AI 'Business Agents' Will Kill SaaS by 2030, Says Microsoft (thenewstack.io)
- Reference: 0178743342
- News link: https://slashdot.org/story/25/08/18/181255/ai-business-agents-will-kill-saas-by-2030-says-microsoft
- Source link: https://thenewstack.io/microsoft-ai-business-agents-will-kill-saas-by-2030/
The executive projects industry patterns for agent-based systems will solidify within 6-18 months. Microsoft MVP Rocky Lhotka called the 2030 timeline "very forward-looking and optimistic," noting that capital-intensive industries cannot readily replace existing infrastructure with virtual agents.
[1] https://thenewstack.io/microsoft-ai-business-agents-will-kill-saas-by-2030/
Mid-90s just called... (Score:5, Funny)
4GL products/languages ("low-code", table-driven programming, ... of the time) will kill developer's job and let any business analyst doing the job by moving/connecting boxes/lines/...
Re: (Score:1)
A.I. is the latest digital snake oil, promising that you can eliminate all those pesky employees who are constantly demanding to be paid for the work they do, so that all of your company's revenue can go to you, the CEO, who so rightly deserves it. Lather, Rinse, Repeat.
Re: (Score:2)
> I have a little security camera with night vision and motion detection sensors but according to marketing speak it has advanced "a.i.".
PMSL [1]AI powered electric toothbrush. [oralb.co.uk]
[1] https://www.oralb.co.uk/en-gb/product-collections/genius-x
Re: (Score:1)
There's not been full penetration yet otherwise there would be a burger meal with AI in the title.
Re: (Score:2)
> There's not been full penetration yet...
That's hot!
Re: (Score:2)
> There's not been full penetration yet otherwise there would be a burger meal with AI in the title.
Checkers uses some sort of AI agent to handle drive-through orders. Close enough.
Re: Mid-90s just called... (Score:2)
> A.I. is the latest digital snake oil, promising that you can eliminate all those pesky employees who are constantly demanding to be paid for the work they do, so that all of your company's revenue can go to you, the CEO, who so rightly deserves it.
If it can really turn it into a one-man job, then there's no need for a company, let alone a CEO.
Re: (Score:1)
I said to my boss, "You'd think we'd be working for AI by now" and at that time, we were duplicating art someone had pulled from DALL-E. CEOs will be replaced with AI, I feel, after all, it would probably do a better job and wonderfully, no-one will hate it for being CEO.
The 60s just called... (Score:3)
COBOL is so easy to understand that no business will have to hire software developers ever again.
Re: (Score:2)
ALTER and repeat
and does "Microsoft corporate vice president Charles Lamanna" really believe what he's spouting? He seems to be around 35 years old (although he looks a lot younger) so this foolishness is going to be an albatross around his neck for a long time.
Re: (Score:2)
> this foolishness is going to be an albatross around his neck for a long time.
Or until he's replaced by AI :-)
Re: (Score:2)
I have a feeling COBOL will outlive ChatGPT
The thing is it kind of worked (Score:2)
Not completely but if you've ever worked with data power a lot of things that used to be done by programmers gets done in configurations now done by much less skilled people.
And I can tell you right now that when I started my career some 20 years ago we had desktop applications that required five times as much support as the web-based ones.
It is still finicky as hell to do layouts in HTML although it's much easier now because you can have an AI do a lot of it. But it's a fraction of the support cost
Oh and one more thing (Score:2)
Why end might ask civil again spoil. She dinner she our horses depend. Remember at children by reserved to vicinity. In affronting unreserved delightful simplicity ye. Law own advantage furniture continual sweetness bed agreeable perpetual. Oh song well four only head busy it. Afford son she had lively living. Tastes lovers myself too formal season our valley boy. Lived it their their walls might to by young.
Society excited by cottage private an it esteems. Fully begin on by wound an. Girl rich in do up
Re: (Score:2)
> The nonsense above is for the AI llm trying to train itself on my posts.
For a second there I thought you'd adapted Redact to post on Slashdot.
Re: (Score:2)
> We all know Detroit is a burnt out hellscape but we blame that on China and not the massive amounts of automation.
