Developers!👏 Developers!👏 Developers!👏 Has there ever been a more iconic moment when a software company CEO confesses their love for the app builders than the chant Steve Ballmer did live on stage back in 2000?

No matter if today’s software developers usually prefer a non-Windows device to write code on, Microsoft still seems to remember the importance of the audience that builds new things. Being a huge corporation, it also keeps forgetting this every now and then, alienating these builders via shortsighted commercial decisions. Leading to the continuous need for winback campaigns, such as — quite literally, trying to get developers to come back to Windows.

Microsoft at its core remains a platform company, no matter if it gets a huge chunk of its revenue from selling per-seat M365 licenses to users who just… use the apps. The reason they have hundreds of millions of paid M365 seats is not the superior features of Office apps. It’s how most things, especially in the world of business technology, connect with Microsoft tech in one way or another. It didn’t all just happen because of Windows becoming the default OS back in the 90s. A big reason is in how well Microsoft has played the game of inviting others to build on top of their software.

Power Apps was and still is a success because of this very game. The massive distribution channel in M365 helps, sure, but it’s hardly the only reason. Driving SharePoint power users into becoming Power Apps makers that build basic apps on top of list data has been an achievement others haven’t pulled off at that scale. Microsoft didn’t need to have the most feature-rich low-code toolkit to win this market. What they had to do, though, was deliver an experience that motivated citizen developers to keep building.

Today, I’m going to look at the state of the app building market from a 2026 perspective. A decade after the core canvas app experience was launched, how well do Microsoft’s tools stack up against what the citizen developers can discover outside the ecosystem? Specifically, I will be looking at the biggest player in the prompt-to-app market today: Lovable.

I don’t know about you but I’m addicted to hoarding slide decks. Whenever there’s an event, be it by Microsoft or community organizers, the first question on my mind is “where can I download the presentations?” This has resulted in my private M365 tenant having a lot of .pptx files. Well over 2K, if these Copilot Search results are accurate:

Throwing decks into folders and hoping I’d one day come across the exact info I wanted when the need arises is a naive strategy — yet it’s pretty much been my modus operandi for years. Paying for M365 Copilot and having Work IQ as the semantic graph around the files doesn’t magically solve the problem. If I haven’t properly processed the information upon the initial intake phase, the chances of my brain knowing what to look for and where when facing a new challenge are slim to none.

Maybe I could do something more with AI to address this challenge? After all, it’s a scenario that users in bigger M365 tenants will surely have encountered. If I can experiment with my private SharePoint archives and come up with something smart, I might be able to help my customers make better use of this technology.

Instead of trying to organize thousands of files all at once, I decided to start from recent events that were still pending for decks to be downloaded. I established a fresh new EventDecks document library in SharePoint and began envisioning a better process than what I had before. I wanted both a better data structure and less clicks to get things in their right place.

Getting data into SharePoint can luckily be handled via the Graph API these days, without you needing to get your hands dirty with the GUI. AI coding agents will be fluent enough with a well-known target like SharePoint document libraries. Show them your source data, ask them to plan the useful metadata columns, then give the green light for modifying the library schema. In the end, you can have a reusable PowerShell script that will handle uploading new sets of data to that customized library. Like here with my EventDecks library:

This of course requires the kind of rights and skills that aren’t likely available to most information workers who interact with SharePoint. Getting to this point is therefore already a bridge too far for the typical M365 user. Now, there may well be first-party M365 Copilot agents that make this process easier. As an example, the SharePoint list agent does something similar with non-document scenarios, but I haven’t personally tested it. Depending on which Microsoft MVP you follow, the list agent either changes everything or is pretty “meh”:

For the SharePoint site-level admin, there are a lot of options on how to tweak the content and UI without touching a CLI, though. At worst, it involves exploring the many layers of legacy menus in SharePoint. At best, it’s a matter of just pasting in a bit of text from your AI chatbot. Let’s focus on the latter example.

You can do a lot of neat things with the list formatting these days. By providing the syntax reference link to your favorite AI coding agent, it can generate the JSON output that will transform your boring SP list into something much more creative. Below is an example of the Session Cards view that I was able to build with a few prompts back & forth:

In the end, the actual website experience is still controlled by the underlying CMS engine of SharePoint. Your list may have fancy cards in it but the way you search and filter for the content, as an example, is not that easy to change on an ad-hoc basis. That’s where I personally was most disappointed with what is doable via just chatting with Claude. Some of the guidance it gave on configuration options was not correct, demonstrating how the deeper you go into any Microsoft enterprise software, the more it takes tribal knowledge from human professionals to get the architecture right.

