Are enterprises abandoning custom AI apps for third-party solutions?

Andreessen Horowitz's 2025 survey of enterprise leaders finds that enterprises have reversed their early build-first posture toward AI applications. "Early in the AI product cycle, enterprises largely opted to work directly with AI models and build their own applications," the report states, but "we've seen a marked shift towards buying third party applications over the last twelve months as the ecosystem of AI apps has started to mature." It summarizes the market move as one where "off-the-shelf solutions are eclipsing custom builds and rewarding AI-native third party applications."

a16z attributes the reversal to two forces. First, model performance and cost keep shifting, so getting incremental ROI out of a given use case requires constant re-evaluation and optimization across models — work the report says is "often best tackled by a dedicated AI application team instead of an internal team." Second, maintenance burden: "in a space as dynamic as AI, companies are finding that internally developed tools are difficult to maintain and frequently don't give them a business advantage — which further cements their interest in buying instead of building apps." In other words, the pace of model change turns a build decision into a standing maintenance cost rather than a one-time engineering win.

This gives a buyer-side voice to Does easier tool-building actually solve enterprise adoption problems?: Evans's argument is that falling build costs don't resolve the harder problems of knowing what to build and getting it adopted, and a16z's CIOs describe acting on exactly that friction by abandoning in-house builds for vendor apps rather than solving it internally. It also corroborates Why do most enterprise AI pilots fail to deliver returns? from the other side: NANDA's case analysis finds buy-vs-build is the variable that tracks pilot success, and here enterprise leaders independently report moving toward buy for the same underlying reason — build doesn't hold up against a fast-moving model layer.

The excerpt reports a shift in enterprise intent and framing, not a controlled comparison of outcomes — it does not quantify how many enterprises switched, how much AI spend now goes to third-party apps versus internal builds, or whether bought applications actually perform better than the ones they replaced. It is also an investor's report: a16z has a commercial stake in a market where enterprises buy from AI-native startups rather than build in-house, which is reason to read the framing as directional rather than neutral. At the strength the evidence allows, the finding is that enterprise buyers increasingly believe build is a losing bet against the pace of model change, not that buying is proven to deliver more value.