Blog

Insights & ideas

Stay ahead with expert articles, industry trends, and actionable insights to help you grow.

10 mins read

December 3, 2025

Is Agent 365 the moment enterprise AI becomes real?

Agent 365 is the moment AI enters the enterprise stack with real identities, permissions, and governance. Before this becomes your new operating model, you’ll want to understand what’s coming.

TL;DR

A365 is Microsoft’s new identity, governance and management layer for AI agents, giving each agent its own permissions, lifecycle, audit trail and operational controls. It's a signal that AI isn’t a side feature anymore; it’s becoming a governed, scalable digital workforce inside the enterprise stack. Instead of scattered pilots and experimental bots, enterprises get a unified way to build, manage and scale agents across CRM, ERP, HR, finance and data workflows. This is the shift from “AI as a helper” to “AI as part of the workforce,” and it raises a simple question: are you preparing your processes, data and governance for digital labour, or will you be catching up later?

How will Agent 365 reshape the way organisations work?

Most organisations spent the last year wrapping their heads around Copilot: what it can do, where it fits, and how to introduce it without overwhelming employees. But while everyone was busy figuring out prompts and pilots, Microsoft was preparing something far bigger.

Agent 365 is the moment enterprise AI stops being a clever assistant and becomes a managed digital workforce.

There’s an important detail that wasn’t obvious at first: the A365 icon sits inside Microsoft’s AI Business Applications stack, the same family as Dynamics 365 and the Power Platform. What looked at first like a Modern Work / Office feature is actually positioned alongside enterprise-grade business applications.

And they gave it the “365” name. When Microsoft attaches “365” to a product, it becomes part of the workplace operating system. SharePoint, Teams, Excel, Dynamics. These aren’t just tools, they’re the foundation of daily work. This isn’t accidental positioning; putting agents in the 365 family, Microsoft is sending a clear message:

AI agents are not experiments anymore. They are part of the enterprise stack.

And this has huge implications for IT Ops, Security, CoE teams, and business leaders.

From scattered bots to a unified agent ecosystem

If you’ve worked with Copilot Studio or any of the early Microsoft agents, you know the experience hasn’t been consistent. Agents lived in different places, were created in different ways, and had different capabilities. Some behaved like chatbots, others like automations. A few acted like full digital workers, if you were brave enough to give them permissions.

Agent 365 is the first attempt to bring order to this chaos. Instead of agents scattered across the Microsoft ecosystem, there will be one place to see them, manage them, and govern them. Microsoft calls it the Monitoring Admin Center, where agents are treated like real operational entities.

For the first time, IT teams can:

see all agents in one view

assign their own permissions

scale them independently

isolate them if needed

monitor activity

apply governance policies the same way they do for users

This is the shift organisations have been waiting for. AI is no longer a set of small tools you sprinkle across teams. It becomes a proper enterprise layer, where you can administer, secure, and scale agents.

Copilot vs Agent 365

What’s the difference? A useful way to think about it:

Copilot is the interface where people talk to AI.

Agents are the products the AI performs the work with.

Copilot will remain the interaction layer used across Microsoft products, but the deeper AI ecosystem (the one that will actually power work) is Agent 365.

This means that agents are moving into infrastructure territory.

A unique identity for every agent changes everything

The most important and least understood part of the announcement is Microsoft Entra Agent ID.

Until now, most AI agents have run under user identities, app registrations, or custom service accounts. Agent ID introduces a new, first-class identity type in Entra that is purpose-built for agents.

With Agent ID, an enterprise agent can finally have:

its own identity in Entra

its own assigned permissions instead of inheriting a user or app profile

its own access and governance policies, including Conditional Access

its own lifecycle management (creation, assignment, decommissioning)

its own auditability, with logs that show what the agent did and when

its own compliance surface, so organisations can apply the same Zero Trust, monitoring and oversight they use for other identities

In short: Agent ID gives agents a proper identity layer, separate from people and apps, and creates the foundation for secure, governed, enterprise-grade agentic automation.

