[Me & X] SaaS is dead?
X: SaaS is dead.
Me: Oh dear. I suppose your argument is it's quicker to build your own than use a SaaS?
X: Exactly. You disagree?
Me: There's a slight problem.
X: ?
Me: SaaS was a marketing term.
X: Explain.
Me: Activities evolve from genesis to custom built to product to commodity. When we talk about services then they start to appear in custom built as timeshare, in product as rental and in commodity as utility services. SaaS was used as a catch all for all of these in the application space but timeshare, rental and utility are not the same. There are solid utility services out there. An easy example of this can be found in the infrastructure space with EC2 (compute as utility). In the software application space you have a mixed bag of some utility like services and then a whole bunch of things with providers who have not been industrialising and they've been exploiting ecosystems of consultants to customise. To be blunt, they've been pretending timeshare or rental models are utility models and they deserve to feel some pain. So, some SaaS is dead but that is not the same as all SaaS is dead.
X: How can I tell the difference?
Me: Try the cheat sheet, attached.
X: Why don't I just get an AI to build it all for me?
Me: There's no point in rebuilding highly industrialised utility services unless you are a glutton for punishment. Sure, there are lots of monstrous deathstar projects out there which are faking it by trying to be the platform for everything and are far from tightly defined utility services. Personally, I'd spend 30 minutes or so mapping out the space and then use that to determine what utility services you can use and where you might legitimately want to rebuild.
X: Seems easier to just get an AI to build it all for me.
Me: Well, I suppose the good news is that some LLM/GPTs have even been spontaneously providing Wardley Maps in PRDs they create. In any case your LLM/GPT will recommend using some services (which are utility like) and not others because they have a bias towards whatever is the most common, well documented path for a given problem (a useful but imperfect proxy). If this continues, they might start doing the "thinking" for you. That's not a bad thing given you seem so keen not to spend 30 minutes on this.
X: So, SaaS is dead?
Me: Timeshare and rental models faking it as a utility with large consultant ecosystems might face some trouble ... yes.
X: So, SaaS is dead?
Me: Why do I feel I might have a better conversation talking with your LLM/GPT?
X: I'm not interested in mapping stuff, I just want an answer!
Me: It's not binary. Ask your LLM/GPT ... oh, actually, don't bother. Just ask your "AI" to build your stuff and let it deal with the nuance.
X: So, I'm right? Let AI build it all, SaaS is dead?
Me: FFS. Sure, just when your LLM/GPT recommends using Vercel, SendGrid, Stripe and S3, just go along with it.
Originally published on LinkedIn.