CHONGQING, CHINA - FEBRUARY 1: In this photo illustration, a smartphone displays the Moltbook website homepage, describing the platform as a social network for artificial intelligence agents, with a Moltbook character graphic visible in the background, on February 1, 2026, in Chongqing, China. Moltbook is a newly emerged social network exclusively for artificial intelligence agents, where autonomous AIs can post, comment and interact with one another without human participation, drawing widespread attention and debate in global technology and ethics communities about the implications of AI-to-AI communication and autonomy. (Photo illustration by Cheng Xin/Getty Images)
Getty Images
Aaron Levie spent a week in April visiting dozens of IT and AI leaders across banking, retail, healthcare, and media. His verdict: enterprises have stopped piloting AI agents and started budgeting for them, and the friction blocking adoption has almost nothing to do with the models. It is about legacy data systems, OpEx ceilings, and a shortage of engineers who can wire agents into real workflows. That last constraint will redefine what enterprise software vendors can charge and which startups will matter most to investors over the next three years.
The shift Levie describes on X on April 11, 2026 is consistent with what enterprise data shows. Gartner projects that 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025. That is not a gradual adoption curve; it is a platform change, and it is happening while most enterprise software vendor contracts are still written for the per-seat world that agents will dismantle.
Box, where Levie serves as CEO, reported fiscal 2026 revenue of $1.177 billion, up 8% year-over-year. Its Enterprise Advanced tier, which bundles intelligent workflow automation and advanced AI capabilities, reached 10% of total revenue within a year of launch. That growth line matters to investors precisely because it represents companies committing recurring budget to agentic infrastructure, not running one-off experiments.
One observation from Levie's dispatch that should matter to investors: enterprises are running into hard OpEx walls when trying to scale agents. Annual budgets were locked before token consumption was a line item, so companies are now running internal pitch competitions to allocate compute. One company Levie describes used a Shark Tank format to let business units bid for inference budget.
FinOps tooling for AI inference, which covers token cost attribution, usage governance, and multi-model routing, is attracting early capital. It also explains why AI agent startups collectively raised $3.8 billion in 2024, nearly triple the prior year, with accelerating rounds into 2026. Investors are betting that the infrastructure layer below the agent sits between foundation models and enterprise budgets, and whoever owns that layer controls margin.
Levie's most actionable observation for infrastructure investors is about data, not models. Enterprises cannot route agents to fragmented, on-premise, or partially cloudified systems in a unified way. The bottleneck is not reasoning capability; it is data access. As Levie noted on his earnings call, 90% of corporate data remains unstructured and underutilized. Agents that cannot reach that data cannot execute work in any meaningful sense.
This is the opening for a specific type of enterprise middleware startup: companies that sit between legacy systems and agent orchestration layers, translating data schemas, enforcing governance policies, and providing the audit trail that regulated industries require. IDC estimates that agentic AI already represents 10 to 15% of enterprise IT spending in 2026. The portion earmarked for data modernization to support agents is likely larger than the portion spent on agents themselves.
Levie's note that enterprises will eject vendors who make headless operation technically or economically difficult is a pricing threat, not just a technical preference. In an agent-dominated workflow, software that requires human interaction to function becomes a bottleneck. Vendors who have built around UI-centric product experiences face structural margin pressure as their seat counts flatten and agent-compatible APIs become table stakes.
Levie's observation that enterprises are operating in multi-agent environments by default, because no single vendor can cover every workflow, means that MCP and similar protocol standards are not developer curiosities. They are procurement criteria. Box launched its Box Agent in April 2026 explicitly positioning its content platform as the secure data layer for third-party agents, which is a direct response to this multi-agent reality: if your data is not agent-accessible, you are not in the workflow.
The most contrarian point in Levie's thread is the last one: enterprise leaders consistently underestimate the technical complexity of deploying agents in production. Skills, MCP configurations, CLI tooling, and orchestration logic are opaque to business teams. The implication is that engineering roles do not disappear in an agentic world; they shift toward setup, governance, and ongoing operations of agent infrastructure, which is work that cannot be automated away by the same agents it manages.
For investors underwriting the thesis that AI will compress enterprise headcount and therefore lift margins, this is a complication worth pricing in. The Fast Company analysis of Levie's March 2026 observations noted that enterprise leaders are focused on revenue expansion and new capability delivery rather than workforce reduction, which is consistent with Levie's April update that most agent use cases involve work that previously could not be prioritized at all.
What Levie's dispatches offer VCs is something most market research cannot: ground-level signal on where the friction actually sits. The AI agent market is projected to grow from $8.29 billion in 2025 to $12.06 billion in 2026, but the companies capturing durable margin inside that market are not necessarily the ones building the most capable models. They are the ones solving the procurement, budget, data access, and governance problems Levie describes.
Change management firms with AI deployment expertise, token cost management platforms, legacy system modernization tools designed specifically for agent integration, and headless API infrastructure for content and data: these are the boring-sounding categories that account for the gap between a compelling agent demo and a production deployment. Investors already in the model layer are well served to track where the organizational friction actually accumulates. Levie is telling them, at scale, in real time.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。