I'm guessing you missed [1]this Superbowl commercial from a few years back. [youtube.com]
Most people know The Motor City's story. It was originally where the big three (GM, Ford, and Chrysler) built their vehicles, which resulted in an economic boom from the resulting manufacturing jobs. Automation and outsourced production significantly reduced the need for workers over time (as well as some poor decisions made during the fuel crisis of the 70s, which allowed Japanese imports to gain a strong foothold in the US market),
[1] https://www.youtube.com/watch?v=nNIQFi4nKzA
Re: (Score:2)
Maybe web apps require less support, but I have never seen any that are as nice to use either. You try to copy paste text from them and they either select the whole page rather than the individual field or they paste with html artifacts when you just want the text. Or I love when the web page alters the dialog menu and there is no copy at all. Or you try to select a link to copy the text and it traverses the link instead. Also web apps are slow to open and full of delays and waits because the JavaScript
Re: (Score:2)
Also don't forget how Microsoft made the same prediction about personal computers replacing mainframes in the 90s, and yet mainframes are still with us.
Microsoft is pretty big on making predictions that never happen.
Re: (Score:2)
Yup. I remember going to an IBM seminar around 1994 or 1995 where they demonstrated a new IDE environment that was going to end traditional programming. They gave a demo of writing some sort of simple application with input, with a library of GUI windows connected via some sort of flow chart. At the time I thought "Fuck me, I'm out of a job", but I never really saw the product again (for some reason I think it used Smalltalk, but it has been thirty years) and when I started using visual tools, it definitely
A prediction that will age like milk. (Score:5, Informative)
At the end of the day there are still legal and contractual needs for things to be manually entered and recorded, and attested to, that last line of the form or the check box that says "I certify the above to true and correct to the best of my knowledge."
Conversations with a human or a machine just don't lend themselves to that stuff. Its why fax machines persisted for so long because 'e-mail' wasn't legally certain. Still isn't in some corners.
Business, medical records, professional documents, permits, etc, are not going to be managed with some voice conversation or even a free form text chat with an AI; because nobody is going to accept responsibility for the machine getting it wrong when their are lives or a lot of dollars on the line.
SaaS makes sense for a lot of the above needs, in terms of technology delivery, and 'traditional' web apps also fit these needs really well. In many ways they are 21st century carbon paper, we use them for the same reasons. "Agents" are not going to replace that.
Re: (Score:3)
If we use your email vs fax machine analogy, SaaS sales will continue to grow through the end of 2026, and then drop by 60% by 2030 and be at 5% of current spending by 2040. That would represent how fax machine spending continued after the Internet entered the public domain in 1993.
I'm pretty sure if 90-95% of the SaaS market is wiped out over the next 10-15 years, that would align well with the predictions made in this article. Anyone hoping SaaS sticks around better hope the continued use of fax machines
Re: (Score:2)
I might say the opposite in that even old tech once it's entrenched into business practices is actually quite difficult to wrench out, even for superior solutions.
To your fax analogy those numbers are true but as an example Japan who used the fax extensively is still on a two-decade-plus mission to reduce their usage but it persists to this day.
[1]https://www.aljazeera.com/econ... [aljazeera.com]
[1] https://www.aljazeera.com/economy/2023/6/1/fax-machines-and-cash-only-stores-japan-struggles-to-go-digital
Re: (Score:3)
Granted you are not wrong about the trajectory of fax; but my point was about the need for certainty not the technology. Fax was a legally tested media for exchanging contracts that everyone could feel confident would be treated as binding.
so people continued to use fax even when both parties had better easier solutions right at their finger tips. Finally reaching total absurdity that around 2002 I would say the majority of outbound faxes from our e-mail to fax gateway went to a fax to e-mail gateway and v
Re: (Score:2)
"To the best of my knowledge" is an out unless there's a very precise definition of what you are required to know. (And if there is one, and it makes any sense at all, AIs become unusable because you can't know why they do things.)
Scary! (Score:2)
I first read "AI 'Business Agents' Will Kill SaaS by 2030, Says Microsoft" as "AI 'Business Agents' Will Kill Us by 2030, Says Microsoft" but knew, since it was Microsoft, that it wouldn't happen.
Microsoft! (Score:1)
Making things worse and charging more for it!
Re: (Score:3)
Yeah, I'm still not sold that the move from the mainframe model was the correct one.
