You may have seen the old graphic describing what RevOps does. Marketing, sales, CS, support, sometimes product, each one a column doing its own thing with its own data and its own definitions. Then a big bar running across the bottom labeled RevOps, tying it together so the customer doesn’t feel like they’re dealing with five different companies inside of one.
That picture is out of date now, and not in a good way.
What happened
Companies tell their team some version of “go use AI.” Which is the right call. Individuals moving faster is good for everyone and I’d tell them the same thing.
But watch what it does. One person builds a n8n flow for their own lead research. Somebody in CS vibe codes a little dashboard with its own database. A rep has 40 chat threads full of positioning language nobody else has seen. Someone in marketing wrote a set of AI instructions that finally sound like your brand, and they live exclusively within the instructions doc of their personal ChatGPT account.
Multiply that by 50 or 100 people. You’ve got the same silo problem the graphic was about, except now it’s at the individual level instead of across a few teams.
Taskford has a definition of this I keep coming back to:
AI sprawl happens when AI tools and agents spread across teams without clear ownership, shared rules, or proper integration.
Source: AI Sprawl: How Too Many AI Tools Undermine Control and Delivery
Ownership is the missing piece. Most companies I walk into have some kind of AI policy doc floating around. Almost none of them have named a person who owns how things fit together.
The context split flipped too. It used to be that maybe 80% of your company knowledge lived in shared docs and 20% lived on personal machines. I’d argue that’s reversed. The richest strategy and process work happening at most companies right now is sitting in personal chat histories and one-off tools that aren’t accessible at large.
Why this one is worse than the old version
The team-level silos were at least stable. One team changes a process or workflow a few times a year at best.
Individual silos move daily. Somebody rebuilds their workflow on Tuesday because a new model came out. The person who built the CS dashboard leaves and the dashboard breaks and nobody knows what it queried. Two people automate the same handoff in different directions, and now your data for a critical KPI has two different formats with two different definitions.
The thing that used to cost you a bad quarter of reporting and missing targets (still solvable by a RevOps function) now costs you a bad week, over and over.
Somebody has to tie it back together
That somebody is still RevOps. Same function as before, just at higher volume and smaller unit size. What used to be a big project for five or six teams every week or two is now dozens of small changes for dozens of individuals, multiple times a day.
Which means the way we work has to change, because the old playbook can’t move that fast.
A few things that have actually helped at the companies I work with:
- Cut the big playbooks into micro-playbooks. Nobody’s reading a 40-page process doc before they build an automation. They’ll read the one page that says what the deal stages mean and where the data goes.
- Put the company context somewhere shared and make it the default input. Brand voice, policies, ICP definitions, stage definitions, what we say about pricing. If people have to write that themselves, they will, differently, 50 times.
- Say out loud what individuals are free to build and what has to come through ops. Anything that writes to the CRM or touches a customer, that’s ops. Personal research and drafting, go for it, but log information useful to the team somewhere accessible.
- Have somewhere for the good stuff to land. Somebody’s private workflow that saves them four hours a week should end up in front of the other nine people with that job. Encourage sharing, don’t be too critical of attempts at AI automation that don’t work perfectly.
The part that’s easy to miss
None of this is a reason to slow the team down. The companies that clamped down on AI are behind the ones that let people run, and they’re not catching up by writing a policy doc.
But if you enable 60 people and nobody owns putting the pieces back together, in a year you’ll have a hundred small automations, four versions of the truth, and no idea which ones matter. I’ve walked into that and it takes longer to untangle than it took to build.
So enable everybody, and give one person the job of connecting it. That role used to be a quarterly thing. It’s daily now.