Inside the new Power Apps: The future of App Development

No need to mince words

Microsoft Ignite is this week and yesterday Ryan Cunningham (future unemployed executive) showed off the future of the Power Platform via their new portal at https://vibe.powerapps.com/.  It is a complete fusion (hahahahaaaaa) of AI w/ human tools to build applications within the Power Platform framework.  It is fully targeting non-developers to build applications with only AI tools for assistance.

He and his bunch of also future-unemployed executives go on to explain how AI will do all the things you humans can.  Just tell the AI what you want and it will do everything for you.  Except you is now a completely different audience from anyone who has worked in this platform to date. 

Why bother selling your well-informed small audience on your new vision when you can go after a much larger audience who doesn't know any better and has no idea how many broken promises lie in your wake.

TLDR

• AI all the thingz (if a prompt doesn't work, prompt it again!)
• Old mode is still useable for those of us who insist on getting value from our investment (for now)
• New approach allows SharePoint vs. Dataverse in certain ways
• Most of this enforces Model (database driven) vs. Canvas (UX driven) 
• This is more of a marketing stunt to use the successful "Power Platform" existing framework to build out new things but slap the word "Power" in front of it (Power Pages <> Power Apps <> Power Automate <> Power Automate Desktop)

What About the Current Approach?

Well, they've got an entirely new URL for their new approach.  So I'd say they're going to leave the current stuff "as is" for some period of time.  That period?  I dunno man.  2-5 years probably (check back here in 2027).  Unless the AI bubble pops before then and perhaps then they'll pretend it was all a joke and put everything back the way it was.

https://make.powerapps.com and https://make.powerautomate.com seem to be continuing, but some of the weird new front-screens and stuff will probably just try to tease you over to the new https://vibe.powerapps.com site.  More on this later.  

NOTE: Maybe they'll make a new Power Platform crypto coin and let you earn some by watching their propaganda videos.  Wait...did I just create a new revenue stream for Microsoft?

Regardless, I think you can continue as-is for now w/o needing to change today, but I also think this lights a fuse under the "do we want to follow where this is going" question that I think everyone should be asking themselves.

I believe we're already seeing some of the wheels come off the bus though on this platform with some of the outages and slowdowns that appear to be increasing.  I don't think Microsoft is quite so focused on stability anymore, which is one of the key pillars of the existing platform.  Therefore, I think we all have to be honest with ourselves if we don't align w/ their future plans and start reviewing alternative paths.

Agentic AI Approach

So they're following an Agentic approach, which means that they're splitting up the "AI" (it is Artificial in that it is copying things you and I have built previously but it definitely is NOT Intelligent), into task-specific roles to try and more easily troubleshoot/fix their own BS behind the scenes because the "AI" isn't doing what they expected it to do.  

If you're trying to imagine what this looks like, just picture a Product Manager, a Developer, and a Support Agent.  Each one has their own role and is trained to do this.  To try and narrow down all the BS these "AI" Agents are doing wrong, they are mirroring what humans do and splitting the tasks.  So for your app building, you'll have Agents that represent similar roles.  

Microsoft published out a Q&D demo of this here:

While there are aspects of this that I've asked for for years, I wasn't specifically asking for "AI" to do any of this.  My main beef w/ the PowerApps is that I always wanted something that helped answer the question of: "What are the business rules this app should follow and how/where are these maintained/edited?"  I think that is a relevant question for business-facing application development.  Who writes these rules (and if they know WTF they're doing when writing rules), how they are managed, and what the downstream impact of changes made to them is kind of the purpose of these kinds of platforms.

In theory this lets the business self-manage.  In reality this breaks a ton of stuff downstream and the "AI" will not be fixing this automatically.  The person who changed the rules will then have to prompt the "AI" to fix things they also do not understand. Whether or not that the person knows what to prompt and whether or not the Agents understand what was asked... /shrug.

App Builder

If you're following this product, then you've probably seen App Builder.  I properly sh*t on this here.  There remain some scenarios where maybe people might use this and have it work for them.  However, what continues to be the issue is that this will break the brains of non-devs or devs who don't have a full Dev/QA/Prod setup/staffing for post-go-live changes.