You’re no longer tying a bot to a user’s permissions and hoping nothing goes wrong. You can now manage a digital worker with the same clarity as a human one, without the HR paperwork.

For IT Ops and Security teams, this is the part that makes scalable AI realistic. Without clear identity, real autonomy is impossible. Agentic ID is the foundation for everything Microsoft wants to build next.

Tools turn agents into real digital workers

Early AI agents were impressive but limited. They could answer questions or summarise documents, but they couldn’t do much.

Agent 365 changes that by introducing a real tool model: secure, isolated, pre-defined capabilities that agents can invoke to complete tasks on your behalf.

This brings a new class of role-specific agents. Some use cases we expect to see soon:

An agent with invoice-reading capabilities can take on routine finance tasks.

An agent that can post into your ERP can handle basic accounting work.

An agent that can update your CRM can manage SDR-level activities.

In other words: your business systems stay the same, but what your agents can do inside them expands dramatically.

The tools define the scope of work, and the governance layer defines the boundaries.
Once those two connect, something significant happens:

AI stops being a helper and becomes a decision-maker. That’s why companies need structure, identity, and controls before they deploy anything serious. And this is exactly what Agent 365 provides.

Microsoft will ship out-of-the-box agents

Microsoft doesn’t hide the direction anymore: they’re building their own out-of-the-box agents for major business functions.

Expect products like:

Sales Development Agent

HR Lifecycle Agent

Customer Service Agent

Finance/ERP Agent

Fabric Data Agent

Security and Compliance Agents

These will be real, supported Microsoft products. And they will almost certainly be licensed per agent, just like every other 365 workload.

This will raise important organisational questions:

"How many agents do we need?"

"Which roles replace manual steps with agents first?"

"Should we start with one per department or buy bundles?"

"What does ROI look like at the agent level?"

Licensing will likely become more complex; but the value will grow even faster for organisations that introduce agents deliberately, not reactively.

Where businesses will see early wins

In the next 12 months, the most realistic value will come from processes that already run inside Microsoft systems and already require repetitive, structured work:

Sales teams cleaning pipelines

Finance teams processing invoices

Customer service teams triaging cases

Data teams preparing datasets

HR teams onboarding people

Anywhere a human currently moves structured data between structured systems, an agent will do it faster, cleaner, and more consistently.

And the mistakes to avoid

Agent 365 brings enormous potential, but, like every major Microsoft release, it also comes with predictable, avoidable traps.

As with every AI initiative, readiness is key. Before you commit to licences, tools or departmental rollouts, make sure you’re not walking into the same issues that slow organisations down every time a new solution arrives.

Don’t skip process mapping.
Use frameworks like MEDDIC or Value Architecture Design to ensure you’re automating a clean, well-understood workflow instead of scaling a broken one.

Don’t buy more agents than your teams can adopt.
Start small. A controlled pilot with a handful of agents will always outperform a large purchase no one is ready for.

Don’t roll out everything at once.
Introduce agents gradually so users have the space to understand how each one fits into their workflow before the next arrives.

Don’t skip process mapping.
Automating a broken process only creates a faster, more expensive version of the same problem. Map the journey first, then bring in the agent.

Don’t underestimate data quality.
Agents make decisions based on the information you give them. If your CRM, ERP or SharePoint data is inconsistent, the agent’s actions will be too.

Don’t assume governance will “figure itself out.”
Without clear ownership, shadow agents, over-permissioned tools and ambiguous access boundaries will appear quickly.

When these pitfalls are ignored, the same uncomfortable questions always come back:

“Why isn’t anyone using what we bought?”

‍“Why isn’t this delivering the value we expected?” ‍

“How did this agent end up with access to everything?”

The organisations that succeed aren’t the ones who rush. They’re the ones who pause long enough to define clean data, clear ownership, intentional design and a rollout plan that respects how humans, not machines, adapt to new ways of working.

