AI agents and automated workflows are becoming real rather than theoretical. Many of the companies struggling with them are not lacking tools. They are lacking foundation.
What is usually missing:
• Disconnected data
• Siloed systems
• Undocumented processes
Something I have been saying to peers for a while still holds. Sometimes you have to go slow to go fast.
Document processes first. Align cross-functionally. Test automation, but tie it to a clear strategy.
And a hard truth. Connecting data should be measured in months, not years, especially when much of that work should have been done years ago to improve the customer journey.
The problem with "go slow to go fast" is that it is easy to say and almost impossible to fund, because the slow part has no visible output. Nobody has ever been congratulated for a quarter spent settling definitions.
The way I have seen it actually work is to attach the foundation work to something that already has a deadline. An integration. A platform migration. A reporting rebuild leadership already wants. The definitions get settled because something visible depends on them, rather than as a project of their own that has to compete for attention against work that produces demos.
The alternative is the pattern most organizations are already in, where the foundation work is agreed to be important, scheduled for next quarter, and has been scheduled for next quarter for two years.
Foundation work rarely fails because somebody decided against it. It fails because nothing ever forced it to be now.