Curious, Not Technical

Self-serve data sounds like a great idea until you actually hand someone a warehouse and a login. Access was never the problem. Comfort was. Every team I've worked with already has more data than anyone's looking at, and most of it goes untouched. It isn't locked away.

I'm not technical. I don't want to be. What I care about is behaviour: why someone uses a product the way they do, in the context they use it in. For most of my career, if I had a hunch about that, I'd write it down, do some user testing, check GA, take it to an analyst, wait, and hope the question I asked was still the one that mattered by the time the answer came back. That gap, between having a hunch and being able to check it, is where good instincts go to die quietly in a backlog.

Databricks is the first tool that's actually closed that gap for me. I can ask a question in plain language and get a query back. I can drop into a notebook someone else built and just read it, no fear of breaking anything. I can build a dashboard I actually understand because I built it, instead of inheriting one where I don't know what half the filters are doing. None of that required me to learn to code. It required the tool to meet me at the level of curiosity I already had, instead of demanding a level of technical fluency I was never going to reach.

That's the actual case for self-serve data, and it's narrower than the hype suggests. It isn't "give everyone a warehouse and productivity will follow." It's: if you want people to actually understand the systems they're designing for, instead of inheriting a version filtered through someone else's dashboard and someone else's assumptions about what's interesting, you have to make asking a question as cheap as having one.

Comfort is the design problem here, which is maybe why this is the tool that got me, of all people, excited about data. I've spent a career arguing that good tools remove the tax between a person's intent and their ability to act on it. Self-serve data is the same argument, applied to curiosity about a system instead of curiosity about a screen. The question isn't whether self-serve data is good, more people understanding the systems they're part of is always good. The question is whether we're building the on-ramp for people like me, who were never going to take a SQL course, or just handing out warehouse keys and calling it done.

The execution tax (neurodivergence angle)

Ideas Were Never the Problem

For as long as I can remember, my head has moved faster than my hands could follow. Ideas arrive in bursts: connected, half-formed, ten at once. By the time I've wrestled the first one into a sentence, spelled right, in the right order, three more have already shown up wanting attention. I'm dyslexic, which is the actual mechanism here. My brain builds meaning associatively, in leaps. Written language demands one word at a time, in order. For years the tax on every idea was translation: could I get it out fast enough, clean enough, to be taken seriously, before I lost it to the next idea already crowding in.

Design became my workaround before I had a name for it. Visual thinking doesn't make you queue your thoughts in a line. A sketch, a board, a flow can hold five ideas at once, the way my head does. I built a career on the one channel where speed of thought and speed of output were finally close: I could hold the whole system and the detail at the same time, and show it, without going through the bottleneck of spelling it out.

But even design has a translation tax. Every idea still had to become a deck, a doc, a brief someone else could act on. That's where the gap reopened. I knew exactly what I meant. Getting a stakeholder, a team, a brief to know it too, in words, at the speed I was thinking, that was the part that drained me.

AI didn't give me ideas. I have plenty of those. What it closed was the distance between having one and getting it into a form other people can use. I bring the thing that was never the bottleneck: knowing what I want, the taste to tell a strong direction from a weak one, seeing a system before anyone else does. AI brings the thing that was: fast, clean execution, on demand, in whatever shape the moment needs, a brief, a mockup, a structure, without me slowing my thinking down to the speed of my typing.

That's the part of the "AI helps neurodiverse people" conversation that gets flattened into something smaller than it is. Assistive tech usually means helping someone keep up. This is different. The output pipe finally matches the size of what's coming through it. My ideas didn't get better. My ability to make and communicate them did, and those were never the same skill, even though most workplaces are built as if they are.

I don't think I'm an edge case. Anyone whose thinking has ever outrun their tools, because of how their brain works, because they were never trained in the "right" format, because English isn't their first language, is about to feel this same shift. The scarce skill was never spelling, or syntax, or knowing the software. It was always knowing what you want to say and having the judgement to know when it's right. That part doesn't get automated. It just finally gets heard.