The future of work will be humans + agents

Agent 365 is the moment Microsoft finally aligns its tools, its platform, and its vision:
every person will work through Copilot, and every system will be executed by agents.

The question for organisations now is simple:

Are you preparing for a future where digital labour is part of your workforce, or will you be retrofitting governance after the agents have already arrived?

We can help with the clarity, structure, and safe adoption you’ll need. Join our free webinar where we'll walk you through how to get AI-ready in 90 days.

‍

Soft blue and white gradient background with blurred smooth texture

Filter

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part 2

Industries

Technologies

Solutions

June 3, 2024

2 min read

Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part 2

‍

In this hierarchical model, we start with epics, representing the overall project, and break down the development cycle through various levels right up to testing. This approach helps organize work, set priorities, and track progress. Let’s dive into each element and its significance:

Epic: At the highest level, the epic represents the project itself. This category encompasses the overarching goals and the project framework. Features and user stories under the epic serve to achieve specific objectives.

Feature: Within an epic, features reflect customer needs. These are concrete requirements and expectations expressed by customers that we aim to meet throughout the project.

PBI (Product Backlog Item): These are elements of the product backlog, which can be issues or user stories.

Action: Specific activities that need to be completed to achieve the project’s goals.

Issue: Problems or bugs identified during the project, as noticed by customers.

User Story (US): Detailed breakdowns of customer requirements. These are short, simple descriptions that outline the functionalities and benefits customers expect from the product. User stories help developers understand and accurately fulfill customer needs.

Task: Specific tasks derived from user stories and features that the project team must complete.Bug: Software defects identified during development. These can be issues found by either customers or developers.
Build: Development tasks aimed at creating a new version of the software.
Test Case: Test scenarios that specify what tests need to be executed to verify different aspects of the software.
Test Plan: A comprehensive plan that includes all available test cases and their results.
The process model illustrated here covers every step of the software development cycle, from requirement gathering to testing. This aids project teams in effectively managing development activities, improving software quality, and ensuring project success. This model not only organizes needs and work but also facilitates communication with customers.
By following this structure, Visual Labs ensures that all aspects of the project are covered comprehensively, promoting efficiency and clarity throughout the development process.

Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part 1

Industries

Technologies

Solutions

June 3, 2024

2 min read

Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part 1

At Visual Labs, we leverage Azure DevOps, a powerhouse tool in software development and project management. It empowers teams to efficiently manage their work from development to release. One critical element that can significantly influence project success is the structure of the backlog. A well-organized backlog not only ensures task and requirement transparency but also helps teams set priorities, respond effectively to changes, and closely monitor project progress.

In our upcoming blog series, we will showcase the backlog structures we use, highlight the differences between them, and discuss the typical work items in these backlogs.

In this blog post, we’ll delve into how we create and efficiently manage an Azure DevOps backlog. We’ll explore practices and tips to help teams maximize their backlog’s potential. We’ll cover how to handle requirements and tasks, and how to ensure the backlog reflects the current goals and challenges of the project. Join us to discover how we enhance project management efficiency with Azure DevOps!

At Visual Labs, we distinguish between two different backlog levels:

1. Delivery Backlog:

This type of backlog contains tasks related to project or product development. It includes software development, implementing new features, bug fixes, user stories related to customer needs, and any other activities that directly impact the product or project’s output. The goal is to support the continuous development and delivery of the product or service. Tasks in this backlog typically have higher priority as they directly contribute to customer value.

2. Admin Backlog:

The admin backlog encompasses tasks related to the project’s or team’s administrative, organizational, or internal operations. This includes updating internal documentation, ensuring regulatory compliance, training team members, or any activities that are not directly linked to product or service delivery but are essential for smooth operation. While these tasks might be less urgent or critical from a product perspective, they are vital for maintaining team efficiency and seamless project execution.

