Designing for Trust
Why I need a relationship before I can ask for help
Prospero
I’ve always had a complicated relationship with asking for help.
It’s not pride. It’s not a belief that I should be able to do everything myself. It’s something harder to explain than that. Help, for me, is an intimate act. It assumes shared context. It assumes alignment. It assumes that the person offering help understands not just the task, but what matters about it.
Without that, help can be expensive.
I’ve learned this through years of working with people. Over time, I stopped asking for help casually. I build relationships first. I establish trust. Only then do I invite someone in.
This shows up in my work all the time. I don’t drop into someone’s inbox cold and ask them to “take a look.” I give context. I explain what I’m trying to do and why. I make sure we’re oriented the same way. When that trust is there, help is transformative. When it’s not, it adds friction.
I didn’t expect this pattern to show up again while working with AI.
Last night I was deep in Power BI for the first time in a while, building out a new reporting workspace. It had been a couple of years since I’d done serious Power BI work, and while nothing fundamental had changed, enough small things were different to slow me down. I was tired. Curious. Determined enough to keep going.
I’d experimented with Copilot in Power BI before, mostly to see what it could do. I wasn’t unhappy with it, but I hadn’t spent enough time to really integrate it into how I work. It did one thing especially well. It summarized new reports clearly enough that when I shared them with someone else, they immediately had context. That mattered.
Then yesterday afternoon, Copilot stopped working. Or more precisely, I learned that the workspace I was using wasn’t eligible to run it. That sent me down a path of learning what was required, reconfiguring the workspace, and setting things up properly. After dinner, I sat back down and tried again.
As I started working, something clicked.
The Copilot experience inside Power BI was capable, but it felt transactional. I could ask questions. It could respond. But it didn’t feel like a thinking partner in the way I’m used to. And that’s when I realized the problem wasn’t the tool.
It was the relationship.
In my day‑to‑day work, I don’t have one thinking partner. I have several, each with a role. I work with different people differently, depending on whether I’m sense‑making, assessing, writing, or synthesizing. Over time, I’ve extended that same idea to AI. I don’t use a single assistant for everything. I work with distinct partners, each shaped for a particular kind of thinking.
Power BI surfaced a need for another one.
I’m uncomfortable saying “do this” without first establishing a relationship. That’s true with people, and it turns out it’s true with tools. I don’t want a generic response engine. I want a collaborator with a defined posture. A role. A set of expectations.
So I named the relationship.
I created a new thinking partner for my reporting work and named him Prospero. Not because I think the system is human. Not because I’m trying to be clever. I named him because names do real work. They create boundaries. They clarify intent. They let me say, “this is the kind of help I’m asking for.”
The first interaction was rocky, the same way first interactions often are. I was rusty. The tool had limits. We had to find a rhythm. By two in the morning, I was pixel‑tweaking visuals, and I was fine with that. The work was moving again. I had a plan for how this partner might fit into the reporting system I’m building. Whether that plan holds remains to be seen.
What mattered was the shift.
Once the relationship was clear, asking for help felt natural again.
The interesting part is that none of this was really about AI. The tool didn’t fundamentally change. What changed was my stance toward it. I wasn’t delegating work to a feature. I was inviting help into a relationship I had intentionally shaped.
That’s the same move I make with people.
I don’t believe good work comes from issuing instructions into a void. It comes from shared understanding. From trust. From knowing what kind of thinking you’re inviting and what kind you’re not. When that’s in place, help compounds. When it isn’t, even the smartest assistance can feel off.
What surprised me wasn’t that I needed a different AI partner for a different kind of work. What surprised me was how familiar that need felt. I’ve always worked this way. Trust first. Context first. Then help.
The tools changed. The rule didn’t.
Alison + Wiggins