Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Will CRMs exist in five years?

CRM's assumption—that a person types in what matters—is already false.

[Photo: Getty Images]

Profile photo of Christopher O'Donnell

In 2014, I sat with a small team of engineers and designed the fields for what became HubSpot’s Sales Hub, a customer relationship management (CRM) tool that grew into one of the systems most business-to-business sales teams use every day to track their deals.

We spent weeks arguing about dropdown menus. What should a deal stage be called? Should there be five stages or seven? What happens when a sales representative forgets to update a field for three weeks straight?

We built something genuinely useful. It solved a real problem for a decade of business software, and I’m proud of it. I also now believe it, and every CRM built like it, will not exist in anything like its current form five years from now.

Why it won’t exist

From day one, we built an assumption into the CRM that was completely reasonable in 2014. It was that a human would decide and type in what mattered in a customer relationship.

This is simply no longer true today.

A useful comparison is what already happened in engineering. AI writes code remarkably well now because it relies on raw material from software engineers: An organized and structured codebase, written and read by every developer from the same source of truth. When an AI tool looks at a codebase, it retrieves something real, then acts on it.

Sales and customer-facing teams never had the equivalent. A CRM was a record of what somebody had time to type up after what happened, if they remembered and thought the effort was worth it. I spent a decade watching the fill rate on well-maintained sales pipelines, and the number was paltry. Writing up a call has always been the lowest-value use of a salesperson’s afternoon, so it gets done last, done badly, or not done at all.

Over the past two years, that gap has become so obvious that most CRMs have added some kind of AI feature on top. Paste in a transcript, get a follow-up email drafted, ask a chatbot to summarize a deal.

While useful in the moment, they share the same flaw: The AI can only reference and carry context for data already in the CRM, whether a contact’s record or case history. But the systems rely entirely on humans entering data and can’t retrieve or accumulate anything outside their own systems. Next week it starts over just as blind as it was the week before, because the system underneath it still isn’t built to hold what happened.

The next problem phase

Retrieval, speed, and action matter now. Can a system find everything true about a customer relationship instantly, move fast enough to matter, and do the next right thing without a person assembling the picture by hand first?

Optimists assume someone will bolt a better AI layer onto the CRM everyone already has to solve the problem. Pessimists assume the whole idea of systematically tracking customers is on its way out. I think both are wrong, for the same reason.

Look at what happens to any software category whenever underlying assumptions stop being true. For example, if you remember early marketing suite software, companies ran five, ten, fifteen tools and stitched them together with another application and spreadsheets. Nobody wanted that.

Eventually somebody has to put the pieces back together into one system a business can actually run on, because nobody wants five logins to understand a single customer. In the past, this happened through new builds or acquisitions stitched together.

I see it already in how serious operators behave. I talk to revenue leaders who built their own notes and scripts folders that do a better job pulling customer context together than the CRM they’re paying for. They got tired of being the ones who had to remember what happened, and started teaching a piece of software to remember it for them.

That’s a person doing exactly what a rebuilt category should be doing automatically, and a sign that the CRM category is coming apart.

A better solution

I’ve spent the past two years on the second version of what I built first. I started Day AI because I wanted to build on the assumption that’s actually true now. Most of what matters about a customer relationship already exists somewhere, in a call, an email, or a message. The challenging job is to find that data, hold onto it, and act on it—wherever it lives in your messaging, mail, calendar or recordings—not wait for someone to type it up later.

The point is bigger than any one company, including mine. The CRM open in your browser right now will probably continue working for a while yet. But the assumption underneath it is already false, and every quarter that passes widens the gap between what the system holds and what actually happened.

While I don’t know exactly what replaces CRM as a category, I do know its replacement won’t be what we’ve spent 20 years calling CRM. And I’d rather be the one saying so now than the one still defending the old assumption after everyone else can already see it’s gone.

Christopher O’Donnell is the founder and CEO of Day AI.