Prompt: Fix the Org

Once upon a time, I had a colleague who kept a single post-it stuck to his monitor that read ‘Fix the Org.’ Whenever we hit some bureaucratic roadblock or sudden top-down shift in priorities, he’d point to the post it and say “Oh right, forgot to fix the org.” And we’d laugh.

This was an organization highly invested in its Agile SAFE transformation, working to create smaller empowered teams, and human-centered design practices. I got to meet some great thought leaders that would swing in seemingly from the rafters to speak to us about best practices … in general, never specifically. Because drawing a map is fun, navigating terrain is hard. It also requires actually being on the journey.

For me, this dissonance crystallized in a workshop I attended on how to conduct better meetings. The expert leading it had been a former employee and had written a book on how to have better meetings. At the end of the workshop, I joined a cluster of folks to thank him. Someone asked how he had leveraged his techniques at the org specifically, if he had any tactical advice. He sighed and admitted that he hadn’t been able to make headway because of cultural challenges, but had learned a lot that informed his book. Luckily, the book was free. Thanks to the org.

I’m seeing a similar Elven map versus the Hobbit footpath unfold with AI. Start-up leaders are peacocking about their AI advantages, the efficiencies gained (I’m aware of the Meter study but it’s based on older models which is unfair). Then you scratch the surface and see the absolute emerald greenfield they’re operating in, versus say, a bank, responsible for money movement in the millions, a code base approaching Mordor, and don’t forget the SEC.

How do you even begin to create markdown files that outline that byzantine complexity? It took at least 12 years to write Lord of the Rings and no stakeholders were involved. Fun exercise, map your org to the nine races: Elves, Humans, Dwarves, Hobbits, Orcs, Trolls, Ents, Maiar, and Valar. I bet they all exist and you may even need to add some of your own.

I'm seeing numerous experiments that streamline code generation or design creation. Then I inevitably get asked to create a product spec so the experiments can have better context — and I don't even know where to begin.

I mean, I could make something up, but if you actually want to test something you need a realistic problem, and those don't just come from a product person typing. There's a whole menagerie of people you need to get information and then buy-in from. More than half of them are absolutely not going to read your Markdown file, but will view a powerpoint. Some have a spreadsheet template you need to use and do not want to hear a word about it. Most have been part of an Agile transformation for at least five years, logged onto Jira once, and haven’t been back. They’ll approve your designs if you grab time on their calendar and talk them through it while they eat Sweetgreen.

And that’s when you hit ‘fix the org.’

A product team 10x with Claude that still lacks empowerment to own their roadmap, manages deadlines and not outcomes, has to navigate misaligned business objectives, is just a product team with a new tool and unrealistic expectations.

AI is only as good as the context it’s given. And in most enterprise organizations, context isn’t a markdown file. It’s a journey through Middle-earth.