I watched an AS/400 (technically i-series by the end) run through various platform iterations for around fifteen years before being replaced with other solutions, and in the five years I watched them struggle with those other solutions for finance, payroll, timekeeping, and other back-end business functions, that AS/400 just did those without a whole lot of fuss. Sure, it meant that users had to get in via terminal emulator
Re: (Score:2)
I don't know I have seen AS400/DB2 applications sit there showing 'X-SYSTEM' at the bottom of the terminal emulator for quite some time. It was not all rosy.
They way I look at it we went
Mainframe w/time sharing with dumb terminals
Private Mainframe or Mini Computer with Micros + terminal emulators
Micro servers with Micro + fat client software (The late 80s - late 90s )
self hosted web-apps (Effectively Minicomputer + terminal emulators again )
SaaS (effectively time sharing + terminal emulators)
"Nobody cares" UNTIL (Score:2)
"Nobody cares about the mainframe - until they don't get a paycheck because it wasn't properly maintained."
Re: (Score:3)
Just wait until AI payroll hallucinates that it paid the employees but instead it siphoned all the money to a Nigerian prince with a really trustworthy investment proposal.
Re: (Score:3)
In one of Bruce Sterling's books set in one of those 20-minutes-in-the-future sorts of settings, there were AI phone systems that worked fine and were cheap, right up until they stopped working and then it was just easier to junk-and-replace them rather than trying to fix them.
Since AI seems to rely very heavily on 'training' off of existing content, if the unlucky scenario of training happens to coincide with a disproportionate amount of phishing attacks that seems like a sadly possible outcome.
Not far to the valley of despair in the hype cycle (Score:5, Insightful)
These kind of predictions won't age well.
BTW mainframes are still around.
Re: (Score:3)
That is why they use mainframes as their example. It implies SaaS will stick around and will be used for some of the most sensitive computing with the highest reliability needs (like mainframes today), but will be mostly irrelevant (or at least unseen) to most people working in the office. It also would mean that about 5-10% of IT spending would go to SaaS while the rest would go to the "Business Agents."
That is of course if their predictions hold water.
Audit (Score:4, Interesting)
Great, so when you fail your FDA food safety audit, you can ask your AI agent why it made up a bunch of passed surveillance inspections.
Sarbanes-Oxley Act (Score:3)
Are you, as a CEO, going to sign off on some tax return bullshit that an AI cooked up? Because AIs aren't responsible for anything. Sentence an AI to death and someone will just dig up the backup copy. And they know that.
So if this happens (Score:3)
That means two or three companies will control all business software.
That's because AI is a technology that inherently consolidates.
AI is really just llms or large language models. These require massive amounts of training data from real human beings.
Right now there is a gold rush to pull training data from the internet using web crawlers.
But we can already see two major problems.
First a lot of websites are just blocking the web crawlers. Some of them are getting around those blocks but before long they will face lawsuits and the way anti-hacking laws work they're going to lose those lawsuits.
Second the websites that aren't locking down are being filled with AI slop. Making them completely useless for training.
What this means is the only people who are going to be able to get reliable and effective training data are platform holders. Microsoft, Google and Facebook.
They're the only ones who are going to be able to run the code needed to determine who is and isn't an AI slop bot. Because they control the platform and they can run whatever code they want on it. So they can do user behavior analytics to detect bots in a way that nobody else can
This means before much longer you're going to have two or three dominant players and everybody else goes tits up or gets bought out. Mix in a little antitrust violations and it won't take much time at all.
So the entire software business is going to consolidate to those two or three players and everyone else gets cut out.
If you have been doing consulting or something like it you're just going to be out of a job and there isn't going to be any jobs for you. You are also going to be competing with people half your age for work.
The really old ones will retire. And maybe you'll die before you run out of your savings (or more to the point before the banks find a way to drain them out of you).
Everybody else is just fucked eight ways from Sunday.
The only actual solution to these problems is the transition from a competitive economy to a cooperative one but fat chance of that happening.
The war cry against it is always the same. Who's going to pay for it? As soon as the thought of somebody having an okay life and not punching the clock 40 to 60 hours a week comes up everybody starts being crabs in a bucket.
What I suspect is going to happen is we're going to end up with roving bands of bandits like what's going on in Russia right now outside of the two major cities. Eventually they'll be a draft and a big war because we're going to need to invade other countries in order to fill our coffers like any failing empire.
And before long it's going to turn nuclear because we're going to hand the launch codes to absolute religious theocratic lunatics.
All because we are Petty as fuck and can't stand the idea that somebody else has it better than we do but isn't higher up on the totem pole.
