Is Software Losing Its Head?
The podcast discusses how enterprise software is fundamentally shifting from being built around human interfaces to becoming 'headless'—optimized for AI agents to access data and logic directly. While incumbents like Salesforce are rebranding existing APIs as agent-ready, the real opportunity lies in new companies building between legacy vendors, leveraging AI to bridge organizational silos and capture the exceptions and context that make enterprise software truly sticky.
Summary
The episode explores the concept of 'headless software'—enterprise applications designed for AI agents rather than human users. Seema Amble explains that traditional software was built around human workflows and UI interaction, but in an agentic world, the user interface becomes irrelevant; what matters is the underlying data, logic, and business rules.
The speakers discuss what has historically made enterprise software sticky: human adoption, muscle memory, workflow integration across departments, the need for a single source of truth, and compliance requirements. Salesforce's recent 'Headless 360' announcement exemplifies this shift, though the speakers debate whether it represents genuine change or merely rebranding of existing APIs. Steven Sinosky argues that in enterprise software, network effects exist primarily within organizations, and adoption often happens through power users who discover productivity gains and create informal viral loops.
A critical theme is the difficulty of replacing systems like SAP or Salesforce. These platforms aren't just databases—they encode decades of customized business logic, exceptions, and processes specific to each customer. Attempting to replace SAP with 'Postgres plus APIs' fundamentally misunderstands enterprise software's value. The speakers illustrate this with examples: automobile manufacturers' operations are literally defined by their SAP implementations; insurance companies run on 50-year-old systems because the cost and risk of replacement is prohibitive.
The discussion shifts to what AI agents actually need from enterprise systems. Beyond data access, agents require 'context graphs'—the undocumented rules, exceptions, and edge cases that exist in employees' heads and company culture. A salesperson's handling of accounts differs by region, industry, and individual preference; these aren't captured in CRM fields but are essential for agents to operate effectively. The speakers emphasize that almost everything interesting in enterprise is exception handling, and capturing these exceptions through observation (voice recordings, email analysis, interaction logs) represents a new opportunity.
The speakers explore the productivity paradox: automating one process doesn't reduce total work—it reveals and creates new challenges. Amazon's customer service automation (accepting wrong shipments without return requirements) didn't eliminate customer service work; it shifted focus to preventing errors in warehousing and logistics. Similarly, if expense reporting is automated, new analysis layers emerge around business travel optimization. Legal contracts won't get shorter with AI assistance; they'll become more sophisticated and comprehensive.
Regarding startup opportunities, the consensus is that directly competing with incumbents is futile. Instead, the biggest opportunities exist 'between' established vendors—addressing handoffs between departments (sales and marketing, finance and operations) that legacy software doesn't optimize. New companies can also build by observing workflows (through voice agents, computer-use agents, document ingestion) to systematically capture context that incumbents won't disrupt their existing products to handle. The speakers note that network effects in enterprise software historically meant adoption within a company; chat interfaces are now creating internal viral loops as power users discover productivity gains and share them.
About this episode
Seema Amble, Steven Sinofsky, and Elena Burger unpack one of the biggest questions facing enterprise software: what happens when AI agents become the primary users of software instead of humans? The conversation explores the rise of "headless" software, why APIs and agentic workflows are reshaping enterprise applications, and whether traditional SaaS products are becoming systems of record rather than systems of engagement. They discuss Salesforce's Headless 360 announcement, MCP, enterprise software architecture, and why AI may fundamentally change how businesses interact with their data. Along the way, they examine what actually makes enterprise software sticky, why replacing systems like SAP and Salesforce is harder than it appears, and where startups have the greatest opportunity as AI reshapes the software stack.
Key Insights
- Salesforce's 'Headless 360' announcement is primarily a marketing rebrand of APIs that already existed; no fundamental technical change occurred, but it signals an industry acknowledgment that agents will access software through APIs rather than UIs.
- Enterprise software stickiness derives not just from UIs but from codified business logic, compliance requirements, downstream dependencies across departments, and the need for a single source of truth—all of which create organizational inertia.
