AI tools work best when they have access to relevant business data. Snowflake CEO Sridhar Ramaswamy argues that business applications risk becoming just “dumb data pipes” for AI’s “big brains.” Is this a legitimate concern?
There is no doubt that this fear exists. The dark red stock prices of Salesforce, HubSpot, ServiceNow, and many other software companies over the past few months suggest that investors are far from resting on their laurels. While all IT vendors are preparing stories around AI and can be agent management platforms, Ramaswamy believes this is not the approach of AI vendors.
According to Snowflake’s CEO, model makers want to create a world where all data is easily accessible. Therefore, the applications in which that data is used are essentially ballast. Why use an AI assistant built into your ERP or CRM platform when you can access the same data via central intelligence?
freedom of choice
Ramaswamy’s warning carries weight. As a central data platform, Snowflake is a very useful partner for AI players, as is Databricks. Snowflake recently signed a $200 million partnership with OpenAI. Therefore, models such as GPT-5.2 will be available directly within Snowflake Cortex AI.
In this way, Snowflake avoids becoming a “dumb data pipe” for LLM. But Ramaswamy claims his company fears that organizations will stop using AI agents built by software vendors. These specialized agents would certainly have added value, such as being more accurate, working more securely, and being easier to use. For experienced users of existing platforms, this is already the case. Solutions like NetSuite and Salesforce offer AI capabilities as extensions to familiar systems, so deploying these capabilities often happens without migration.
Ramaswamy believes that customers have the final say in this regard. According to the Snowflake CEO, if you want to consult central AI and ignore traditional enterprise apps, you should be given that option. The implicit message: Software vendors only need to include requested features or features that are so attractive that they pale in comparison to alternatives.
simplicity is incredibly complex
The defenses you get from software players are now familiar. Organizations are not greenfield operations, so you can’t simply replace your business suite with “AI.” Processes and policies are formed around the current IT stack, and the migration away from that assumed legacy will result in a world where not only data but also the day-to-day operations of the company need to be transferred.
However, the tug-of-war over the center of AI is in full swing. It is not without reason that vendors insist that their solution should be a central AI system, for example because it contains a huge amount of data or because it is the most important application for a particular department. So far, AI trends for these vendors have revolved around the adoption of AI chatbots, easy-to-set-up or off-the-shelf agent workflows, and automated document generation. At several IT events over the past year, attendees have floated the idea that old interfaces may disappear as all employees interact with data through AI.
Even when using AI, every detail has to be checked, so its simplicity is incredibly complex. You may need tabs to understand exactly what the data is showing. If this is just data in a “dumb data pipe”, it suggests that the data actually acts as a stream or water, a substance that needs to move from A to B (B in this case is the AI tool). But that data is not clear enough for AI systems to process in a single way. Screens within business applications do more than just represent data as truth. Humans also need to interpret that data, but only through an AI abstraction layer, which is already guided and focused on convenience. This simplification of data will negatively impact many people.
Also read: Oracle NetSuite introduces the missing piece of the puzzle for its platform