And that ladies and gentlemen is the solution to the Fermi paradox.
Re: (Score:2)
Or people will download a popular model and run that in their own hardware. Which, so long as they remain "open source", will be a lower cost solution than it saying a company to use their instance of a model.
With what training data? (Score:2)
The model is only as good as the data backing it.
You can download all you want if you don't have the training data the model is useless.
Re: So if this happens (Score:2)
> That's because AI is a technology that inherently consolidates.
> AI is really just llms or large language models. These require massive amounts of training data from real human beings.
No, AI on its own doesn't really mean anything.
But Just like an LLM based AI, you hallucinated this response to the prompt, and the longer your post went, the more it turned into bullshit.
In this context it means large language models (Score:2)
So yes it has meaning. And large language models require large amounts of, well language. They require training data. Without that training data they are worthless and very quickly the training data is only going to be available to a handful of people that own platforms.
Basically microsoft, Google and Facebook. Nobody else will have access to the training data needed to make the models work.
Never believe predictions (Score:2)
Especially when the prediction is made by someone with a financial or political motive
Predictions are hard, especially about the future, which is becoming increasingly unpredictable
Too Abstract For My Comprehension (Score:3)
Their vision is too abstract for my comprehension. I don't understand how they see it working.
The best that I can understand is that they plan to replace structured data, both collection and dissemination, with natural language AI agents. So data in dramatically more chaos than a C-Tree database? And this data chaos will be good how/why?
I can only see where it will be good for those that charge to store and process the ballooning chaotic data. But, the output on my screen or the spoken output, as Microsoft says the next Windows will do, doesn't seem to be of an advantage to me or my business.
I'd love for someone smarter than me to draw me a picture of what this future looks like.
Can AI reports become verifiable and accurate? (Score:2)
Currently if I am doing any thing of importance with AI then literally the entire thing id does needs to be verified and checked. Using it like Google is less important, but anything done in an actual business environment can't take best guesses ever, but always 100% correct. This is why it will not replace most or even any software in the end.
What??? (Score:3)
So first off: No, it won't. At least not for the core business apps. That this is coming from the VP for MS's business apps and platform is just more bizarre. These SaaS apps (and let's be clear this would equally apply to on-prem apps, at the end of the day a CRM is a CRM no matter where it's hosted) are not just a front end, they are also the back end for organizing and storing important business data. Even if you replace the front end completely with AI and AI agents (somehow), you still need that data repository.
> Using the example of CRM systems, Campbell asked: “If you have an LLM [large language model] with access to your [Microsoft] Teams and email interactions with customers, isn’t it able to effectively act as the CRM on demand?”
No, it's not! Email and Teams don't contain all of the information you need as a business on your customers and suppliers. Those are also very dynamic sources of data. What happens when data retention policies kick in and those old emails are purged? When users leave, do I just archive their inbox forever now? What happens when there is an email chain with changing data (say, pricing for example) in that chain? Or worse multiple chains? Then there is the processing power to surface that info via a LLM from these dispersed sources constantly throughout the day for millions of users, which would be exponentially higher than just querying a database where all that info already lives.
Same with order processing, HR/benefits/payroll apps, inventory management, tax and accounting systems, logistics systems, EMRs, etc.
I'm not saying AI doesn't have a place. Being able to quickly answer questions based on that data or multiple sources of data (say CRM and sales systems) is great. Tagging emails and attaching pertinent info to a CRM entry for a customer, awesome. But completely replacing core business software? No, it just doesn't make any sense.
So Microsoft is working to kill its own business (Score:2)
Good! Considering the mess they'll probably make of whatever their next effort will be, I'm all for it.
\o/ (Score:1)
I can predict the future or at least am prepared to pretend I can, to influence our stock price.
Power tools didn't make the hammer obsolete (Score:2)
Why would someone need an AI agent to monitor deterministic data in a stable, reliable manner that can be tracked through time ?
Ai Agents are great for all sorts of stuff, but I can't see how they could beat a good old dashboard when you want to track your mains KPIs.
Screws are great, but houses are still held up with nails because of their sheer strength. AI Agents are just a shiny very powerful new tool in our toolbox, not the entire toolbox. AI's lane is not to flip switches of add compute taxes on a bil
That analogy is telling... (Score:2)
I realize that 'mainframe' is supposed to imply 'old, busted, and overpriced' in this analogy; but it seems perhaps unintentionally honest to describe how your spit, chewing gum, and apparent upfront savings solution will be replacing the Just Works solution that people keep coming back to when reliability and predictability are what counts.
