13 hours Ago
Louis Landry , CTO at Teradata. AI is at a crossroads. Enterprises have spent the last several years exploring AI technologies, but most are still searching for revenue-generating opportunities.
This year, I expect business AI to take a major leap forward. Agentic AI helps make AI practical and commercially valuable by dedicating AI agents to complete tasks with minimal guidance. Agentic AI is gaining traction for automating research and business processes, enabling workers to hand off rote work and focus on higher-value activities.
Faster, more accurate results will simplify ROI calculations while unlocking business opportunities. But deploying agentic AI isn’t magic—businesses need to process data at a higher level to make agentic AI possible. That means CTOs must prioritize selecting and implementing a vector store, a key step in making agentic AI accurate and trustworthy.
A vector store is a structured database abstracting large quantities of data for easy retrieval, turning the 80% of unstructured data into vectors—numbers, categories and relationships AI can understand. Agentic AI depends upon vectors to obtain contextual guidance and quickly produce useful outputs. By searching vectors and their relationships, vector stores empower enterprises to extract deeper insights from their data.
Vector stores are the water of generative AI, agentic AI and systems that include retrieval-augmented generation (RAG), so building them now will give businesses the foundation they need for the data platforms of the future. They expand the utility and versatility of data in any environment, enabling searching and cross-referencing at massive scale, so users can draw upon resources from across an enterprise to quickly, deeply and accurately surface insights that previously might have been siloed (at best) or impossible to research. Not all vector stores are the same, however.
Some are limited by scalability, cost or adaptability to new technologies. With the exception of existing databases that have added vector store capabilities, many offerings are vector-only, which means they don't enable users to see any other context. These issues are important to understand up front, because the bigger the problem agentic AI will address, the more important the scalability of the vector store becomes.
Enterprises need to choose a platform capable of high scale and low latency, ideally with a unified structured/unstructured database. This database should be able to bind real enterprise data with trends, analytics and predictive insights, enabling a seamless integration of both unstructured context (like search and similarity analysis) and structured data (such as KPIs and metrics). This approach removes the need to separately query multiple data repositories, allowing for a more powerful, comprehensive solution.
Agentic AI is expected to help businesses solve long-standing problems while freeing teams to focus on higher-value work. For example, agentic AI can process virtually unlimited numbers of documents (including multipage PDFs and the most complex unstructured data) to identify themes and trends. Combining vectors with agentic AI, CTOs can offer users a platform that instantly loads and queries data, allowing businesses to quickly test AI agents and drive revenue.
Instead of wasting AI on abstract tasks, businesses can leverage agentic AI to answer new queries and accelerate sales. For instance, AI-augmented call center services can help human operators personalize customer interactions and decision-making. In insurance, agentic AI allows operators to analyze a customer’s profile and policies, offering tailored suggestions: “Here are several ways we can cost-effectively enhance your coverage to address your latest needs.
” Agentic AI can also run temporal queries, enabling insights into how data has evolved over time. For example, by analyzing vectorized satellite images, an environmental scientist can track changes in vegetation or detect potential natural disasters. Similarly, a doctor could analyze a series of medical images to monitor a patient’s condition, such as tumor reduction or recovery progress.
Vectors play a critical role in making AI trustworthy. As abstractions of data—the inputs for AI-producing large language models (LLMs) and RAG systems—vectors can be checked to ensure AI produces accurate results. They also improve explainability by allowing tracking and timestamping of changes.
Using millions of vectors enables AI agents to holistically consider all or most of a business’ data, helping ensure agents optimize for both the task and the business. I believe the winning solution for businesses will be an enterprise-class vector store—one that's trustworthy, scalable, cost-efficient and integrates with existing databases. It should be robust enough to support temporal data and queries, which will soon become much more common than today, enabling individuals to compare historical and current data in searches.
By fusing vectors with unstructured data, an enterprise vector store can optimize the performance of that data. AI will research answers using rich context to understand input, then deliver more accurate and effective results, blending text responses with photos, audio and video content. The richer the context represented by vectors, the more accurate and effective the results will be.
An enterprise-class vector store should offer several benefits. Beyond connecting structured and unstructured data together in one place, it should be able to scale multidimensionally and support high concurrency across vast numbers of vectors. It should also automate processes, enabling businesses to use agentic AI quickly without managing complex setups.
For many businesses, there's no question that agentic AI will be a game changer, unlocking the value of data that's been underutilized for years. But without vector stores—specifically those that integrate with existing databases—enterprises will struggle to understand their data, connect AI initiatives to broader goals and use agentic AI to grow revenues. I believe that choosing a scalable, fast and easy-to-use vector store will be one of the most impactful AI investments any CTO makes this year.
To achieve enterprise-class performance, I recommend opting for integrated solutions rather than standalone vector stores. A well-designed vector store will not only connect and vectorize all business data but also simplify the automation of key business processes, driving immediate ROI while supporting the company’s long-term mission. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives.
Do I qualify?.
