Field Note
Cable, Streaming, and the AI Remote
How enterprise software became a collection of specialized tools, and why AI might bring it back together.
How enterprise software became a collection of specialized tools, and why AI might bring it back together.
For a long time, I assumed a lot of the complexity in our enterprise software was specific to our organization. Temporary solutions that became permanent. Systems added during different phases of the business. Knowledge that disappeared as people and teams changed.
And I’m sure some of it is.
But this week, I was talking with another product leader, and we were dealing with remarkably similar problems. That conversation got me thinking about how companies buy and manage digital tools, and I kept coming back to cable TV.
We traded cable for streaming
For years, enterprise software worked a lot like cable. You bought the package. One vendor provided commerce, content, search, analytics, and sometimes hosting. Some parts worked well and others didn’t, but they came together, and replacing one usually meant dealing with the rest. When something fell behind, you found a workaround or waited for the vendor.
Now we buy software the way we stream TV. Commerce from one provider, search from another, content from yet another, each selected because it does one job well.
Our own B2B commerce platform is going through that transition now, and it’s given me a firsthand look at both the flexibility and what comes with it.
I understand the appeal. B2B commerce is complicated, and every business handles it differently. Customer-specific pricing, contract terms, purchasing permissions, inventory availability, and fulfillment are often fundamental to how a business operates. Fitting all of that into a single packaged solution can mean years of customization and compromise. Choosing the right tool for each job makes sense.
But there’s a catch.
We became the cable company
Think about how many streaming subscriptions you have. We dropped cable for more choice and control. Now we have five subscriptions and five apps, and we still can’t find the show we’re looking for. We’ve become our own cable company.
Enterprise software is doing something similar. We gained the freedom to choose better tools, and we took on the job of making them work together.
Each system can work exactly as designed while the experience they create together is wrong. A product might be measured by the carton in one system and by the piece in another. A sales rep might log in on a customer’s behalf, and the analytics platform credits that activity to the customer. Neither platform is broken. The information is still wrong, and someone has to figure out why.
Some of these problems aren’t new. They’ve been around for years, buried under custom integrations, workarounds, and business processes that outlasted the people who designed them.
So how much of this complexity comes from decisions made over time, and how much is just part of running a modern enterprise platform? I suspect it’s both.
Everyone’s building AI into their piece
Almost every software vendor is building AI into its products. Search platforms use it for relevance and recommendations. Content platforms use it to create and organize information. Analytics platforms let you ask questions instead of building reports.
That has real value. But each tool works with the information it can see.
Your search platform might know which products are related, but not whether a particular customer is allowed to buy them. Your analytics platform might spot a buying pattern without knowing the account restrictions or inventory limits behind it.
I’ve been working through a version of this with product recommendations for our platform. Recommending related or alternative products sounds simple. Then you ask what happens when the product isn’t available to that customer, isn’t priced for their account, or can’t fill the quantity they’re ordering.
You’ve recommended something the customer can’t actually buy. Do that enough times and they stop trusting the recommendations.
It doesn’t matter how intelligent the recommendation is if it doesn’t account for the rules behind the transaction. And those rules often live somewhere else entirely.
That’s my concern with adding AI to each tool on its own. We could make every part of the platform smarter without making the overall experience any better.
The AI remote
There’s another direction this could go, and it’s the one I’m most curious about.
Streaming broke the cable package apart so we could choose. Now the services are bundling again, and we want to search across all those catalogs without opening every app. We wanted the flexibility of streaming without the headache of managing it.
AI could play that role for enterprise software, giving us one way to work with the applications we already have.
For someone running a digital platform, that might mean asking why an order failed without tracing it through commerce, ERP, inventory, and reporting one system at a time. For a customer, it might mean asking for an alternative product that’s in stock, priced for their account, and able to fill the quantity they need.
Either answer requires working across several systems. That’s a big step beyond a chatbot that only knows its own platform.
It also brings us back to the original problem. An AI assistant can’t make two conflicting sources agree. It doesn’t automatically understand years of business rules, customizations, and exceptions. If it can’t reach the right information, or doesn’t know which system to trust, the answer can still be wrong. It may just sound more confident.
That worries me more than an obvious integration failure. When a system fails visibly, someone knows to investigate. When an assistant gives an answer that looks reasonable, how quickly would we notice that something underneath it was wrong?
What I’m still figuring out
This has reinforced how far product ownership reaches beyond what a customer sees on the screen. We talk a lot about features, experiences, and roadmaps. The real experience depends on decisions and data spread across several systems, often owned by different teams.
Someone has to understand where that information comes from, how it moves, and what happens when two systems disagree.
A product manager won’t own every integration, but we need to understand the business logic well enough to notice when the pieces aren’t working together. Otherwise we ship features that technically work and still don’t solve the customer’s problem.
And that responsibility doesn’t disappear with AI.
AI might become the remote that connects all these different systems, but someone still has to push the buttons. Someone has to provide the context, question the answers, and understand the business well enough to recognize when something doesn’t make sense. I think that judgment matters more as AI takes on more of the work.
I don’t think moving away from packaged platforms was a mistake. The flexibility is valuable and often necessary. And not every problem comes from that shift.
What I find interesting is how familiar these conversations with other product leaders are starting to sound. Different companies, platforms, and industries, with similar trouble in the systems, data, and rules that hold everything together.
So I’m curious. Are we all dealing with different versions of the same problem? And as AI becomes part of more of these systems, are we actually solving that complexity, or just finding new ways to work around it?
Get notified when I publish new research