Summary:

The delivery backlog focuses on product development and delivery, while the admin backlog handles internal operations and administrative activities. Both are crucial for successful project management and teamwork but concentrate on different aspects.

Industries

Technologies

Solutions

June 2, 2024

3 min read

DOJO

Background

The forum for operational development at Visual Labs is called Dojo. These sessions are held every Tuesday morning for one and a half hours. Originally launched in January 2023, Dojo was a half-day event held every two weeks and could only be attended in person.Participation in Dojo sessions is optional, although certain colleagues are often expected to attend due to their expertise. The topics of Dojo are determined by quarterly goals, and the exact agenda is set by management. These topics cover a wide range of areas, aiming to contribute to organizational development.

Why is it necessary?

Despite being a small organization, we work in distinctly defined technological areas (ERP, CRM, BI) that often serve the same clients. It is essential to have a common forum where we can gain insights into each other’s work or the company’s operations (e.g., ERP or BI colleagues are welcome at a CRM hackathon). This allows for collective thinking about how Visual Labs can develop as an organization.

Topics Covered So Far

Among others, the following topics have been discussed in Dojo sessions:

Using Azure DevOps (see: Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part One)
- Feature delivery process
- Managing work items
- Defining work items
- Handling statuses
Refining financial planning and billing processes
Hackathon for improving our internal CRM
Brainstorming on the development of our increasingly cramped office
Solution-seeking:
- For technical challenges
- For broadly defined, exciting client needs
Power Platform Starter Kit
- Launching the PowerPlatform Center of Excellence (CoE) Starter Kit
- Agile sprint review and planning ceremonies for the Visual Labs enhancement of the PowerPlatform Starter Kit ([[CoE+]]), where we also tested our agile wings 🙂
Customer relationship and competency training for POs, conducted by our regular coach, Klára Sugta
Knowledge sharing on the following topics:
- Customer visits
- Project launches

What Has Dojo Given Us?

Key achievements:

Creation of a Dojo Handbook that defines our main processes and operations, which we keep up to date.
Further development of the Microsoft-developed Power Platform Center of Excellence Toolkit [[CoE+]], cataloging Azure resources, DevOps projects, and areas that can be assigned to teams and users.
Strengthening and integrating the competencies outlined in the Competency Matrix into our daily routines.
Introduction of monthly financial planning with weekly tracking.
Development of our own CRM system (based on Dynamics 365 Sales and Project Operations, in collaboration with CoE+).

Other Forums for Organizational Development

Team Retrospectives
- While Dojo generally deals with cross-departmental or company-wide topics identified by management, retrospectives at the team level focus on system-level improvements in individual collaboration.
- In the future, it will be important to create a development backlog (Kaizen) to provide an overarching view of areas needing improvement, to which initiatives can be assigned and receive the necessary buy-in and resources.
1on1s
- The most straightforward opportunity to foster bottom-up ideas is through 1on1s, where topics can ripple up to management level.

The Future of Dojo

Dojo will continue to be the main forum for organizational development and will remain optional. We aim to facilitate participation in organizational development asynchronously and ensure that more written records of Dojo discussions are kept. Often, colleagues cannot attend sessions on topics they find interesting or do not find it worthwhile to sit through the entire discussion but would find a "tl;dr" version useful.If you want more information about our Dojo sessions or the topics discussed, follow our blog, as new articles on these topics are continuously published. If you have any questions or comments, don't hesitate to contact us!

Industries

Technologies

Solutions

May 22, 2024

3 min read

Dojo

Background

Participation in Dojo sessions is optional, although certain colleagues are often expected to attend due to their expertise. The topics of Dojo are determined by quarterly goals, and the exact agenda is set by management. These topics cover a wide range of areas, aiming to contribute to organizational development.

Why is it necessary?

