Building a Human-AI Partnership That Earns Trust — Part 1
July 8, 2026
Trisha Merriam's research opens the series: organizations blame agents for errors while no human takes accountability. Chris Carolan and Erin Wiggers trace it to industrial-age structure, landing on AI-Human Partnership over Replacement. Chris coins "automation without abdication."
Moments from this episode
6 short cuts from this conversation.
Key takeaways
Treating AI as a black box you cannot influence is what lets a team disown its errors — you control the setup, the tools, the skills and the flow of information, so the output is still your responsibility.
Say "human-AI partnership, not replacement" out loud as a leader, because in the absence of that sentence people will read any AI investment as a threat to their job, which is what the visible examples have taught them.
Blaming the agent is not new behavior: for two hundred years, work crossing between people with different expertise has produced the same reflex — if I did not do it, it is yours.
Put an approval gate in front of anything that leaves your organization, so nothing reaches another human without your explicit permission and your ability to edit it first.
Accept the cost of that gate — a message that goes out a little late with your blessing is worth more than an immediate one nobody checked.
Give an agent exactly the permission set of the person operating it: if that user cannot edit a record, neither can their AI.
Guide an agent on three axes — scope (what is expected of you to be successful), access (do you have what you need to deliver it), and context (do you know how this is actually done, beyond the best practices sitting in training data).
Standardize the work before you hand it to an agent: a library of templates with a worked example of what good looks like caps how far anything can wander.
Run this test on your own stack — for each place AI touches the work, ask what the blast radius is if it goes wrong and how likely that is; the pair that scores highest on both is square one.
Use three simple gates for what may leave without your specific approval: does it go to a client, does it go to your boss, does it go to somebody who tends to pick my work apart.
In a new system, visibility is what produces trust — certainty and accountability both rest on it, and automation that hides work before people know the rules of the game undercuts adoption.
If you are downstream of somebody else's AI, ask to help QA the result as soon as you hear it is being built; you can shape the context even when the system is not yours.
Transcript
This transcript was generated by AI and may contain errors. It is provided to help you skim and search the conversation.
Keep learning
You just watched the work. Now build it.
The next step is doing it yourself — live, with people on the same path.
Get the next conversation in your inbox
One signal-dense note when a new episode lands. No noise.