Access to Business Context Makes or Breaks AI Success 

Something interesting is happening in enterprise AI. When I say that, you might immediately think of the latest AI model or the newest chipsets. Those are important, but there’s something else that’s equally important. It’s just as hard to find, and it has just as much of an impact on AI’s success in production as models and hardware. 

That differentiator is context, specifically access to business context. Without it, AI models are good, but not quite good enough to succeed in complex, enterprise production environments. In this sense, context makes the difference between AI results and AI aspirations. 

All of this means that the race to access context is heating up. 

How do you get context? That’s a key question, and it’s one that I’ve been thinking deeply about for some time. 

The argument goes like this. 

If the question of AI is really one of context, the question of context is really all about access. Seen in this way, access becomes a central bottleneck in the race for context. Organizations that can access their business context easily will have a huge advantage in the future when it comes to AI success. Meanwhile, those whose business context is locked away in data silos will have a problem. 

This begs the question, why can’t businesses just access it easily, and what’s needed to get there? In practice, that might sound simple, but in actually existing enterprise environments, accessing contextual data is difficult. 

There are at least three reasons for this. 

1. Where is your business context?

First, in many enterprises, business context is hard to pin down. It isn’t found in one known, identifiable place. Instead, it’s spread out across multiple data sources, some of them using legacy technology, others operating in the cloud, others not really on the map at all. 

Making sense of this complexity presents a challenge because if you can’t identify business context, it’s very hard to leverage it. 

2. How is your business context formatted? 

The next problem is a data formatting problem. Context is messy, organic, and always changing. It is literally contextual, and that means the form it takes is often equally messy. 

In practice, this might mean that context is held in non-traditional formats that can be hard to parse, like PDFs, chat logs, slide decks, Slack messages, emails, call transcripts, and video demos. Context can be almost anything. That’s what makes it so valuable, but it’s also what makes it a challenge to access. 

3. How is your context governed? 

The next problem is governance. Because context is spread out and formatted in complex ways, its governance is equally complex. In practice, this means that AI agents might not always have the governed access to the context they need to perform the tasks they are required to perform. 

Complexities also beget more complexities. Solving this problem also gets exponentially more difficult the larger an organization’s data estate becomes. 

Getting universal access to business context

Any way you look at it, the AI business context problem is really a data access problem. And data access problems can be solved in one of two ways. 

The first is cumbersome and traditional, and the second is a more elegant solution that we see more enterprise customers adopting today. 

Let’s start with what you will probably be told you should do, but you absolutely shouldn’t. Data centralization. This is the old, one-size-fits-all standard solution from the data industry’s past.

Why doesn’t it work? Although you could centralize all of your business context, in reality, the effort required to succeed will be herculean in both scope and cost. In reality, these types of projects have always been unwieldy and prone to overruns of every type. 

With AI, this problem is even more pronounced. Business context is spread across your data estate for a reason. Centralizing it is neither practical nor cost-efficient. This is already the case with analytics, and it’s doubly so with AI. 

A more elegant solution comes in the form of data federation.

A more elegant solution comes in the form of data federation. I’ve been talking about the value of federation for years in the context of analytics, but when considering the needs of an AI context layer it makes even more sense. Federation allows organizations to access data wherever it lives, regardless of its location, whether on the cloud, on-premises, or a mix of both. 

For the problem of accessing the context needed for AI, data federation is a perfect approach. Business context exists across your data estate for a reason. Those spreadsheets, chat logs, and Slack messages can’t reasonably be centralized, and even if you could centralize some of them, new ones would be generated. With data federation, you don’t need to do any of this. You can just access the data where it lives. 

All of this means that data federation is the answer to every enterprise’s need for AI context. It provides the universal access needed to make AI work in production, and it fits the unique needs of business context from a data perspective.

As the AI revolution continues to heat up, I expect eyes to turn more closely to the context layer needed to make AI succeed, and to data federation in particular, as the best way to unite context and AI together.