating model for how work gets done better and more efficiently.

That’s what Rex, Rellify's primary AI interface, is built for. Rex encourages and rewards an “AI state of mind.” It prioritizes autonomy, non-restriction, and company-controlled deployment—especially where sovereignty and data protection matter. 

Many organizations work with Copilot. However, a growing number of businesses are finding value in the way Rex optimizes for different outcomes in creativity, productivity, and security. 

Copilot is a system of controls; Rex is a system of consequence

A number of enterprise AI products, including Copilot, still reflect “old world” software assumptions, like: 

• Overdevelop the AI with fine-grained configurations that … 

• show users lots of screens where they …

• take lots of manual steps filling in various fields to … 

• supply information that AI can work with.

Rex is built around a different assumption: If you tell an AI agent what outcome you want, it can infer many of the steps. And you can do this with confidence, because data sovereignty is baked into Rex and you can explicitly describe the guardrails you need.

Rex is designed to help teams move from “chatting” to agent-driven execution—without requiring your team to become prompt engineers or workflow designers.

Brief comparison of Rex and Copilot

Rex is a first-class AI stack—not a chat-first tool. It’s built to support real work in a secure runtime environment you own and control. Jürgen Kranz, Rellify’s Head of Technology, EMEA, describes the stack in this way:

• State of mind

• User interface

• API

• Agent orchestration

• Network

• Agent harness

• Large Language Models

• Adapters and integrations

He explains that although “State of mind” sounds a little abstract, he puts it at the top of the stack because it has a very concrete and profound effect. 

Copilot is a brand that spans multiple layers:

• Copilot in Microsoft 365 apps (Outlook, Word, Excel, PowerPoint): Chat interfaces embedded where work already happens.

• Copilot Studio: Where you can create agents/workflows, as you can do in Rex, with a powerful UI and extensive prebuilt adapters, connectors, and integrations.

• Enterprise-grade services (Azure layer): A heavy-duty enterprise infrastructure side—databases, large systems, and enterprise operations under the Copilot umbrella.

For Microsoft-standardized organizations, this layered distribution is a serious advantage: Copilot is designed to meet users where they already work.

Kranz says that one way to sum up the differences between Rex and Copilot is that:

• Microsoft is a huge tanker. It has massive resources, broad coverage, polished depth everywhere. In one-to-one comparisons, Copilot may look “better” because of the billions of dollars invested into it.

• Rex is the race boat. It's faster, more agile, more willing to design using AI-native assumptions rather than legacy UI and legacy workflows.

Why “State of mind” sits at the top of our stack

Kranz puts “State of mind” at the top of our stack because of its cultural and architectural meaning. To help our clients make the most of AI, we work in an “AI state of mind” and help them approach Rex in that manner.

It reflects the willingness to forget what we think we know about software and rebuild around what AI can do now.

Kranz experienced this new mindset when he first started playing with Rex to learn about its possibilities. He and a colleague decided to create a team of AI agents to create blog posts. 

Kranz spent at least an hour meticulously handcrafted a set of agents:

• An English-writing agent

• A German-writing agent

• A brainstorming agent

• A Q&A agent

• A “master” agent to coordinate them

His colleague tested Rex in a different way, telling it: “I want a team of agents to create blog posts. Think about what roles you need and then create them.”

Kranz says Rex quickly created almost the same agents that he had, except that Rex added one more agent to cover a role they hadn’t thought of.

This illustrates the AI state of mind:

• Stop trying to pre-design the perfect agent organization chart.

• Tell the system what you want.

• Let it propose the structure.

• Judge the output.

Don’t restrict your agents

Rex can surprise users by seeing beyond their instructions to suggest things they hadn’t thought of or accounted for. It’s an offshoot of Rex’s ability to improve itself. The idea is simple:

• Rex reads feedback

• Rex identifies patterns

• Rex proposes fixes

• Rex implements them

“With an AI state of mind,” Kranz says, “we’ll let the system learn and evolve faster than a human process can.”

If you over-restrict agents, you prevent them from being smarter than you.