It requires more intricate back-end to make this "appear" so simple.

I do enjoy that during the full presentation, Emma Cooper (future unemployed person) also demos the "make me a dark mode" app.  Which is just about the dumbest f*ing demo thing to show off.  I mean, your platform made this hard.  You asking the AI to make it easy is showing how much your team sucks at building features people want.

However, i will point out that they do build out parts of this on top of SharePoint.  Which is much better than originally shown off.  This is definitely an improvement.  Tiny clapping hands achievement unlocked! 

NOTE: Maybe we can find Emma a job cleaning up after the robot barista in their planned AI dystopian future.

However, this still brings us into the Model Apps vs. Canvas Apps conversation and my previous points on strict vs. loose data coupling between front/back-ends.  Although this is SharePoint, it again means we now have a VERY tightly coupled back-end that then controls the front-end.  This remains a design-choice and there are scenarios where this works well.  However, I continue to advocate for Canvas over Model  for most applications simply because it slows post-deployment changes or makes changes much more risky on this platform.

I suppose the "AI" is supposed to then resolve all of this but...good luck I guess.

Vibe.PowerApps.Com

Ryan starts this one off on how this is going to change the "Operating System" of departments.  Which given how much Microsoft is f*ing up Windows 11 (Windows ME, Windows Vista, Windows 8...), I think you know what kind of future this portends.  Simon (a PM) then saunters in to show us how a non-technical person builds things (but of course needs help connecting his device).

NOTE: Why didn't he ask Copilot to do that for him?

This is the "new portal" that is intended to replace make.powerapps.com (Oh sure!  You can keep using the old portal, just like VB6!).  It is supposedly their new business-facing portal where non-devs will work.

Now THIS takes me back to what I expect.  Any minor praise I gave them above for connecting App Builder where it can work w/ SharePoint is now reversed as you know the default will be Dataverse.  They are desperate for this to be used so they can charge you extra.  And they do this for non-tech users so people don't know they're racking up charges.

You know, just last weekend we had someone in our security team screw up and grant admin access to our PowerApps environment to someone who ultimately deleted it and destroyed about 80% of the production apps in our org (we were able to recover w/ some work).  However, because we completely ignore Microsoft's attempts to make us use Dataverse, this person DID NOT GAIN ACCESS TO OUR DATA AND ALSO DID NOT DELETE ANY OF OUR DATA.  Merging the security for apps/data definitely simplify things, and also makes it more catastrophic when things go wrong.  Sometimes firewalls between things help limit the damage when the inevitable screwup happens.  I look forward to the first time someone deletes "an app" and winds up deleting all of the associated data tables.

However, let me take just a small moment to NOT sh1t on what they've built here.  The workflow for building an app isn't terrible if you just removed the "AI" from it.  Giving people a kind of divided series of steps to kind of think through what they want to build is not bad on it's face.  

This does give you some control over the UX w/o having to re-prompt, which is nice.  Thanks I guess for doing the thing you should have done immediately.  But I also will say that for an app that is supposed to let non-tech users to build things, it certainly feels like you need to be technical to work in this environment  Don't get me wrong, it removes a lot of "coding", but it is still "technical".  I would never punish any of my HR users (even the technical ones) by sending them to this portal as it currently exists.

Regardless, it is reasonably nice to build a "workflow" vs. building "an app" in some respects.  However, I really feel like this is is just a completely different approach and a completely different product.  It still appears they're creating React web pages vs. Power Apps proper.  So I'm definitely feeling they're heading in this direction.  

I don't have to hate this approach, but I remain dubious of them doing this unless they really do resolve the problems/concerns w/ how/where front/back-end logic breaks down and how/where the code for certain things exists.  I do not hate them if this auto-builds the back-end validation but allows front-end style construction.  On the contrary, it's my dream feature! PowerApps kind of did for us originally as there is/was no such differentiation between front/back-end. You just built it in the front-end and it was secured to the point where end-users couldn't change things.