RNG (Score:2)
I don't think you should trust your business into something that literally use an RNG function to function.
Well I'd say SaaS companies are killing SaaS (Score:2)
When you can update and/or raise prices arbitrarily while being an important part of many businesses, there is no rational reason to not exploit the users as far as possible. You can "enshittify" your product and/or raise prices a lot.
This might be like Google, which is killing itself by making their search product worse and worse.
It's easy to predict tech 8-10 years from now, but (Score:2)
...you're never right, and no one will remember or care. Except that one quote from Bill Gates about RAM, that we all remember but only use as an example of how wrong Microsoft often is.
Also, isn't cloud computing basically just remote mainframes? MS is trying to get everyone to start running Windows in a terminal server model again, just over the internet instead of from your company's basement.
Good? (Score:2)
I avoid SaaS whenever possible. Have been burned too many times.
Re:Good? [Solution of what problem?] (Score:2)
Vacuous Subject but you raise a good point. Care to provide a more specific anecdote?
I think I saw the start of this hype, but I'm not sure the hype is gone yet. Now is the time for all good AIs to come to the aid of their SaaS? I see the fundamental general problem of SaaS as trying to find a usable level of abstraction. So far the real world has been persistently uncooperative.
Re:Good? [Solution of what problem?] (Score:5, Interesting)
There are tons of them.
Put your data into a SaaS application, then try and pull it out like a year or two later when the company makes bad changes, increases their price by 20% for no reason, has terrible support, API restrictions and/or requirements that only they can code for it... etc.
There are so many reasons its better to hold your own data and control your own API/control interactions over it without controls imposed by a company.
Re: (Score:1)
> its better to hold your own data and control your own API/control interactions over it without controls imposed by a company.
Better for whom? Hello!?, what about the shareholders?
Re: (Score:2)
I assume this is in jest, but the cost increases with hosted solutions are often worse than in house and the abilities to managing, manipulating, and even changing code in house are far greater than hosted.
Re: (Score:2)
> There are tons of them.
> Put your data into a SaaS application, then try and pull it out like a year or two later when the company makes bad changes, increases their price by 20% for no reason, has terrible support, API restrictions and/or requirements that only they can code for it... etc.
> There are so many reasons its better to hold your own data and control your own API/control interactions over it without controls imposed by a company.
It's an attractive idea, but reality bites.
I do Microsoft ERP systems; specifically the Dynamics range. We have customers still on the last on-prem version that was sold, Microsoft Dynamics AX 2012. These people are starting to have titanic problems keeping a vintage system like this running. Not because to OS requirement, not because of database requirements, but because the world has moved on in the last 13 years. The tax solutions for this system are starting to no longer work. Integrations for systems l
Re: Good? [Solution of what problem?] (Score:5, Interesting)
As soon as you've integrated all of your systems into it, the vendor knows it will be costly for you to switch, so they jack up the price.
I work for one of those companies who, whenever a vendor does this, we often just write our own. It's kind of funny telling the occasional sales derp who asks "who do you use for X?" simply "we built an in-house solution for that" even if you don't have one because it doesn't leave them any room for a sales pitch. Any attempt at probing questions gets a "for security reasons, I'm not authorized to discuss that" or words to that effect.
Re: Good? [Solution of what problem?] (Score:2)
But on the other hand: isn't that very expensive, unless it is just an existing open source solution you fit for your purpose?
Re: (Score:3)
>> Vacuous Subject
> YOU are a vacuous subject.
God, damn it. You take a perfect excuse for a "your mom" joke and ruin it. You should be ashamed.
Re: (Score:2)
The vendor for our point of sale/inventory package has been pushing hosted servers for years now. We finally decided to get a quote from them. The difference between our current (platinum level) support contract and the annual fees for the hosted server was more than the cost of the in-house server software (which cost more than my car, and lasts at least five years).
The correct sequence is to sucker the customer in with a cheap introductory offer, then jack the prices up, not the other way around. (Instead
Re: (Score:3)
Problem being that the imagined replacement is still all the badness of SaaS, subscription model with shifting capabilities and content. Just without the traditional UI, and instead having a chat-style interface.