Consolidating the modern tech stack with AI

AI’s Impact on the Enterprise Tech Stack: Consolidation, Agents, and a Human-Centric Future

The discussion at Web Summit Vancouver 2026 opened with the observation that many companies don’t have a cohesive “tech stack” but rather a “blob” or “morass” of numerous applications. Moderator Mr. John Koetsier highlighted that the average company uses about one piece of software per employee, with large enterprises often running over a thousand different applications, both official and unofficial. This complex landscape sets the stage for AI’s transformative role.

Mr. Wen Sang, Co-founder and COO of Genspark.ai, suggested that AI is fundamentally changing this dynamic. He posited that lighter applications will be completely absorbed by AI, while heavier systems, like Salesforce, will become “headless,” meaning their user interface is detached, and AI agents will interact with them via APIs. Mr. Sang believes AI will become the new user of software, leading to an explosion in SaaS usage due to AI’s speed in utilizing these products.

Mr. David Shim, Co-founder and CEO of Read AI, drew a parallel to the early internet’s “Cambrian explosion” of websites, which eventually consolidated. He anticipates a similar consolidation in AI tools, where companies will establish a “system of record” for specific functions like meetings or sales. This consolidation will be driven by the need to manage costs and simplify the overwhelming number of available solutions.

Mr. Koetsier elaborated on “headless” software, explaining that instead of human users manually inputting data into a user interface, an AI agent can connect via APIs to perform tasks, such as updating customer records in Salesforce. This shift promises greater efficiency and accuracy, moving away from slow, mistake-prone manual processes. Salesforce’s move to a headless architecture, allowing integration with tools like OpenClaw, exemplifies this trend.

Mr. Avery Pennarun, Co-founder and CEO of Tailscale, described AI as a “universal translator” capable of bridging not only human languages but also disparate software systems. He noted that connecting various enterprise applications, like Salesforce to a Snowflake database for dashboards, has historically been a monumental challenge for IT teams. AI now enables such integrations in minutes, potentially leading to even more software being adopted, but with seamless, computer-managed connections.

The panel then addressed the concept of a “SaaS apocalypse.” Mr. Shim argued it is “overrated,” citing companies like Data Dog experiencing re-accelerated growth due to AI. While some companies might see slower growth, he emphasized that the money isn’t disappearing but rather shifting, leading to lower valuations for those not adapting. Mr. Sang, however, pointed to Salesforce’s significant market cap decrease as evidence of a real impact, stressing that companies must leverage AI to create new value or risk decline.

Mr. Sang highlighted a shift from a “tool-centric” world, where individuals learned dozens of tools, to an “agent-centric” one. He envisions users commanding an AI agent, like Genspark.ai’s “claw,” to perform tasks across various applications without manual context switching. This allows humans to focus on strategic and creative work, offloading busy work to AI. Mr. Shim agreed, foreseeing AI operating within existing workflows, subtly guiding users with optimized options.

Mr. Pennarun expressed a more idealistic vision for the AI-native stack, seeing it as an opportunity to create human-centric systems. He believes AI’s ability to follow instructions empowers individuals to build custom software tailored to their exact needs, rather than being constrained by mega-corporations’ offerings. This means users can deploy their own chatbots to interact with websites, transforming frustrating support experiences into empowering ones.

Regarding maintenance, Mr. Pennarun noted that digital assets are unique in not degrading over time, yet software often stops working due to underlying platform changes. He suggested AI could allow users to instruct their custom-built tools to remain unchanged, eliminating maintenance needs. Alternatively, he proposed that code generated by AI could become disposable, simply regenerated when requirements shift. The panelists concluded by advising businesses to embrace AI for strategic tasks and build customizations around robust, approved AI platforms.