Gartner Digital Workplace Summit 2026 made one thing very clear:
The AI-native workplace isn’t a technology race. It’s a readiness race.
In Part 1 of our Gartner Summit recap, we explored the surprisingly human side of AI adoption: trust, governance and the growing problem of AI slop.
But in London, another conversation kept surfacing.
Organizations want to scale Copilot. They want AI agents. They want the productivity gains they’ve been promised.
There’s just one problem: their digital workplace often isn’t ready for them.
01
The Copilot problem nobody wants to talk about
Microsoft 365 Copilot promises to transform how employees find, create and consume information.
The ambition is clear:
A workplace where employees can ask questions, generate content, discover knowledge and automate tasks instantly.
But the reality?
Many organizations are discovering that Copilot is only as powerful as the environment underneath it.
And for many Microsoft 365 tenants, that environment is not ready yet.
Less than 25% of knowledge workers currently have access to Copilot. Not necessarily because organizations lack interest or budget — but because the foundations Copilot depends on are often not mature enough.
The challenge isn’t the AI.
The challenge is everything AI needs to work with:
- outdated content
- unclear ownership
- duplicated information
- inconsistent permissions
- overshared SharePoint environments
Gartner highlighted just how significant this challenge has become:
say they need additional technical controls to properly manage and secure Copilot agents.
believe they currently have the right governance structures in place.
are primarily reducing risk by restricting rollout to trusted, low-risk users.
Restricting access might reduce exposure, But it doesn’t solve the underlying problem.
Because Copilot’s value depends entirely on the quality of the information it can access and the controls protecting that information.
Deploy it on top of a clean, governed digital workplace and it can be transformative.
Deploy it on top of a typical Microsoft 365 environment filled with outdated, duplicated and overshared content?
You don’t get a smarter workplace, You get an AI engine that can surface the wrong information faster — and with confidence.
That’s not an AI failure.
That’s a governance failure that AI made visible.
The next digital workplace problem
Welcome to Agent Sprawl.
First came site sprawl. Then Teams sprawl. Then app sprawl.
Now AI agents are creating the next governance challenge.
02
More agents, more risks
As creating agents becomes easier, organizations risk accumulating hundreds of them with overlapping purposes, inconsistent permissions and unclear ownership.
And agents introduce another dimension: they don’t just access information.
They can act on it.
Oversharing 2.0
Information can move through Human-to-Agent and Agent-to-Agent interactions in ways traditional permissions weren’t designed around.
Data loss & Content sprawl
Copilot is another exit point for organizational data — internally to people who shouldn’t see it, and potentially externally to bad actors. Every ungoverned SharePoint site is a potential leak.
New attack vectors
Copilot agents don’t just answer questions. They act. That action capability makes them a new entry point for attackers who know how to manipulate agent behavior or exploit weak access controls.
Response accuracy
Access to more information doesn’t automatically make an agent smarter if the underlying information can’t be trusted.
Garbage in. Garbage out.
Now at conversational speed.
03
AI readiness: the foundation behind the intelligence
Organizations often approach AI readiness as a technology question:
Organizations often approach AI readiness as a technology question:
“Which model should we use?”
“Which Copilot licenses do we need?”
“Which agents should we build?”
But Gartner’s message was much simpler:
“AI is only as good as the environment it runs in.”
Before asking what AI can do, organizations need to answer:
1️⃣”Is the information current?”
2️⃣”Who owns it?”
3️⃣”Who should access it?”
4️⃣”Can employees find it?”
5️⃣”Can AI trust it?”
AI readiness is ultimately about ensuring information is:
✓ trusted
✓ findable
✓ up to date
✓ properly governed
All before AI consumes it, and this is where the intranet takes on a new role.
NOT just as a communication channel.
NOT just as a homepage.
But as the trusted knowledge layer connecting people, content and expertise.
The AI readiness path
The AI-native workplace isn’t a technology race. It’s a readiness race.
And AI readiness starts long before the first agent goes live.
04
Agent 365: right problem, unfinished answer?
Microsoft’s response to the agent governance problem is Agent 365. Conceptually, it addresses a very real need.