Why don’t we just turn it into a proper app then? That’s a great question, and the very reason I started writing this newsletter issue. Because when working with tools from the Microsoft cloud, it’s not that straightforward. My idea of the feature set that an EventDecks app should deliver to be better than raw SharePoint was:

Bouncing around these ideas with Claude resulted in the following plan of what would need to be built:

Turning a SharePoint document library into a canvas app should really be a simple wizard you click through. In practice, achieving a complete UI instead of a small companion app would be a lot of work. The worst part is that it would be mostly manual work for me to click around the canvas app studio GUI. Yes, I’ve tried letting GitHub Copilot build a canvas app for me via the canvas MCP server bridge. It’s been a huge waste of time and tokens, so I’m not going to use that path until the tech matures considerably.

Don’t get me wrong: there’s a lot that can be built on the Microsoft stack. In a high-value, business critical app, it can make perfect sense to spend some time on crafting things via the low-code tooling. It’s a deterministic, proven way to build apps, and it offers the kind of security and governance capabilities that few (if any) vibe coding products could match. The same Power Platform stack that runs the biggest CRM systems in the world is not going to run out of enterprise-grade features for your custom apps anytime soon.

In this case, we’re dealing with public decks from public conferences, though. So, let’s try a different route and embrace the Vibes.✌️

What could things be like if we start from the app maker experience instead of the existing platforms and products? This is what startups without the enterprise software moat need to figure out. The Swedish company that started in 2023 as GPT Engineer has grown into Lovable, which now has 1 million new app projects created every week.

There are also built-in connectors to Microsoft services like SharePoint, making this a great choice for my example scenario. Using nothing but the 5 free daily AI credits that Lovable gives to everyone, I was able to generate an Event Decks app that met all my requirements. And it literally took 15 minutes from prompt to done. Below is the result:

I spent way more time getting the SharePoint list customization JSON to work properly than it took Lovable AI to generate the whole package. That’s because the coding agent was free to build things in its native language, rather than trying to learn something new and specific to SharePoint. If Lovable generates a few hundred thousand new React apps every week, that’s a lot of data and practice rounds for the AI to improve its skills.

I find the standard SharePoint search experience to be unintuitive, requiring me to always think about “where do I click to get the right filters available?” With the app that Lovable built, I had complete control to set the context-specific search filters where I wanted them to appear. It’s like creating an online store catalog search for a scenario where you are both the merchant and the buyer of the items to be shown.

With content coming from public events and speakers with other online information to be discovered, letting AI enrich the UI with more than just strict document metadata should be doable. I didn’t have to configure or ask anything specific from Lovable’s app builder to achieve the speaker and event card showing up with AI-provided summaries:

The deck viewer control could also be adjusted if needed. Yet the real beef in using an app like this would come from the ability to append the viewed decks with your own tags and notes. That functionality was also shipped by the app builder agent, making data flow back from the viewer page to the SharePoint library. No need for bouncing between different views like the native SharePoint UX requires since now I had full control over what’s shown in the sidebar next to the slides.

All in all, the experience of building apps this way remains… simply lovable. Back in September 2025, this was how I started with the concept of what became the XrmToolBox Plugin Catalog site:

I talked about the journey from idea to live site. As it so often is with personal projects like this, you don’t set out to do a specific deliverable with a predefined schedule and budget. No, you start with just trying things out and you keep going because what you’ve managed to achieve builds up motivation to finish it. If everything feels too tedious and there’s no visible progress, you quickly find better things to do. But if you manage to reach the “hey, I’m onto something” level, that is just priceless:

This “OMG it works!!!” moment is exactly what powered the rise of Power BI and later the low-code app building experience we today call Power Platform. I’m going to repeat one of my favorite quotes about what enabled the citizen developer/analyst movement, taken from the GA announcement of Power BI back in 2015:

The problem is in what happens when the wow-factor is assimilated into the everyday reality of enterprise software. Nowhere is this more obvious than with Power BI. In my 2025 article that examined the rise of Fabric and the fall of Power BI (gift link), I gathered some evidence from the partner and customer community on how the complexity and cost of Fabric was being pushed onto them by Microsoft with little practical upside. Especially the lower end of the SMB market now seems to be left behind when their data needs aren’t big enough.

