Skip to content

Debates

Must an agent-led software company own the tools, workflow and data itself, or can it access them through a core provider?

6 recorded positions from 4 people, first said Jun 24, 2024. They do not agree — the readings below are what each one actually argued.

Must own tools workflow and data full stack

David Luan · Jun 24, 2024

In the agent space it is extremely important to own the entire vertical stack, from end-user interface down to the foundation modeling layer

Owning the interface gives tremendous leverage in this era of AI, and making agents reliable enough for work requires tuning the whole end-to-end system, including the model layer

Scope: specific to agents, not model-selling businesses

32:09 20VC: Why Foundation Model Performance is Not Diminishing But Models Are Commoditising, Why Nvidia Will Enter the Model Space and Models Will Enter the Chip Space & The Right Business Model for AI Software with David Luan, Co-Founder @ Adept

David Luan · Jun 24, 2024

The only way to handle the variability and edge cases of enterprise workflows is vertical integration of the model with the use case, which is why Adept will do better than companies focused on a single narrow vertical problem

Generalizing across all the edge cases of how different employees and companies do a task requires controlling the model, not just the application

33:05 20VC: Why Foundation Model Performance is Not Diminishing But Models Are Commoditising, Why Nvidia Will Enter the Model Space and Models Will Enter the Chip Space & The Right Business Model for AI Software with David Luan, Co-Founder @ Adept

Vlad Tenev · Jul 14, 2025

Robinhood's in-house customer support AI is best in class, stronger even than dedicated customer support AI companies

7:09 20VC: Vlad Tenev on Robinhood's $85BN Resurgence | Tokenizing SpaceX & OpenAI | Building Nine Revenue Lines Over $100M | Why Crypto Will Be Robinhood's Biggest Revenue Line | Why Stablecoins Will Replace Banking Rails

Vlad Tenev · Jul 14, 2025

For a company like Robinhood the hard part of support AI is system integration, and most vendors haven't figured out how to do deep integration with back-end systems to take actions

Answering queries and taking actions requires deep integration with proprietary back-end data and systems

Scope: 'most vendors'

7:24 20VC: Vlad Tenev on Robinhood's $85BN Resurgence | Tokenizing SpaceX & OpenAI | Building Nine Revenue Lines Over $100M | Why Crypto Will Be Robinhood's Biggest Revenue Line | Why Stablecoins Will Replace Banking Rails

Harry Stebbings · Feb 9, 2026

To win in an agentic world you have to own the tools, the workflow and the data — the full stack

Podium built a $100M+ agent-led business off a traditional SaaS base by owning the full stack, as Salesforce does

57:27 20VC: Is SaaS Dead in a World of AI | Do Margins Matter Anymore | Is Triple, Triple, Double, Double Dead Today? | Who Wins the Dev Market: Cursor or Claude Code | Why We Are Not in an AI Bubble with Anish Acharya @ a16z

Also on the record

Anish Acharya · Feb 9, 2026

You don't have to own the full stack — you can access a core provider via tool use; the real determinant of agent leverage is the degree of ambiguity in the work

BPO-style jobs are bundles of well-defined tasks pulled off a queue and are well set up for agent replacement, whereas high-ambiguity work resists it

57:58 Tool use access suffices ambiguity level decides leverage

Your assistant can query this graph directly — 6 positions here, 19,646 across the corpus. Add 996.fm over MCP.