Topics Covered So Far

Among others, the following topics have been discussed in Dojo sessions:

Using Azure DevOps (see: Optimization from Top to Bottom: How We Structure Backlogs at Visual Labs, Part One)
- Feature delivery process
- Managing work items
- Defining work items
- Handling statuses
Refining financial planning and billing processes
Hackathon for improving our internal CRM
Brainstorming on the development of our increasingly cramped office
Solution-seeking:
- For technical challenges
- For broadly defined, exciting client needs
Power Platform Starter Kit
- Launching the PowerPlatform Center of Excellence (CoE) Starter Kit
- Agile sprint review and planning ceremonies for the Visual Labs enhancement of the PowerPlatform Starter Kit ([[CoE+]]), where we also tested our agile wings 🙂
Customer relationship and competency training for POs, conducted by our regular coach, Klára Sugta
Knowledge sharing on the following topics:
- Customer visits
- Project launches

What Has Dojo Given Us?

Key achievements:

Creation of a Dojo Handbook that defines our main processes and operations, which we keep up to date.
Further development of the Microsoft-developed Power Platform Center of Excellence Toolkit [[CoE+]], cataloging Azure resources, DevOps projects, and areas that can be assigned to teams and users.
Strengthening and integrating the competencies outlined in the Competency Matrix into our daily routines.
Introduction of monthly financial planning with weekly tracking.
Development of our own CRM system (based on Dynamics 365 Sales and Project Operations, in collaboration with CoE+).

Other Forums for Organizational Development

Team Retrospectives
- While Dojo generally deals with cross-departmental or company-wide topics identified by management, retrospectives at the team level focus on system-level improvements in individual collaboration.
- In the future, it will be important to create a development backlog (Kaizen) to provide an overarching view of areas needing improvement, to which initiatives can be assigned and receive the necessary buy-in and resources.
1on1s
- The most straightforward opportunity to foster bottom-up ideas is through 1on1s, where topics can ripple up to management level.

The Future of Dojo

If you want more information about our Dojo sessions or the topics discussed, follow our blog, as new articles on these topics are continuously published. If you have any questions or comments, don’t hesitate to contact us!

My Journey with CI/CD in Power BI: A Personal Tale of Transformation Part 2

Industries

Technologies

Solutions

May 6, 2024

3 min read

My Journey with CI/CD in Power BI: A Personal Tale of Transformation Part 2

Implementing This Miracle: If I were to start all over again, I'd tell you this:

A) Get cozy with Git, simulate real-world scenarios as a team during a demo session, and document/record everything.

To better prepare for collaborative development, we crafted a scenario in which multiple developers would work simultaneously on the same file, aiming to rehearse a range of actions we had previously encountered. This included creating different measures at the same time and then merging these into a single file, adding, removing, or modifying visuals and integrating those changes, reverting to a previous version to serve as the basis for new features, and managing version transitions between DEV, TEST, and PROD branches. Through this exercise, we sought to simulate and navigate the complexities of real-world collaboration, enhancing our team's ability to handle various project management tasks efficiently.

Commit early and often, and
Treat your commit notes like a diary of your project's life. Writing comprehensive commit descriptions will enhance clarity and facilitates easier navigation through the project's history for both you and your teammates.
Branch Carefully! Always verify you're working on the correct branch to avoid unintended changes in the wrong areas of your project.
And never forget the power of communication. Despite the technical tools at your disposal, effective teamwork hinges on constant communication to ensure alignment and collaboration.

These practices didn't just make our projects better; they made us better developers and teammates.

B) Bumps Along the Way: Sure, we faced our fair share of surprises.

Challenges Encountered Since Adoption:

Despite the myriad benefits, several challenges have emerged since adopting source control in Power BI, particularly given the preview status of this feature. I'm going to highlight the few most common ones.

Custom Visuals Compatibility: Custom visuals used in reports need to be installed separately by each collaborator. Don't forget to let your colleagues know which custom visual you added to your dashboard as its name won't show up only the following error message which is a bit hard to dechypre.