Ultimately, I feel like this change to React vs. the standard Power Apps interface is a two-fold benefit:

• Allows their "AI" to be trained identically to normal web pages so faster go-to-market
• Reduces overhead on web servers 

Neither of these are for you and I.  They are for Microsoft.  This benefits them vs. us.

I certainly can see a version of this that makes it an easier transition for web developers into this world and opens up some reasonable paths for them that might be slightly easier than the baseline Power App model, but I didn't seem to have a lot of trouble w/ any web developer moving into this platform except when they insisted on building everything like they built web pages in 2010.  Where they simply couldn't stop doing the technical things they'd been trained to do back then on a platform that didn't require it.

I do think it is worth watching this specific space closely.  I do not believe in any way you should use this for production or believe this is the future unless you work for Microsoft, but I do believe it shows us how Microsoft believes the future should look.  You certainly want to understand if you agree with them or not.

Is There a Clear Path to Success?

No.

What else do you want me to say here?

Unless you believe Microsoft is always right (Windows ME, Windows Vista, Windows 8, Windows 11...), I think there are huge questions in staffing, guardrails (vendor vs. internal ones even), ROI, and the internal reskilling required before we even get into if this all ultimately costs less per unit of value.  

Today's Low Code solution does solve very real problems conceptually and while it is something the enterprise fought, they eventually accepted the improved delivery pattern for Low vs. Standard code approaches.  The design principles remain fairly straightforward with pushback primarily being old habits that stopped people from being successful when they failed.

However, maybe you are the kind of person who thinks they'll definitely build this plane in time before you hit the ground.  I'm just not so inclined to follow along this time around.  This isn't just one risky bet, it's thousands of them.  I think most people are betting that they can make excuses when it all fails and say "everybody else said it would succeed"!  No.  Everybody else is not saying that.  If you've been around tech for >20 years, you should know this and know it well.

This might all eventually work, but their current vision for that new site is so full of holes as to make it a colander.  Their vision all depends on AI at every step.  There is no success w/o it.

Final Thoughts

You can watch the full session here. You might want to play a drinking game for every time Ryan ejaculates more hype for "AI" and how it's going to "save the world".  But I'd suggest you do a low-alcohol drink so you make it through the video.  But then again, you might wish you passed out 15 minutes in.

I cannot state strongly enough how all of this pisses me off.  Not just because I don't necessarily agree w/ their direction, but how they are attempting to turn the Power Platform as it exists into abandonware.  Whether or not they are successful in some ways come down to all of us and if we choose to lean into the direction they're trying to tempt us to follow.

The primary problem that the existing PowerApps/Automate solution solves for is that it allows us to deliver applications that can be secured and delivered to an array of endpoints within the enterprise (and to to a limited degree outside of the enterprise) very easily compared to more traditional methods.  It is pretty slick and quite nice to be able to have a native app on a mobile phone access the camera.  It is not a tiny thing for us.

However, the rest of...this...isn't a problem we're having.  We are already much more responsive to the business and they are VERY happy w/ how we are retooling to be more responsive to their needs.  Shoving the IT department out of the way to tell us we're all a bunch of slow pokes and Microsoft is now going to solve all their problems when Microsoft was a good chunk of why IT was previously slow isn't a way for any of us to want to jump on board.  Which is a large part of how/why they are pushing things the way they are through the more common end-user-facing applications.

I think there is again a question here for us to consider:  do we trust Microsoft more than we trust our own IT departments.  I will always admit that our local IT teams aren't always aligned w/ organizational goals, but I think this is definitely a wakeup call for everyone.  AI is attempting to make the machine the trusted advisor for business users.  I think that isn't simply a slippery slope, but one w/ a spiked pit at the bottom of it.

I'd say this will be the Microsoft Bob of products, but it will be so much worse than that because lots of people will bully others into using it since they need others to verify their delusional belief in AI instead of a canine version of Clippy.

What the AI advocates hear in their head:  

What the rest of us hear: