AI use

So who owns the outcome?

AI, sure. But who's accountable when it's wrong?

Three posts on org culture. This one's about who actually gets a say when AI enters the room.

Everywhere I look right now, someone's telling me to inject AI into something. Use it or get left behind. Adapt or become irrelevant. Every second article has the same energy: the train is leaving the station, and I'm apparently meant to feel the panic on cue. There's no reward in heaven for early adopters, and there's no rapture happening in your on-boarding flow either. Just a lot of people making very confident decisions on a very short timeline.

Full disclosure: I'm an Erika Hall fan, I read her books, blog, LinkedIn posts. She posted something recently that put language to a feeling I'd been sitting with for a while — that this kind of urgency isn't strategy, it's marketing wearing an inevitability costume, and some of the loudest people pushing it have lived through enough tech cycles to know better. That landed exactly the way she meant it to.

With all the "adapt or die" articles out there, not one of them has convinced me that what I actually do for a living has fundamentally changed. Tools, sure. The core? No. I'm an experience designer. My job is to be the conduit between a person and the thing they need — ethically, safely, in a way they can actually understand. That was true before anyone put AI in the flow. It's still true now. The bells and whistles changed. The job didn't.

So before the mandate gets to set my questions for me, I'd rather set my own.

Should this even be here?

Not "can we use AI" — of course we can, everyone can, that's a procurement decision, not a real question anymore. The real question is what it's actually doing. Helping? Facilitating? Quietly replacing a step that used to involve a human making a judgment call, without anyone flagging that as a risk worth examining? "AI knows better" isn't an answer. It's a deflection wearing one.

Does anyone on the other end actually want this?

Not "does it perform well in testing." Does it resonate. Who's delighted by the new conversational flow, and who's exhausted by it — and are we honestly checking, or just assuming enthusiasm because the demo killed in the boardroom?

Accessibility isn't a footnote here, it's the whole test. A chatbot that makes an SVP's demo look slick can be a genuine obstacle for someone who just wants to know what their loan payment is, in plain terms, without negotiating with a personality. Conversational doesn't automatically mean better. Sometimes the most human interface is the plainest one: a clear number on a clear screen.

Does it actually make sense?

Financially speaking. You know, the model costs, token pricing, third-party dependencies you didn't have before, SLAs you're now trusting for something that used to be a static screen and a form field. "It's innovative" isn't a line item, so somebody needs to be doing that math honestly before deployment, not finding it in a vendor invoice eight months later.

There's an older discipline for this, and it still applies: before you fire an existing person, process, or system and replace it with some new combination of the three, audit what you actually have, what you actually need, and the real cost of the gap between them. Nobody's running that audit when the instruction from three levels up is just "use AI here." That's a mandate, not a decision.

And now the big one...

Who's to blame when it gets it wrong? "The AI made me do it" is not as good a line as "the dog ate my homework". It's already showing up in courtrooms, minus the punchline; litigation over AI-mediated decisions that misled people, misinformed people, or made a call nobody can trace back to an accountable human. "The model did it" is a terrible legal position, and a worse thing to say out loud to the person it hurt.

None of these four questions has a home. They get asked in hallway conversations if they get asked at all, answered by a UX backlog that ships after launch if it ships at all, negotiated by whoever signed the vendor contract with no connection to whoever owns the customer relationship, and ignored entirely until a lawyer needs the answer.

Sound familiar?

It's the listening circle again. The thing that matters most to the person with the least power in the interaction — whether they understood what happened to them, whether they had a real choice, whether there's someone accountable when it goes sideways — has no formal seat at the table where those decisions get made. Not because anyone's malicious. Because the table got built around risks to the institution, and nobody drew a box for whether the human got what they were owed.

Model risk has a seat. Compliance has a seat. Legal has a seat. The customer's actual comprehension, actual choice, actual way out: an empty chair everyone assumes someone else is sitting in.

I saw the gap and it irked me enough to actually do something with the irritation — write it up properly, argue it out, get it peer-reviewed. It's now published internationally. A case for putting someone in that chair with real authority: standing to test a flow before it deploys, and halt it when it doesn't meet the bar. It's in Touchpoint, the SDN journal (also on my LinkedIn if you want the humble-brag version). I'll do a full post on the paper itself soon.

Whoever has the least power in an interaction is usually the one whose experience of it is least likely to have a formal owner. The mandate says use AI everywhere. Nobody's mandating who's accountable for what it does to the person on the other end of it.