Your AI Should Be Able to Use Our Software

A few weeks ago, I wrote about the shift from an integration fabric to an intelligence fabric. The idea was that as AI becomes more capable, enterprise software has to expose more than APIs for other software to call. It needs to expose capabilities in a way that AI can discover, understand and use safely.

That is starting to look less like an architecture discussion and more like a product strategy question.

At Dreamforce, Salesforce replayed a clip of Parker Harris making a remarkable statement: "Why would you log in to Salesforce again? Maybe you never will?"

For the co-founder of one of the world's largest enterprise software companies, that is not a throwaway line. It reflects a bigger shift in how software companies need to think about the relationship between the product and its interface.

The interface still matters. Salesforce is not abandoning its UI, and neither are we at Exchange Solutions. But you no longer have to access a product's value exclusively through its own screens.

That distinction matters.

Every enterprise software company is adding AI right now. Your CRM has an assistant. Your analytics platform has one. Your service desk has one. Your productivity suite has one. Individually, many of those experiences are useful. Collectively, I am not convinced the future is every software vendor owning a separate AI relationship with the same employee.

Enterprise customers will make their own choices about AI. Some will standardize on a major enterprise platform. Others will build internal agents around their own data, workflows, and security model. Different teams will still use specialized tools where they make sense, but software vendors should not assume that because we own an application, we also get to own the AI interface.

That is becoming an important product question for us at Exchange Solutions.

We are investing heavily in AI inside our own loyalty platform because our domain context makes the experience materially better in some places. A marketer may want to understand why member behaviour is changing, identify an audience worth acting on, determine what kind of intervention makes sense, understand the economics behind that intervention and move toward execution.

Exchange Solutions already has deep context around the member and the loyalty program. We see transaction behaviour, points balances, member status, engagement history, scoring and the actions available through the program. When we can use that context to reduce complexity and help someone make a better decision, building intelligence directly into the product makes sense.

Our Member Scoring & Intelligence capability is a good example. Marketers shouldn't need to understand all the modelling and scoring machinery underneath the experience to ask a useful question and reach an actionable audience. In that situation, our domain understanding makes the AI experience better.

But there is another side to this.

A client may already have an enterprise AI that understands much more about its business than Exchange Solutions ever will. It may have access to customer service interactions, merchandising information, credit data, internal documents and other context that sits outside the loyalty platform.

That broader context becomes much more useful when it can work with what we know.

If a marketer asks their enterprise AI, "Which of our high-value members are showing signs of declining engagement, and what should we do about it?", they should not have to know which application has each part of the answer.

Their AI should be able to come to us.

That is why we are actively working on our first Model Context Protocol release at Exchange Solutions.

I wrote about MCP in the earlier intelligence-fabric piece, so I am less interested in explaining the protocol itself here. The connectivity is increasingly becoming infrastructure. The harder question for us is what Exchange Solutions should make available through that interface and when the client should still come into our experience.

That is a product decision.

For example, if a client's AI needs to understand a member's points balance, recent loyalty behaviour or program status, there is no obvious reason the user should have to open Exchange Solutions and navigate through a set of screens to retrieve that information. If they are already working inside a trusted AI environment, our platform should be able to provide it there.

The calculus changes when the work becomes more consequential. Building and reviewing an offer, understanding its economics, evaluating the audience and preparing an action that may affect hundreds of thousands of members may still be better inside Exchange Solutions, where we can provide the domain-specific context, controls and visibility that make the decision understandable.

That distinction is more useful than simply deciding whether something is "AI-enabled."

Where our domain context creates a materially better experience, we should build the intelligence directly into Exchange Solutions. Where the client's broader context matters more, and the capability can be safely exposed, their AI should be able to use us directly.

Sometimes both will be true.

The important part is that we should not force the customer into our AI simply because we built one.

That requires a change in how enterprise software companies think about product ownership. For a long time, owning the application effectively meant owning the interface into the application. We designed the screens, workflows and navigation because that was how users reached the capability.

AI weakens that assumption.

A customer may decide that the primary interface into its technology environment is an AI it controls. In that world, the application still matters. The data still matters. The business logic still matters. The domain knowledge still matters. But the interface into those capabilities may no longer belong entirely to the software vendor.

That is what makes the Salesforce comment interesting to me. "Why would you log in to Salesforce again? Maybe you never will?" is not about killing the interface. It is about separating the platform's value from the requirement to use it through its own UI.

We are thinking about Exchange Solutions in much the same way.

We will keep investing in our own user experience because workflows still exist where direct interaction, visualization, configuration, and review are the best way to work. We will keep building AI into the platform where our understanding of loyalty allows us to create something genuinely useful.

But Exchange Solutions also needs to work well when the customer chooses a different intelligence layer.

That is the shift.

Enterprise software needs to work with the AI the customer chooses, not just the AI the software vendor provides.

And if we really believe that, then the product cannot stop at the interface.