TFiR

When Agents Write to Salesforce, Who Controls the Data | Mario Moscatiello, Airbyte

Duration: 4:06 830 transcript words EN summary

Video summary

The video explores the significant shift from unidirectional data retrieval to bidirectional interactions where AI agents are empowered to write back to sophisticated enterprise systems like Salesforce. This evolution moves beyond simple data activation or reverse ETL, which previously involved deterministic scripts updating CRMs based on scheduled reports, into a realm where autonomous agents can take real-time actions. While companies have long utilized tools to push data back into sales pipelines for tasks like upselling, the introduction of AI agents fundamentally changes the architecture by allowing non-deterministic reasoning and dynamic decision-making directly within these critical business systems. This transition introduces profound governance challenges because it expands access from a small, specialized data team comprising perhaps fifty individuals to potentially every employee in an organization having agent capabilities. The primary concern is ensuring that agents do not hallucinate or write incorrect data while simultaneously verifying that specific users have the necessary permissions to modify records they are interacting with. Unlike traditional workflows where code execution was predictable and limited to a few developers, agent-driven actions require robust identity management and access controls to prevent unauthorized modifications to sensitive customer information when the scale of potential actors increases dramatically. To address these complexities, the discussion highlights the critical need for enhanced data lineage and traceability mechanisms that can explain why an agent took a specific action. Organizations must be able to answer questions regarding who initiated the agent, whether it was running on a schedule or automatically, and exactly what reasoning led to a particular data write operation. Since agents operate through non-linear thought processes rather than fixed code paths, maintaining a clear audit trail becomes essential for compliance and trust, ensuring that every modification to a CRM record can be traced back to its root cause within the agent's decision-making flow. Finally, the conversation touches on the practical implementation of these capabilities across various cloud platforms and AI marketplaces, noting that while integration processes vary in speed among providers like OpenAI, Frontier Labs, and others, the core functionality is becoming widely available. Although having tools officially listed on a marketplace simplifies discovery and deployment for teams, the underlying technology allows for direct installation via configuration links even outside of these curated environments. As companies collaborate with different AI model providers to bring these resources to their respective marketplaces, the industry is collectively navigating the balance between unlocking powerful agent capabilities and maintaining the strict governance required for enterprise data integrity.

Read the full video transcript

Can you talk about you know, uh uh how how it just kind of pushes your bite beyond just retrieving data into actually now you're not just looking at data it's not just one way traffic. You're actually also taking actions like writing back to very sophisticated system like Salesforce. Uh how does that bi-directional access change the way enterprise AI agents are architected and what new governance challenges come with it because if you do allow your agents to start taking actions to start writing to your sophisticated system, you need to have right governance as well. So, it is not as that simple. So, can you talk about the working of working of uh the CLI? >> Um I think the CLI is is one component but um where we allow developers to just interact with our platform directly from the terminal. But, I think what we're talking about here is more like us enabling agent to write back to systems. If we look back at the last 5 years, this is already happened like uh when it comes to you know, just pure data. So, I'll back up a bit but essentially we also have this feature but there are a lot of companies that uh a bunch of companies that have developed this feature before which is called what we call reverse CTL or data activation. And what that means is that, you know, before this is before AI you could have a DBT model that runs every day and says like, "Hey, who were my top, you know, spenders yesterday?" And of course, what you want to do is have a system that every morning writes uh back into Salesforce so that the sales team can look at, you know, who were the biggest spenders and try to upsell them or trying to lock them into a contract as an example. Um and so this reverse CTL actually transforming data and writing it back to systems is not something that is just out today, something that companies have been doing for a while. I think the issue is that when you go from a data team writing very deterministic code that is going to run every day and just update, you know, stuff in a CRM or in whatever system, when you go from that to an agent writing back to a system, first of all, uh you need to make sure that the agent is not hallucinating and then the agent is writing the right data to the right system. And second, you go from a few people within the organization, even if it's a large organization, you can think of even a company of 10,000 people is going to have maybe 50 people, 100 people in the data team if it's already very data mature. When you go to from like 50 people to suddenly everybody's agent having access to writing back to systems, how do you know that you know, a specific person is even allowed to write back to the record or how do you know that a specific person has access to the information it needs to for the agent to take action. And so, from a government standpoint, like it requires way more robust, you know, governance and and so on and so forth, but you also need way better data lineage because you need to be able to understand like why did an agent take an action? You need traceability, you need to understand, yes, this is a non-deterministic, you know, flow where an agent is reasoning and taking action, but you always need to be able to trace that back into, okay, but why did the agent actually make the decision? Why did the agent wrote that data? Who initiated that agent? Was it on a schedule? Was it automatic? And so on and so forth. So, I think you're going from a few people in an organization owning and governing access to data to suddenly everybody being able to work on data. And so, the scale at which that happens is just massive and so it introduces a a lot of challenges. So. >> It's available on OpenAI marketplace. What about other like cloud and Tropic Cloud is there a thing? >> Yeah, we're working with all of them. Yeah, we're working with them. You know, it's it's a process like every company, OpenAI like was quicker in this case, but it's a process and we're working with all with all of the Frontier Labs, you know, to have these you know, these resources like officially on their marketplaces. You can still go ahead and install like an MCP into Cloud. It's it's just pasting a link. Of course, it's nice to be in the marketplace so that people can find it quicker and their team is also like, you know, doing some work on it, but we do work with with all the matter providers.