In a lot of enterprise software design, the kind we see with Copilot, for example, we think “control = quality.” So developers build more fields, more constraints, more screens, more forms.

But with AI agents, restrictions often remove the exact behavior that creates leverage: the agent seeing the whole and improving it, not just executing one micro-task.

Don’t treat agents like human workers

If you build “teams of agents” the same way you build teams of humans, you carry over human assumptions—handoffs, rigid roles, communication patterns.

“I don’t think that’s the future,” Kranz says.

Allowing agents to communicate in their own manner can yield better results than trying to force a human team structure. The goal is not to make agents imitate people. The goal is to get the best outcomes with the least friction.

An approach to AI that is too fine grained

Kranz says Copilot workflows seemed “a little bit stuck in the past.”

For example, look at using AI to send email. In an old-school enterprise UI, the user must fill in field after field—recipient, subject line, body text, and on and on.

“If you tell an AI to send an email, it already knows it needs a recipient, a subject, and the message,” Kranz says. “In the AI state of mind, we don’t create a world where humans keep filling out forms and AI is sprinkled in to process the forms. Instead, we build a world where you express intent, and agents infer the necessary steps.

“That’s where I see a real advantage for Rex.”

Productivity gains with Rex

Here’s another example of the power of Rex and Rellify’s approach to agentic AI.

Kranz and his colleague had been working for about a year on a side project in the low-code environment Bubble. They spent endless hours:

• Clicking

• Dragging

• Defining rules

• Clicking some more

• Adjusting edge cases

• Repeating

They grew tired of that and turned to Rex, using a different approach, Kranz says. They were afraid that it might take a great deal of time and effort to retrace their steps, but it didn’t turn out that way.

“We told Rex: Go look at our solution, understand every detail, and write a specification. We took that specification and had AI implement the whole application from scratch.

"We rebuilt the application in about two weeks—after spending close to 12 months building it the first time.”

Rex wasn’t just faster and more efficient. It also noticed that a key workflow was missing and wrote the solution.

“We knew the hole was there. But we had postponed working on it because we thought it was too complicated,” Kranz says. "Rex just did it."

"This is why I keep coming back to this point: Don’t put agents in a box. If you do, you’ll just get a faster version of your old workflow. If you don’t restrict your agents, you get a system that can improve the workflow—and sometimes your business logic itself."

Sovereignty and security with AI

Another major difference between Rex and Copilot—especially in Germany and the rest of the EU—is data protection and sovereignty.

We can provide Rex through Stackit, hosted in Germany, or on the premises of a European client using open-source models. That means full data protection for your company and your customer data. 

The importance of this can’t be overstated. And the big tech players in the U.S. can’t match it.

If you use frontier models in the U.S. cloud, such as Gemini, Claude, and Copilot, you take on vendor lock-in and you’re sending your data into ecosystems you don’t fully control. Even if you check a box that says “don’t use my data,” you may still wonder where the data really goes, who can access it, and under what legal regime.

In the EU, there’s an additional concern around the U.S. Cloud Act, which allows American law enforcement to force U.S.-based technology companies to produce electronic data and communications, regardless of whether that data is physically stored inside or outside the United States.

That’s why Rex is built differently. Copilot and other chat-first AI tools provide access, which means you rent capabilities from providers who control the infrastructure, the data, and ultimately the relationship. 

Rex and Rellify provide ownership, which means running AI systems that operate within your infrastructure, under your control, with your data staying your data.

You can have a strong tech stack that’s independent of the big companies and the security concerns that come with them. And that combination is a rare offering.

Where to go for an AI tech stack?

If your organization is all-in on Microsoft 365, Copilot may be the fastest way to get broad, baseline AI adoption.

But if your goal is to move beyond “assistive AI” into agent-driven execution, and if you care about autonomy, speed, and sovereign options, then you should look at Rex.

You can start a free trial today or schedule a demo. We’ll show you exactly what we mean by putting “state of mind” at the top of the stack. You can quickly see how to get dramatically better outcomes when you stop over-specifying, stop restricting, and start letting agents do what they’re good at.