- SAP and similar enterprise systems encode decades of company-specific customizations and business rules; replacing them is impossible not because of technical limitations but because the systems literally define how the business operates.
- Larry Ellison's failed advocacy for an '80% solution' demonstrates that enterprises will not accept off-the-shelf software if it doesn't match their specific operational needs—customization is not optional but essential.
- Agents fundamentally require 'context graphs' beyond data access: undocumented rules, regional variations, exception handling patterns, and permissions that exist in employees' heads and company culture, which incumbents rarely expose.
- Almost all interesting enterprise work involves exception handling—edge cases, special circumstances, and deviations from standard processes—making exception capture the actual competitive battleground, not routine automation.
- Automating a mundane process doesn't reduce total work; it reveals and enables new layers of analysis and sophistication, creating a growing pie of work rather than eliminating work.
- Incumbents will not risk disrupting their existing product lines and revenue streams to build genuinely new agent-first architectures; they will instead bolt AI onto legacy products, creating an opportunity for insurgent companies.
- The highest-value startup opportunity in enterprise software exists 'between' two established vendors—addressing handoffs and communications that legacy software doesn't optimize—rather than head-to-head competition.
- Network effects in enterprise software operate primarily within organizations, not between them, as demonstrated by chat interfaces creating informal viral loops when power users discover productivity gains and share them with colleagues.
- Observing and recording how humans actually work (voice calls, email interactions, document handling) provides a new systematic way to capture business context that AI systems need, creating a data exhaust from operational improvement.
- Permission and access control in multi-agent systems, and determining when multiple agents can write to a system of record simultaneously, represents a new class of unsolved problems as enterprises transition to agentic workflows.
Topics
Transcript
There are many things that made software sticky, but a lot of it had to do with the fact that it was built around the way a human interacts. In an agentic world, do you actually need that? The data, the logic, everything stored below it is really where the value is. There's this wild underestimation about you could vibe code your way into enterprise software. Larry Ellison at Oracle, he went on a rant about how enterprise software was so stupid because everybody customized it. The minute you automate the most mundane thing and think you have it all squared away, whole new things appear. Misconception right now is that you can just have Postgres database and APIs and…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The a16z Show
The Top 100 Consumer AI Apps: Who’s Actually Paying?
A16Z's seventh edition Top 100 Consumer AI Apps Report reveals that while half of Americans use AI, only 4.5% pay for subscriptions, with spending heavily concentrated among developers and power users. The report highlights emerging personal agents as a major trend, significant divergence between ChatGPT, Claude, and Gemini, and substantial untapped opportunities in categories like dating, social, shopping, and entertainment.
The $1 Trillion AI Buildout | State of Markets
A16Z partners discuss the $1 trillion AI infrastructure buildout, arguing it's driven by real earnings growth rather than inflated valuations, with adoption still extremely early at the enterprise level. They highlight opportunities across consumer agents, robotics, autonomous vehicles, and enterprise diffusion, while noting that the market's 90% gain since ChatGPT reflects fundamental business performance rather than speculative excess.
The Personal Agent Race Is Here | Anish Acharya & David Pawlan
A16Z's Anish Acharya and Assistant Benchmark creator David Pawlan discuss the explosive growth of personal AI agents, exploring their capabilities across email, travel, and finance use cases, the infrastructure needed for agent-to-agent interactions, and how these systems will reshape commerce and consumer behavior.
AI Can Write Code. Why Isn’t Software Better?
Diogo Almeida, founder of TypeSafe AI, discusses Jev, a new primitive that embeds AI decision-making directly into software rather than automating software engineering itself. The conversation explores why current AI tools haven't delivered on automation promises, and how probabilistic programming could fundamentally change how software is built and what it can do.
Building a Team at AI Speed | Harvey’s Maggie Landers
Maggie Landers, VP of Talent at Harvey, discusses how the legal AI company scaled from 340 to over 1,000 employees in one year while maintaining startup culture and values. She explains Harvey's approach to rapid hiring, emphasis on progress over perfection, founder leadership, and the specific traits they seek in candidates.