‍

2. Merge Conflicts and Code Loss: Situations have arisen where accepting both changes during a merge resulted in lost code, highlighting the need for careful conflict resolution (screenshot from another blogpost describing the issue).

‍

3. File Opening Issues After Merge: Conflicts within the data model, such as incorrect relationship settings, can prevent files from opening, necessitating reversion to previous versions.

‍

4. Infinite Semantic Model Refresh Loops in PBI Desktop: Unexplained delays in model refreshes post-merge, extending for hours, indicate potential issues with large semantic models.

+1) Data is not stored in repos: It’s good to keep in mind that data is not stored in the Repo, only semantic model and report related code is stored, you still need to refresh your data on Power BI Service.

Conclusion: This journey of integrating CI/CD and source control into our Power BI workflows has been one of the most rewarding experiences of my professional life. It wasn’t just about making our projects more efficient; it was about transforming our team into a more cohesive, capable, and confident unit. As we look to the future, I’m excited about the possibilities that lie ahead, ready to tackle whatever comes our way with a smile and a git commit.

My Journey with CI/CD in Power BI: A Personal Tale of Transformation Part 1

Industries

Technologies

Solutions

April 23, 2024

3 min read

My Journey with CI/CD in Power BI: A Personal Tale of Transformation Part 1

Hello, friends! Today, I'm diving deep into my own adventure with Continuous Integration (CI) and Continuous Deployment (CD) in the realm of Power BI—a journey marked by trials, triumphs, and a lot of learning along the way. For so long, CI/CD in Power BI felt like trying to fit a square peg in a round hole. With a mix of third-party tools and makeshift solutions, my team and I navigated through a maze of compliance and administrative hurdles, often feeling lost in a sea of approvals and support tickets to try introduce a potential CI/CD solution or workaround. Then came May 2023, and with it, a beacon of hope: Power BI project files. This was more than just a feature; it was a revolution that promised to redefine how we approached our projects.

The Before Times: Rewind to before this pivotal change. My team, like many others working in data analytics within complex, multinational landscapes, struggled immensely with the lack of source control in Power BI. Our attempts at version control felt archaic—think "save as" with date and version stamps—and far from agile (shared folders on SharePoints or OneDrive). We were a team longing for simplicity and efficiency but found ourselves bogged down by the limitations of our tools.

A Game Changer: Power BI Project Files: Then, the game changed. The introduction of project files (.PBIP) wasn't just an update; it was a lifeline. This wasn't about saving our projects as mere files; it was about evolving them into living, breathing entities, organized within folders that spoke the language of our data stories through JSON (not ideal, but still an enormous improvement) .

Getting Our Hands Dirty: Embracing this new world, we ventured into the depths of Git with VS Code and Azure DevOps.

Each Power BI file became its own project, its own repository. This structure, while logical, required us to rewire our brains, to rethink our approach to collaboration and version control. Our workflow transformed, becoming more streamlined yet demanding a new level of diligence and precision.

The Rocky Road: It wasn't all smooth sailing. Adopting Power BI project files and integrating Git introduced us to a host of challenges. Merge conflicts became our nemesis, and the occasional quirks of a preview feature tested our patience. But with every stumble, our resolve grew stronger. The benefits—oh, the benefits!—far outweighed the occasional headaches. We were building something resilient, transparent, and infinitely more manageable.

Looking Ahead with TMDL: The next big step ahead is the introduction of TMDL. Storing your semantic model in TMDL (preview feature since the March 2024 update) transforms the code into a format that's not only human-readable but also intuitively organized, making every element—be it measures, tables, or columns—distinctly easy to identify in separate blocks. This clarity significantly simplifies the process of resolving merge conflicts and tracking changes, making the whole experience more straightforward.

‍

Why This Matters: The move to source control in Power BI has been nothing short of transformative. It brought clarity to our development process, made our end products more reliable, and, most importantly, it strengthened our team's bond. We learned to communicate more effectively, to trust in our collective skills, and to embrace the inevitable learning curve together.

To be continued...