Designing the Human-AI Boundary by Role — AI Readiness, Part 5
- Trisha MerriamHost
- Chris CarolanHost
- Erin WiggersHost
November 4, 2026
Part 5 of AI Readiness asks who owns what when AI joins the team. Chris Carolan, Trisha Merriam and Erin Wiggers on defining roles, what success looks like, starting from the outcome, no data entry ever, and negative prompting.
Moments from this episode
8 short cuts from this conversation.
Key takeaways
Start with the roles. Erin says it is crazy how often you have to start there, because most roles, she feels, are vibes-based: you do whatever comes across your desk.
Define the role, then jump to the end. Chris asks what success looks like for that role, and says everything in between, on a human to human level, is different.
The same role is not the same thing everywhere. In Chris's example, a marketing coordinator role that looks the same changes with what is being sold, who is involved, and whether they are using GPT or Claude.
Humans need to say out loud what they are good at, what they aren't, what they want to be good at, and what they don't want to do. Chris asks how to have that conversation in a meaningful way.
Trisha comes at it with "I don't want to do any of it," then asks what only she can do. Chris thinks that is a fair place to start, because you find out which part you do have to do.
Chris sees correcting at the right moment as a human value, and he says we're a ways out before that changes.
Data entry is the one boundary Chris says he is confident can be set almost across the board, at least if your stuff is digital, and the only generality across roles he has found.
Erin's line between AI do it and me do it: if she needs to learn something from it, there has to be some manual part. She also wants to be able to back up what she hands a client.
For Trisha, the more public facing the result and the riskier it is if it's wrong, the more it points to human intervention.
AI is easy to bias, Chris says, and the way you frame it is everything.
Chris compares negative prompting with giving people responsibility without authority, tools, scope or the vision, and says a lot of it is based on not trusting the other thing.
In his telling, a prompt worried about risk gets you a big table of what was avoided, and another ask for approval you said it didn't need. Models get better, he says, and last year's best practices are still baked in.
Erin starts from the outcome, not the input, and works backwards: what do I need to know, when, and what's the easiest way to get it. Sometimes the answer is AI, sometimes the client, sometimes her.
Erin says it is hard to put together a process that arrives at your expertise, because she doesn't have your expertise. AI may get to a V1 draft, and the expert has to come in and contribute.
Chris says frontline people want to sell stuff and help customers, so use AI is never the outcome for them. For the person consulting, coaching or building, he suggests adding with AI on the end of the outcome, and he thinks the standardizable outcomes are unified views and unifying the context.
Chapters
- 00:06Part 5: the human AI boundary by roleTopic
- 00:44Where have we been so far?Question
- 02:22Where do you start?Question
- 03:37Why this boundary is hard to defineTopic
- 05:24Define the roles, then jump to the endTopic
- 07:05How do we have that conversation?Question
- 07:29I don't want to do any of itTopic
- 08:44Building an agent with Breeze, and no period youTopic
- 10:28The ability to correct at the right momentTopic
- 11:30No humans will be doing any manual data entryTopic
- 13:41Is it that easy?Question
- 15:34I don't know, Claude did itTopic
- 16:29The riskier it is, the more a human should lookTopic
- 18:12Negative prompting and responsibility without authorityTopic
- 18:57Nectar and the Claude Code teamTopic
- 22:35The quote update and the big tableTopic
- 24:21Everything gets created from a risk perspectiveTopic
- 25:14This prompt kind of looks like you're worriedTopic
- 27:00It comes down to processTopic
- 27:39After the roles, outcomesTopic
- 30:03Start with the outcome and work backwardsTopic
- 33:16It always comes back to building contextTopic
- 35:00The scroll wheel and muscle memoryTopic
- 37:46Show me what you're doingTopic
- 38:06You just add with AITopic
- 40:00Any generalities across roles?Question
- 42:40Standardizable outcomes and unified viewsTopic
- 43:37How many more AI readiness episodes?Question
- 45:19We're going to keep talking about itTopic
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.