A balanced look at the potential SaaSpocalypse
There are reasons to keep paying for SaaS - even if software is "free"
My TL;DR on SaaS
If the SaaS is reliable, meets your needs, is fairly priced and doesn’t get in the way of you restructuring to an agentic SDLC, go for it.
In my experience the real reason to get rid of SaaS products is if they:
Don’t allow you unlimited agentic access to all of your data
Force you to use their workflows or data structures - not yours
Assume users will be humans vs agents
Don’t support the automated, deterministic agentic SDLC workflows you’re trying to build.
Below is a longer list of some of the reasons you should keep - or replace - your SaaS subscriptions. I will note anecdotally that the bar is now way higher for me to start using a SaaS product (although I still use plenty of them).
Reasons to continue to pay for SaaS
Reliability - It just works. SaaS companies put a lot of effort into uptime so you don’t have to
Complexity - It encapsulates rich domain knowledge. SaaS companies worry deeply about authentication or payments or tax reporting or whatever they do, so you don’t have to - generally it’s not a leaky abstraction and you can focus on your core mission without getting in the weeds with table stakes features like auth.
Cost of Change - It’s installed. If you have SaaS in place, removing it requires replicating the functionality, changing your SOPs and retraining anyone on your team who might need to interact with it.
Network Effects - Some SaaS products deliver additional value by being part of a larger community. I could vibe code enough of substack to meet my needs in an afternoon, but I get value from potential referrals from other engineering leadership properties.
Business Relationships - I could probably rebuild stripe payments fairly quickly. Getting access to the banking networks to settle the transactions would be slower/harder and more expensive.
Data - Sometimes the real value of the SaaS is the industry benchmarking data. It’s not that hard to extract DORA like insights from Jira and GitHub, but most of the players in the software engineering intelligence space are working hard to provide additional insights and benchmarking data which might justify the cost of the subscriptions.
Reasons to not use SaaS
Licensing costs - For large companies the licensing costs may be sufficient to provide an ROI in cost savings alone.
Focused on human usage - People shouldn’t be using SaaS - their agents should. Any SaaS should treat agentic users as first class citizens.
Not relevant in an agentic org - A lot of the things some SaaS products do are simply not relevant with an agentic SDLC. I don’t need Jira or confluence to see the state of my software engineering activities - that can be reported in whatever ways I want by my agentic PMs.
Not optimized for your workflows - Historically we all used SaaS offerings that could do a good enough job of being adapted to any reasonable workflow. The question is whether it’s easier to configure Jira for your needs or generate a lightweight app that allows you to customize your workflow precisely and efficiently.
Data Vampires - Are they sucking the data out of your company? Make sure you don’t commit to any SaaS tools where you can’t agentically access all of the information that you’re persisting.