HN Explorer

Ask HN: How does your organization make build vs. buy decisions now?

1 points by bigblind ↗ HN
With AI, a lot of things are way easier to build than they once were. So I wonder how big organisations now decide whether to build vs. buy. Building is easy, but you still need to maintain and support everything you build in-house. I'm especially curious how this varies by organisation size.

6 comments

[ 1837 ms ] story [ 970 ms ] thread
prasadvara
If I am not wrong, orgs build vs buy for the most part is NOT impacted by AI cycle.

For the most part business care about, is this our core capability OR not?? If core, yes build is what preferred. If not, buy is preferred. Where few occasions they deviate from this due to, to be frank non-technical reasons.

In Software world, building is not the costly, maintaining it is - and this cycle of AI developed software is NOT proven yet. So, so far my understanding is nothing changed in that area.

thorin
Agree with the other reply. Working for enterprises the actual initial writing the code part is probably 10% of the cost or less. Several times larger is the working out what the business requirements might be, which is mainly talking to people and writing stuff down. The support and maintenance part is also massive and if you assume a 5-10 year lifespan costs much more than the initial build.

I think what probably has improved is the ability of the business to prototype ideas for new systems and products quickly and throw them away if they don't work as expected.

[dead] ayda
[flagged]
datatrav
I can agree that writing code was never really the expensive part and that upkeep is, but what has really changed is scope. Noone in their right mind would want to "build salesforce" on their own. But if you have a team that is using 10% of salesforce and paying premium salesforce licensing, then an opportunity has presented itself. I can build a simple CRM customized to a team's workflow for a fraction of the annual licensing costs. That's a much smaller thing to maintain and thats where the math started making sense for us.

our process is workflow-first, not tech-first. We ask if improving the workflow is worth real money and how much and how much does it take to keep it alive for 5 years? If it isnt big, we stop, doesnt matter how easy it would be.

I have 14 applications in production right now, they are incredibly well documented and completely managed by AI agents that I deploy and manage. And yes, they have been externally/independently inspected for security and accuracy.