On paper, Agent 365 offers:
- A unified agent registry
- First and third-party agent visibility
- Agent-level identity and access management
- Centralized visibility into agent activity
Challenge 1
The cost question
Agent 365 sits within Microsoft’s M365 E7 bundle.
For large organizations, moving from E5 represents a significant increase in investment.
The question becomes:
Does every employee actually need this level of capability?
Challenge 2
Governance still needs work
Agent 365 is a work in progress.
The current version lacks the automation that would make it genuinely useful for governance at scale.
Today, admins have to manually refresh pages to detect new agents and manually block the ones that shouldn’t be running.
That’s not governance yet.
That’s reactive monitoring dressed up as control.
Challenge 3
It can’t fix bad foundations
The root cause of most Copilot governance failures isn’t agent visibility: it’s poor information governance in M365 itself. Oversharing. Broken permissions. Sprawling, unstructured content lakes. A control plane for agents doesn’t fix a permissions model that was never right to begin with.
“Agent 365 is a promising work-in-progress with a premium price tag. Organizations need to think carefully before jumping in.”
05
What should organizations actually do?
Fix information governance first
Get your information governance in order before anything else. Classify your content. Fix your permissions model. Implement DLP policies and sensitivity labels. Apply retention policies to the content that needs them. This isn’t glamorous work. But it’s the work that determines for your Copilot deployment to be transformative.
Give every agent a lifecycle
Every agent needs:
- an owner
- a purpose
- defined access
- monitoring
- retirement criteria
If nobody can answer:
“Who owns this agent and what does it have access to?”
then you already have an agent sprawl problem.
Make readiness continuous
AI readiness cannot be achieved through one-off audits and manual clean-up efforts alone. Organizations should look for Digital Workplace solutions that continuously improve content quality, enrich metadata, identify outdated or duplicate information, monitor permissions, and enforce governance policies at scale. The more these processes are automated, the easier it becomes to maintain a trusted knowledge foundation and ensure AI always relies on relevant, secure, and up-to-date information.
Don’t rush into E7.
Assess what your organization actually needs, be honest about what your workforce will genuinely use, and explore whether third-party governance and intranet platforms can close the gap at a fraction of the cost.
06
More AI doesn’t automatically mean a better workplace
One theme came through louder than almost anything else in conversations on the floor: the proliferation of AI tools is actively hurting the digital employee experience.
It was AI fatigue.
More agents
More fragmentation
More switching
The organizations winning aren’t the ones with the most tools. They’re the ones that have unified their AI experience around how employees actually work, with the intranet as the anchor, not an afterthought.
And looking ahead, the shift is only accelerating. By 2030, 40% of enterprise apps will be custom-built up from just 2% today. That same logic is reaching intranets: personalized experiences that reflect role, location, and context; agents built for specific workflows rather than generic assistants. Intranet environments that can be built in hours, not months, because the right templates, webparts, and proven patterns are already there.
This is where being a vendor with years of intranet deployments behind you and a deep understanding of what works and what doesn’t, becomes a genuine advantage.
The takeaway from London
AI has made digital workplace foundations matter more than ever.
For years, digital workplace teams have been improving search, cleaning up SharePoint environments, reducing content sprawl and trying to establish better governance.
AI has suddenly made all of that work business-critical.
So before asking how many agents your organization can deploy, ask a simpler question:
Is your workplace ready for them?
Because AI readiness doesn’t start with the agent.
It starts with everything the agent needs to trust.
Trinity Resting
Product Marketing Manager
Product marketing doesn’t have to be all jargon and slides—Trinity’s here to prove it’s about creating human connections. With 5 years in digital marketing, she’s been scaling, building, and transforming product strategies that resonate with real people. At Powell, Trinity ensures the message isn’t just heard, but felt. She’s passionate about helping teams go beyond the office and build lasting relationships. When she’s not crafting strategies, you’ll find her sipping tea (she’s a self-proclaimed tea-aholic) or working as a professional dog sitter—because, let’s be honest, dogs are the ultimate team players.
