How People Work

How People Work, Live! — Dan Hon and Hillary Hartley — notes

Simon Morgan-Wilson

• 1 min read
Share

On Friday I sat in on How People Work, Live!, Dan Hon's regular live show, a spin off from his workshop of the same name.

Basic premise: how well you work with other people matters as much as how good you are at the work.

This time Dan was talking with Hillary Hartley, formerly of 18F and Ontario's first chief digital officer.

It was live from 5pm BST on Friday, so I didn't have the energy to write full notes. Instead – ahem – here's some things that caught my attention from Dan and Hillary's chat. (Dan might share a full recording at some point.)

Thanks for talking openly, Dan and Hillary!

Recon before persuasion

  • Horizontal leadership is changing things across silos you don't own. No authority, so it runs on influence, credibility and trust.
  • Map people on authority versus trust. Then what they care about.
  • Who can say yes? Who can quietly kill this without me knowing? Whose turf am I on?
  • It's rarely just one person in the hierarchy. It's the committees set up to shepherd things along.

Fixed versus flexible

  • Be plain about what's non negotiable. The outcome, legal review, anything touching people's data.
  • The how is the team's call. Tools, sequencing, research, methods.
  • Can the team act well without you in the room?
  • Half of what's "fixed" is custom. Teams need permission to say so.
  • Judge directors on how few decisions reach them that should have been made closer to the work.

Demos not memos

  • Show the work, skip the status report. Otherwise people give feedback on the thing in their head.
  • Take the people who did the work into the briefing. Hillary took the policy analyst, and the designer who'd sat in on the research, into ministerial briefings.
  • Questions get answered there and then...
  • ...otherwise people give feedback on the thing in their head.

Learning ready, not AI ready

  • Can budgeting, planning and governance move on what we learn?
  • AI failures are quiet. Confident, fluent, wrong. Audit trails weren't built for drift over a few prompts.
  • AI mostly speeds up the governance failures already there.
  • A vibe coded proof of concept is fine. It isn't a pilot, and it isn't on its way to production.