1) Adoption: It's really easy to build AI content or AI agents now. One of the easiest ways to look impressive these days is to show the walls of text an agent outputs. But people don't read any of these, and adoption stays low. There's good work to be done on background agent development where agents already complete tasks. See more here: https://alpergondiken.substack.com/p/agents-are-commoditized 2) Signal: Again, content is easy to produce, but quality is extremely low in many AI solutions out there now. We've seen this first-hand with Glean. There's a good bit of context-engineering needed to get strong signals. See more here: https://alpergondiken.substack.com/p/the-noise-factory 3) Build vs Buy: This depends on a lot of factors like internal talent, budget, etc but ultimately I've found build vs buy decisions are pretty different for AI SaaS vendors vs other SaaS vendors. It's very important to maintain control through connectability while also remove some of the technical ownership from internal shoulders. See more here: https://substack.com/home/post/p-172726907