79 comments
It will probably continue to exist because most businesses are perfectly content to outsource tech problems to someone else for a reasonable fee, and if you think that managing an agent swarm is a seamless replacement for that, I don't know what you tell you. The vast majority of business owners don't want the added complexity in their life.
And for the same reason, most SaaS will probably not migrate to some radically new UX paradigm. Some of it will be vibecoded, but the UX will probably continue to be deterministic. I think this might change once a new breed of "LLM-native" business owners takes charge, but that's going to be a slow process, and it will be hampered by prosaic concerns about interoperability, support, predictable cost, and liability for mistakes (from "the agent did a bit of tax fraud" to "the agent decided to hack my competition").
Again, barring radical superhumanity, in many cases companies do not today want their own bespoke solutions to a problem and wouldn't want it even if the developers were working for free, because the other costs to the business would still be more than they want to pay and more than another business can charge to make the problem more thoroughly (even if not completely) go away.
I suspect this is another reason you may see some companies making grabs at data and points-of-presence that otherwise make little or no sense... they're trying to colonize and defend the sources of contact with the real world before someone else gets there and locks them out.
With insignificant concerns, at times it seems a business might not even care about the cost but that's also the same areas where a cut in quality is ignored if it comes with a cut in costs as well.
A lot of SasS fall into this category, so even if AI is not great, the value of getting a solution in under 5 minutes, simply by talking to a machine, is too great a value for a business.
All those marketing sites that might've been handled by 3rd party agencies in the 80s, became in-house clip art projects in the 90s, became SaaS sites in the 2000s, and will become "CEO told his dreams to the AI, and the AI delivered while they were being driven to work" projects in 2030s.
The fact that it's bespoke or not is besides the point when writing up the requirements means the site is essentially already built and there's no need to vet vendor at all.
Now for significant concerns, SaaS will likely still have a place, but some aspects of it will be minimized. Companies focused on Wordpress, Jira, and Salesforce customizations will fall by the wayside since the main company can just say "talk to our AI agent and it'll customize it for you".
It doesn't say SaaS is being replaced by agents, rather that SaaS companies will be transformed in the way described. Outsourcing will still happen, hence the discussion of headless components, and the continued demand for SaaS.
And part of the argument is that we won't just be waiting for the new breed of LLM-native companies; lots of existing businesses are being LLM-ified right now, through the process described in the article...
That said, your point can be distilled down to: comparative advantage. It still exists and so SaaS will likely still exist. But per the article, it'll probably look pretty different (and a lot less profitable) over time.
I think the reason it'll look different is that it's very hard to make powerful software that's also walk-up usable. The idea that you can provide just the core concepts of a piece of software and allow AI to interact with those concepts more directly, or allow your users to use AI to build their ideal (and ever-changing) interfaces to your concepts seems pretty convincing to me.
Of course, nothing is really preventing the SaaS side from operating the agent swarm, yes?
But to pull something from the post:
> If this is the right mental model, you should expect to see:
> An unusual amount of in-house harness building on both the build side and the sell side
Yup.
> Org structures and individual roles being reshaped around their place in the business harness
Not yet, this is the one I am most skeptical about because it's still way too easy for the harness-driven processes I've seen to go off the rails, so it's still very much human SME-driven. However, the number of butts in seats required is going down.
> AI-native startups beating incumbents in domains where the “moat” can be easily harness-ified
In most cases I've seen the "moat" is not easily harness-ified, and increasingly the money and energy seems to be going into getting to the starting line.
> All software a software company uses (on or tied to the core build or sell paths) needing to be headless so the outer harness can run it
Tech companies are definitely doing it, but I don't think this is actually realistic yet for companies whose core competence isn't software. But this feels like a matter of when, not if.
There are 40,000 McDonald’s franchisees who basically just put money into a coinop machine and the machine returns with exactly the food, drinks, restaurant decor, prices, etc that it wants them to sell.
Toyota’s production system coordinates what gets made and when, detects abnormalities and directs human attention toward problems and improvement. They explicitly eliminate the need for people to continuously watch machines, while preserving human judgment
Every contract is different and the majority of the work is "the boring inefficient stuff" that can't be automated away. That is, arguing with shifty and arrogant middle management in a long series of meetings until you get some concrete requirements. Then realizing the metric shit ton of crap to untangle and test rigorously. It's not just a matter of having patience, but being unflinchingly attentive the whole time for the nuggets of gold that fall out of their disgusting mouths and then pouncing with a plan so you don't go over time and budget. AI simply isn't going to grill people like that and then execute swiftly and precisely. A lot of these jobs really are like a murder interrogation and then finding the bodies.
It's absurd that so many people believe there are unturned stones in this space. If this kind of business could be cookie-cutter, it would have happened over a decade ago and it wouldn't have made any money.
But I believe a premise of the article is that AI-ification will disrupt the power of those middle managers, probably in a few different ways.
One is that the middle managers themselves will be replaced by AI. Another is that more purchasing decisions will be made by engineers building the corporate harness on the buyer side. And, probably people will be more accustomed (i.e. forced) to talk to AIs on the other side, or at least accept that the responses are ultimately constrained by them. Going out for a round of golf with the sales guy just can't accomplish anything anymore.
Every exec team wants this transition to happen, so both buy and sell side will go through great transformation in concert.
If we're talking automation in the way that humans define a process, boundary cases, etc up front then yes you can't really automate away one-off monotony.
If AI is actually artificial intelligence, it will be able to figure that out similar to how a human would.
I suspect for purposes of optimization and performance etc., sometime in the next several months we will start to see new popular ML architectures or variations of LLMs/VLMs that are designed to be _stateful_. So a lot of the infrastructure for managing state and memory etc. gets sucked into the model somehow.
They may end up changing or expanding the concept of an ML model to enable that.
That type of belief is what is making me anxious about AI engineering staying relevant for much longer. I just wonder if pretty soon we need to be able to build and train or customize ML models that just do everything, or at least know how to prompt an agent to do that.
Read the full thread on Hacker News →
Related stories
- The Verge · 0 points · 5 days ago
- Hacker News · 20 points · 6 days ago
- Hacker News · 1 points · 12 days ago
- DEV Community · 0 points · 13 days ago
- DEV Community · 1 points · 8 days ago
- Hacker News · 5 points · 9 days ago