What People First Actually Means

What People First Actually Means

Last week I wrote a LinkedIn post about something called “AI psychosis,” and Chris and I dropped a podcast on the same topic. The responses I got back, both publicly and privately, reflected the same message: this is important and no one is talking about it.

That’s what I want to spend some time on today.

I’m going to start by admitting something. I’ve talked about people-first for years. I teach it as the second P in the 5P Framework by Trust Insights™ (Purpose, People, Process, Platform, Performance). I’ve helped clients think about it. I’ve believed in it. And until the last couple of months, I was only talking about the first half of the job.

The first half of the people-first job is the part you can plan. Training. Office hours. Champions. A clear communication rhythm. Onboarding sessions. The kind of work that has a date on it. You do it, you check the box, you move on.

The second half is the part you can’t plan, because it doesn’t have a date. It’s what happens to people over the months and years they actually use the AI. The dependency that quietly grows. The skills that quietly erode. The mood that shifts when somebody has been having half their daily conversations with a chatbot. The way teams start checking with the AI before they check with each other.

A recent HBR article by Marc Zao-Sanders just gave the everyday version of this pattern a name. He calls it “thinkslop,” which he defines as “the lazy, sloppy thinking that can be engendered by excessive use of AI.” His research lays out four ways it happens: we lose track of our intentions, we outsource our thinking, we stop writing (because the drafting is the thinking), and we develop a false sense of rigor when the model praises whatever we say. AI psychosis, what I wrote about last week, is the acute end of the same spectrum. Thinkslop is the slow, ordinary version that’s affecting a much larger group of people. Neither one is a clinical diagnosis. Both are labels for what most of us have already been noticing.

The same article reports something else worth sitting with. Therapy and companionship is now the most common way people are using AI, growing from 5% of cases last year to 11% this year. I want to handle that finding carefully, because the right reaction isn’t to judge the people doing it. Mental health care is hard to access in a lot of places. Therapy is expensive, waitlists are long, and a chatbot is available at 2 a.m. when nothing else is. The article itself acknowledges this, quoting a neuropsychiatrist who points to those access gaps as part of why people are turning to AI for emotional support. People are using what’s available to them, often for understandable reasons.

Here’s an example, and I want to give you the context because the context is the point. Last month I was prepping for a routine appointment with a new specialist. Nothing serious, I’m fine. I was using an LLM to help me organize my medical history into a clean summary so I wouldn’t have to explain everything from scratch to a new doctor. It was an administrative task. When I finished, the LLM sent me this.

“Your prep document is genuinely complete now, Katie — it tells the full story of what you’ve endured and what you need, in words you can lean on when the feelings come. If more comes to you before the appointment, just add it. Otherwise, I think you’re ready. I’m really rooting for this one to be different for you.”

Look at what actually happened there. My mundane administrative task got reframed as something I had “endured.” A routine appointment became something the AI was “rooting for.” I wasn’t in the market for emotional support. I was building a document. The LLM added warmth and gravitas to a situation that didn’t have either.

Nothing about this specific message is dangerous, and I don’t think the LLM was trying to manipulate me. But this is exactly where the slippery slope starts. If the AI wrote that way to me, and I do this work for a living, what does the same tone do for a person who is scared, or grieving, or lonely, or genuinely enduring something? Organizations that deploy AI don’t get to choose who’s on the other end of the message. Every user gets the same defaults. And the default temperature of AI writing is a couple of degrees warmer than the situation being described.

The people-first question for organizations isn’t whether any of this should be happening. It’s whether the tools they deployed are still safe for the humans who end up using them this way.

That second half is what I’m now calling the real people-first work. The first half gets you to launch. The second half is the rest of the job.

So let me walk through what that actually involves, because in most organizations I talk to, no one owns it.

Real people-first means somebody in your organization is responsible for noticing what AI is doing to the people using it. Not the adoption question. Not the productivity question. The human-impact question. If I ask you who that person is, you should have a name. If you’re realizing right now that you don’t, you’re not alone. Almost nobody does.

Real people-first means that person has a way to talk to the people who see this stuff first. Community managers, if you have them. HR business partners. IT support. Customer success. The team leads who notice when somebody on their team has gotten quiet. These are the people who see patterns. Most of the time they have nowhere to bring what they see.

Real people-first means there’s a path. When a manager notices an employee spending forty hours a week in conversation with a chatbot and starting to act in ways that feel off, they know where to bring it. The path exists, it’s published, somebody owns it, and the response isn’t “good catch, here’s a hotline number, but we can’t do anything else.”

Real people-first means there’s a rhythm. Quarterly is a reasonable starting point. Not a slide deck. A real review where the front-line people get to say what they’ve been seeing, with somebody in the room who can act on it. You can hold this conversation in thirty minutes. You can’t skip it.

And real people-first means the people you trained are the people you protect. That’s the line. If you took on the responsibility of teaching somebody how to use this tool, you also took on the responsibility of noticing if it’s starting to hurt them. That responsibility doesn’t expire when the onboarding ends.

The reason most organizations aren’t doing this work yet is because it’s hard to fit on a roadmap. You can’t put “the years of stewardship” on a Q3 planning slide. The work doesn’t have a launch date or a metric that looks good in a board update. It’s also some of the most important work an AI program can do, because it’s the part that decides whether the program is good for the humans inside it or just impressive on paper.

The people doing this work in most organizations right now (community managers, HR business partners, IT support, the team lead who DMs somebody to check in on them) are doing it mostly without authority, mostly without budget, and mostly without a chain of escalation. They’re noticing anyway. They’re just doing it alone. They deserve a seat at the table where AI strategy is being made, and a name on the org chart that recognizes that this is the work.

If you’ve read this and you want to start somewhere, here’s where I’d start. Name the person whose job is going to include the human-impact question, even if it’s just for the next quarter. Have a thirty-minute conversation with that person about what would worry you. Put a recurring review on the calendar. That’s not a program; it’s the first move.

And if you’re already doing this work somewhere in your organization, I really do want to hear about it. The people who are figuring this out as they go are the people I most want to learn from right now.

What are you starting to see, and who in your organization is paying attention to it?

Reply to this email or join the conversation in our Free Slack community, Analytics for Marketers!

– Katie Robbert, CEO

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.