Why Data is More Valuable than Code
In “Data Rules Everything Around Me,” Matt Slotnick wrote about the difference between SaaS & AI apps. A typical SaaS app has a workflow layer, a middleware/connectivity layer, & a data layer/database. So does an AI app.
AI makes writing frontends trivial, so in the three-layer cake of workflow software the data matters much more.
The big differences between an AI & the SaaS app lie within the ganache of the middle layer. In SaaS applications, coded business rules determine each step a lead follows from creation to close.
In AI apps, a non-deterministic AI model decides the steps using context : relevant information about the lead that the AI is querying from other sources.
The better the data, the better the workflow.
The context is the most valuable component because it ultimately changes the workflow. Models are relatively similar in performance.
For example, an inbound email comes into a customer support desk, “Was I double charged this month?” An agentic workflow would query the billing system, the contract system, & the email drafting tool to draft an email to the customer with distinct language for that persona. This only works if the enterprises’ data is well structured.
Enterprises will be shy about sharing the context with their vendors because of how much value it provides. They may start to structure it & assign a department to manage it because the better its availability, the more effective the agentic systems will be.
Data architecture may become a competitive advantage & the future battleground for software companies will be the access to that context - & the fight has already begun.
Pragmatic Innovator. Currently focused on Distributed Data Management. Control, Security, Rights Enforcement on all data on all devices.
1w"Context is the most valuable component ..." and yet our approach to data management strips the context out from the data and separates it. What is data was stored in a "self-aware" state such that each record was stored, with its metadata, as a fully encapsulated, directly findable data object? I understand that this thinking reverses 40 years of practice but, as an organization, how much do you save by eliminating ETL, MDM, point-to-point integrations etc? I suggest those savings would be material for every organization!
VP, Data & Business Operations | Shaping Data & AI-Enabled GTM for Growth & Retention
3wLove this, Tomasz!
Charting a True North in the AI Economy
1moTomasz Tunguz - would be great to get your take on how AI powered Browsers (Dia, Comet, pending from OpenAI) play into the 'fight for context'.
Data & Analytics, AI/ML, AWS, Architecture & Engineering, DataIku, AWS Sagemaker, DataOps, and MLOps Leader
1moGreat Insights. Fragmented data ecosystems remain a major barrier: data duplication, silos, lack of lineage, poor data dictionaries, and inefficient MDM all contribute to widespread data quality issues. Quality data is king. Yet, many companies are still struggling to clear the first hurdle: transforming their data ecosystem to address these challenges and make their data AI-ready.
Founder and CEO at Reef.ai
1moTomasz Tunguz - we bet our business on this being true. Reef.ai is a data product (no frontend). Vertical data depth (NRR in the case of Reef will underpin the best agents in this domain). Reef created the NRR Intelligence Graph that provides clean, validated longitudinal customer data layered with AI-ready predictions and metadata designed to enable agents to reduce churn, increase consumption, cross-sell, etc. The best frontend agents and apps for NRR use cases will operate on Reef data.