Ash Ashutosh: The shift to agent-centric systems is revealing flaws in data retrieval, AI agents struggle with task completion rates below 50%, and vector databases are key to enhancing knowledge engines | AI + a16z

Vector databases are revolutionizing AI efficiency by boosting data retrieval accuracy from 68% to over 90%.

Listen on a16z Podcast

Share

Add us on Google by Editorial Team May. 5, 2026

Key takeaways

• The transition from human-centric to agent-centric systems is reshaping data retrieval structures.
• Data retrieval accuracy has improved from 68% to over 90% with new systems.
• AI agents face inefficiencies due to their different operational methods compared to humans.
• Agents often perform query expansion to gather information, lacking context.
• Task completion rates for AI agents are below 50%, indicating inefficiencies.
• 85% of AI agents’ efforts are spent on retrieving knowledge rather than processing it.
• Vector databases act like libraries, providing information for knowledge engines.
• Agents can process queries much faster than humans, leading to high token consumption.
• Moving reasoning closer to the data source can enhance data handling effectiveness.
• Current data systems often perform reasoning at the retrieval level, which can be inaccurate.
• The shift to agent-centric systems reveals flaws in traditional data retrieval methods.
• Improved data retrieval accuracy is crucial for the effectiveness of AI applications.
• The inefficiencies in AI agents’ task completion highlight the need for better systems.

Guest intro

Ash Ashutosh is CEO of Pinecone, a vector database platform powering retrieval-augmented generation and agentic workflows at scale. Previously, he founded and led Actifio, where he established the principle that every company is fundamentally a data-driven company. He specializes in building infrastructure that enables enterprises to manage and leverage data as a competitive advantage in the age of AI agents.

The shift to agent-centric data systems

• For years systems like databases and search were designed around human interaction… but with the rise of agents that model starts to break down.

  — Ash Ashutosh

• The transition to agent-centric systems is challenging traditional data retrieval structures.
• AI agents operate differently from humans, leading to new inefficiencies.
• The shift reveals fundamental flaws in how data retrieval is structured.
• Improved data retrieval accuracy is essential for agent-centric systems.

• The accuracy dramatically goes up from I think best case was about 68 well over 90% accuracy and that is just version one.

  — Ash Ashutosh

• Understanding the transition is crucial for optimizing AI applications.
• The new systems enhance data retrieval accuracy, impacting AI effectiveness.

Inefficiencies in AI agent operations

• Unfortunately agents don’t have the luxury they the human gives them a task the agents go there and start trying to perform the task right and they spend a ton of time going through this brute force loop of querying.

  — Ash Ashutosh

• AI agents face operational challenges due to different querying methods.
• Task completion rates for agents are often below 50%.

• Most of the time turns out the task completion rate is less than 50% so half the task return these these agents don’t actually complete.

  — Ash Ashutosh

• Inefficiencies highlight the need for improved systems.
• Agents perform query expansion without context, impacting efficiency.

• Agent does something called a query expansion breaks up the queries and then says okay let me go figure out what this product is and it goes to five or six different systems.

  — Ash Ashutosh

• Understanding these inefficiencies is crucial for system optimization.

The role of vector databases in knowledge engines

• Think of a vector database like a like a library yeah there’s tons of information out there a human being asks for some information appropriate books and pages and documents are given to you and you read through this stuff and you figure out the knowledge out of it.

  — Ash Ashutosh

• Vector databases provide relevant information for knowledge synthesis.
• These databases differ from traditional systems by optimizing for knowledge engines.
• They act as libraries, offering information for AI applications.
• Understanding their role is crucial for enhancing AI efficiency.
• Vector databases support the synthesis and generation of answers.
• They are integral to the functionality of modern knowledge engines.
• The analogy of a library helps clarify their function in AI systems.

Efficiency and token consumption in AI systems

• Agents are very very good at reasoning they can spin up more queries in a millisecond than you can do in an entire day right right and so they brute force their way through which is why you see a ton of token consumed for even the smallest of the i see application.

  — Ash Ashutosh

• AI agents process queries faster than humans, consuming many tokens.
• This efficiency in processing highlights the operational dynamics of AI systems.
• High token consumption is a byproduct of agents’ rapid query processing.
• Understanding token consumption is crucial for optimizing AI systems.
• The efficiency of agents is a key factor in their operational success.
• Agents’ ability to process information quickly is a significant advantage.
• The rapid processing capability of agents underscores their effectiveness.

Enhancing data handling through reasoning

• When you move the reasoning closer to where the data is, closer to where the curation of the data, where the actual processing of the data is happening, you can do a lot more things.

  — Ash Ashutosh

• Positioning reasoning closer to the data source can improve data handling.
• This approach enhances the effectiveness of data utilization.
• Current systems often perform reasoning at the retrieval level, leading to inaccuracies.

• Fundamentally today all of the reasoning is done at the retrieval level… may or may not be the right answer because I don’t even know if you have all the data.

  — Ash Ashutosh

• Improved reasoning mechanisms are needed for accurate data processing.
• Understanding the distinction between retrieval and curation is crucial.
• Enhancing reasoning processes can lead to more effective data systems.

Addressing inefficiencies in AI task completion

• The task completion rate for AI agents is often less than 50%.

• Most of the time turns out the task completion rate is less than 50% so half the task return these these agents don’t actually complete.

  — Ash Ashutosh

• Inefficiencies in task completion highlight the need for better systems.
• Understanding these inefficiencies is crucial for optimizing AI applications.
• The low completion rate indicates significant operational challenges.
• Addressing these challenges is essential for improving AI effectiveness.
• The inefficiencies underscore the need for enhanced data retrieval systems.
• Optimizing task completion rates is a priority for AI system development.

The importance of data retrieval accuracy

• Data retrieval accuracy has improved from 68% to over 90%.

• The accuracy dramatically goes up from I think best case was about 68 well over 90% accuracy and that is just version one.

  — Ash Ashutosh

• Improved accuracy is crucial for the effectiveness of AI applications.
• Understanding the implications of improved accuracy is essential.
• The enhancement in accuracy impacts the efficiency of agent-centric systems.
• Accurate data retrieval is a key factor in AI system success.
• The improvement underscores the effectiveness of new systems.
• Enhancing accuracy is a priority for optimizing AI applications.

The operational dynamics of AI agents

• AI agents operate differently from humans, leading to unique challenges.

• Unfortunately agents don’t have the luxury they the human gives them a task the agents go there and start trying to perform the task right and they spend a ton of time going through this brute force loop of querying.

  — Ash Ashutosh

• Understanding these dynamics is crucial for optimizing AI systems.
• Agents perform query expansion without context, impacting efficiency.

• Agent does something called a query expansion breaks up the queries and then says okay let me go figure out what this product is and it goes to five or six different systems.

  — Ash Ashutosh

• The operational dynamics highlight the need for improved systems.
• Addressing these challenges is essential for enhancing AI effectiveness.
• The dynamics of agent operations are a key focus for system optimization.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.