Making the simple PBI experience better (even “lovable”) wasn’t going to bring meaningful new revenue when Microsoft already dominated the Magic Quadrants for many years. This meant the investments were directed toward building an enterprise data platform in Fabric. If you look at the investment areas in the low-code application platform side, much of it seems to have gone into the Managed Environments side of enterprise capabilities. Meanwhile, many parts of the Power Apps maker experience remain stuck in the past as all bets have been on AI to become the new UI.

Everything grows bigger and becomes more cumbersome. Which in turn makes room for new, nimble challengers to disrupt the market dominated by the incumbents. Ah, the circle of life in software business.

Like I mentioned in an earlier newsletter issue about DynamicsMinds 2026, Microsoft is already internally using Lovable in many teams. They are creating mockup demos of their own upcoming products by using this tool. As of now, it’s not an “us vs. them” scenario but rather something that the software giant from Redmond wants to integrate into their AI app development story.

In addition to the connectors like SharePoint that I used in my EventDecks app, there’s more integration points on the way. Lovable agent for Microsoft Teams is coming soon. As part of the Build 2026 announcement for M365 devs, “Build collaborative agents where work happens”, Lovable has been mentioned among other partner agents like Linear and Cursor. The Agent Showcase page on the partner portal (no login required) presents a short demo video of what to expect:

You might be wondering “isn’t this the same process that App Builder agent in M365 Copilot supports?” Well, it currently covers a tiny fraction of the capabilities that a builder like Lovable can offer. For instance, I could not point the App Builder agent to an existing SharePoint site and ask it to make it pretty. There’s been so little news about App Builder since the preview launch (see my review) that I’m currently not sure the product will ever get out of the Frontier program — unless it gets “reimagined,” of course.

Another experience currently stuck in preview trenches is the Vibe Power Apps builder. It remains restricted to only select English-speaking markets. The explanation from Microsoft representatives has been that the unavailability of Anthropic models in the necessary non-US geos is slowing down the rollout. While capacity and compliance issues may well contribute to this, I somehow doubt those are the only reasons why everything’s not yet full-on Vibes when it comes to Power Apps.

In the grand scheme of things, a company the size of Microsoft will most often be both partnering and competing with other software providers at the same time. For instance, Copilot Cowork has been built on top of licensed IPR from Anthropic. Now, it is pitched as the prime example of Copilot being “the AI that actually does work stuff”, all while Claude Cowork is of course out in the market, too. The co-operation does not remove the competitive pressure. A recent post from Chris Smellie illustrates this everyday tension quite well:

I actually stole this newsletter issue’s cover image from that post by Chris and did some light AI-based editing to it, to make it clear how the photo of the Claude + Lovable billboard was taken from inside Microsoft’s offices. The ad’s slogan of “for optimists turning ideas into reality” was obviously a key factor that prompted the post to be written. It also happened to be at the core of this Perspectives issue draft that I was working on.

Copilot is not built for the creatives, nor for the creators. Whereas that has been a key market for Lovable, as can be seen from the recent builder economy 2026 report. Not only do 2 out of 3 app builders of Lovable come from outside the tech industry. The majority are creating products with monetization in mind — either as the immediate target or a side hustle that could potentially turn into something you can sell. Compare this with Power Apps or Power Pages that are solely targeted for internal tools or portals to append existing customer information systems.

Chris Smellie is right: there truly is no “one-AI-to-rule-them-all”. Even a company the size of Microsoft is unable to cover all the scenarios for AI-assisted app creation that exist out there. The focus of Power Platform will very likely remain on apps that people from the outside world are not meant to access. Founders wanting to do quick testing with public-facing assets and web apps will choose something like Lovable. Between those two extremes remains the murky middle where you should do a critical analysis on what the pros and cons of different tools and platforms are for your AI app scenario.

I made a quick video of my experience on turning the boring SharePoint site into a wonderful Lovable app. Check it out for also getting some insights into how it all works when bringing in additional tools like GitHub Copilot:

After watching it, I also recommend you check out this brand new video of my buddy Sean Astrakhan taking Lovable out for a test drive with a Dataverse backed website:

Watch this space for more insights on what tools like Lovable mean for the Microsoft business applications ecosystem. Meanwhile, if you’ve got any comments, questions or ideas about all this, remember that you can always reply to the newsletter emails or reach out via my company website.