Stop Buying Licenses, Start Building Minds: What an AI-Ready Team Actually Looks Like

You've felt the pressure by now. The board wants an AI story. Managers are forwarding LinkedIn prompt playbooks. Someone in the exec meeting suggests buying ChatGPT licenses for the whole company and calling it a strategy.

In a recent Croissant webinar, CEO and co-founder Zoltan Szalas sat down with Stephany Oliveros (AI strategist, university lecturer, TEDx speaker, and CEO of Lyra AI, a platform that uses transparent, evidence-based algorithms to match women to tech jobs) to talk about what organizations are getting wrong about AI adoption. Her central argument lands squarely on your desk: preparing a workforce for AI is not primarily a technology challenge. It's a people challenge.

That reframe changes who owns the problem, what "AI training" should look like, and which skills you should be hiring and developing for. Here's what the conversation surfaced.

Why isn't an AI tool stack an AI strategy?

Stephany sees the same pattern everywhere: FOMO-driven purchasing. Companies buy licenses, book a Copilot course, circulate a list of "magic prompts," and check the box.

Her objection is simple. AI can't be treated like a line item on a training checklist. "It can't be another checkbox," she says, "because it's so deeply embedded in how it's changing our behavior as individuals and as a society." She points instead to research-backed frameworks, like the six principles of human-AI collaboration developed by Berkeley researchers with the California Management Review, that treat AI adoption as organizational design, not software rollout.

There's a practical reason the tool-first approach fails, too: the tools won't hold still. Frontier models have already made most single-purpose AI apps redundant, and every few months the leaderboard reshuffles. Any training program built around mastering one product is depreciating the day it launches. As Stephany points out, Excel didn't update at this pace, so mastery of the tool made sense. With AI, the durable layer sits somewhere else entirely: "The first foundational layer is who you are as a person and how you can deal with this technology."

Think in architecture, not apps

If not tools, then what? "Instead of thinking in terms of tools, we need to reframe and think in terms of architecture," Stephany says.

Her suggestion: develop employees who think like chief operating officers of their own function. People who understand how work actually flows, where the bottlenecks are, what "good" looks like, and how to document all of it. That diagnostic ability (seeing the system, not just the task) is what transfers across every model release. The specific mechanics of using any given AI assistant can be trained in weeks. The wit to know where the problems lie cannot.

She illustrates the payoff with a story from a chance dinner conversation in Serbia. Executives at one of the world's largest recycling companies described a problem they'd assumed was unsolvable: identifying the mix of metals (iron, stainless steel, copper) in incoming cargo. They had drones, cameras, and data, but no answers. Stephany pointed out that this is a textbook case for computer vision and anomaly detection, the same family of AI that powers self-driving cars. Their reaction: why has no one told us this before?

What does an AI-augmented role actually look like?

The most vivid framework of the session was Stephany's model for how individual roles evolve. She calls it becoming the conductor of your own orchestra.

Rather than one all-purpose chatbot, she describes designing a set of specialized agents, each with a narrow role, coordinated by a human tied to a business objective. Her own newsletter workflow, built with no marketing background, looks like this:

• A research agent running sentiment analysis on blogs, Facebook, and Reddit
• A competitor-analysis agent working the same research from a different angle
• An inspiration agent that reads the newsletters she subscribes to and extracts what top brands do well
• A copywriting agent producing SEO-optimized drafts, checked by a separate SEO audit agent
• A creative-assets agent pulling and generating imagery
• A "manager" agent that defines success metrics, analyzes misses, and improves the system for next time

She still edits. "This doesn't really sound like me" remains a human judgment. But her role has moved up a level. "I've elevated myself as the conductor of this orchestra," she says: the strategist who assigns roles and owns the outcome, rather than the person typing every word.

The implication for your team is significant. Even individual contributors become, in effect, managers. Managers of agents rather than people. The skills that matter are decomposition (what is my job actually made of?), delegation, quality judgment, and accountability for outcomes. Those look a lot more like leadership competencies than software skills, and they should start showing up in your job architecture, hiring rubrics, and development plans accordingly.

AI-ready beats AI-first

Stephany draws a sharp line between two postures. "The goal is not to create an AI-first organization," she says, "but something I call the AI-ready organization."

AI-first (automate everything, deploy agents everywhere) has a hidden economic flaw: today's AI pricing is heavily subsidized. Costs will rise, and organizations that wired AI into everything indiscriminately will find that some of it should have stayed human all along. AI-ready asks the better questions: where are the genuine opportunities, when should we use it, and, just as important, when shouldn't we?

An AI-ready organization is also tool-agnostic by design. If one vendor's model goes down or gets leapfrogged, work continues. That resilience comes from investing in people and systems rather than any single product. Which is exactly the investment HR is positioned to lead.

What does the gender gap in AI adoption teach everyone?

Through her research, Stephany found that women were adopting AI at roughly one-third the rate of men two and a half years ago, a gap that has since narrowed sharply. The reasons women gave are revealing: concerns about whether using AI is cheating, about environmental cost, and about whether outputs can be trusted.

Her insight is that these "skeptical" attitudes are precisely the attributes AI-ready teams need. Trust calibration, ethical questioning, and reluctance to over-rely on unverified output are the raw material of good AI governance: the same instincts behind evaluation rules for agentic systems and bias checks in high-stakes uses like hiring. (She notes research showing large language models advise women to negotiate roughly 20% lower salaries than men on identical prompts. A reminder of why skepticism belongs in the loop.)

For your People team, that reframes inclusion as a capability strategy. Don't just push adoption rates up. Recruit the skeptics into building the guardrails. And for anyone anxious about starting, her first step is disarmingly simple: "Open it and ask it questions." Ideally alongside a community or a colleague, and never with private company data.

Where can you start?

Pulling the conversation together, a practical starting sequence looks like this:

1. Audit your current "AI strategy." If it's a license count and a prompt library, you have procurement, not readiness.
2. Train systems thinking before tools. Teach teams to map their workflows, bottlenecks, and success criteria first. (Stephany and Zoltan both recommend Donella Meadows's Thinking in Systems; Stephany adds The Fearless Organization on the psychological safety that lets creative ideas surface.)
3. Redefine roles around orchestration. Build the conductor model into competencies and career paths.
4. Set governance early. Risk-tier your use cases, watch for bias in people decisions, and put evaluation rules around anything agentic.
5. Make readiness fair. Track who is and isn't adopting, and turn skeptics into system-builders.

There's a familiar shape to all of this. The companies that navigate distributed work well don't treat it as a perk to purchase either. They treat it as a system to design: clear policies, governed access, measurable outcomes instead of scattered stipends and status-quo spreadsheets. Building an AI-ready team is the same discipline applied to a new technology. In both cases, the organizations that win are the ones that design for people first.

Croissant helps companies manage distributed workforces with governed access to 700+ flexible workspace locations across 100+ cities. No leases, no long-term commitments, and workspace spend your team can actually see and manage. Learn more at getcroissant.com.

Watch the full conversation with Stephany Oliveros and Zoltan Szalas: The AI Mindset: Building Future-Ready Teams Beyond the Tools.