Humans + AI: Redesigning Work, Roles, and Relationships for What’s Next

The most honest conversation about AI and the future of work isn’t happening in the technology track. It’s happening in the people track—where HR leaders, elected officials, talent executives, and researchers are grappling with questions that no amount of compute power can answer on its own: How do you build trust across a workforce that is anxious about change? What skills actually matter now? Who owns the decision when AI is in the loop? And how do you bring your whole organization along without leaving anyone behind?

These were the questions at the center of one of Innovation Summit 2026‘s most grounded and human conversations. A panel of five women—spanning a professional sports organization, a county government serving 1.2 million people, one of the world’s most recognized consumer apps, a research-backed consulting firm, and HIKE2’s own human-centered design practice—spent an hour cutting through the noise to talk about what AI transformation actually looks and feels like from the inside. Their conversation was candid, at times funny, and full of the kind of practical wisdom that only comes from people who are living this challenge in real organizations, right now.

About the Session

The panel was moderated by Olivia Smith, Director of Human-Centered Design & Strategy at HIKE2. Panelists were:

• Rachel Brecht, Chief People Officer at the Pittsburgh Penguins
• Sara Innamorato, Allegheny County Executive
• Brittany Mitlo, Director of Talent Acquisition at Duolingo
• Dr. Shannon Gregg, President of Cloud Adoption Solutions and adjunct professor at Point Park University

The combination of public sector, private sector, early-stage talent, and research perspectives produced a conversation that consistently challenged the assumption that AI transformation is primarily a technology problem.

The Productivity Myth: AI Doesn’t Lift All Boats Equally

Dr. Shannon Gregg opened with a finding from a meta-analysis published in the California Management Review—a study of studies, covering approximately 120 research papers on AI and productivity—that immediately reframed the conversation. The finding: productivity gains from AI do not apply universally. They vary significantly by role, by experience level, and by the nature of the work being done.

In software development, junior developers saw strong productivity gains from AI-assisted coding—but those gains were partially offset by the additional review burden placed on senior developers who had to catch errors the junior developers didn’t know to look for. The same pattern appeared among clinicians and help desk agents. In each case, AI made certain tasks faster for certain people, while creating new demands on others—particularly those with the expertise to evaluate AI output critically.

The implication for organizations is significant and often overlooked in AI planning. When leaders project productivity gains from AI adoption, they frequently assume those gains will be uniform across the workforce. The research says otherwise. The people who benefit most from AI augmentation tend to be those who already have strong domain knowledge because they can direct AI effectively and evaluate its outputs. The people who benefit least, or who generate new problems, are those who lack the context to know when AI is wrong.

This isn’t an argument against AI adoption. It’s an argument for designing adoption thoughtfully—understanding which roles will see immediate gains, which will face new burdens, and what investments in skill development are needed to make the benefits more broadly distributed. For HIKE2’s clients, this is exactly the kind of workforce mapping and change management design work that determines whether an AI deployment creates organizational lift or organizational friction.

Critical Thinking Is the New Core Competency

Across every panelist and every topic, one theme kept surfacing: critical thinking is the skill that AI makes more important, not less. At Duolingo, Brittany Mitlo described a deliberate shift in how her talent acquisition team evaluates candidates—with new emphasis on surfacing critical thinking in interviews, because it’s the capability that determines whether someone can use AI tools effectively or just generate plausible-looking outputs that require expensive rework downstream.

Rachel Brecht echoed this from the organizational design side, noting that recent research has raised concerns about cognitive decline in younger generations who’ve grown up offloading thinking to technology—and that organizations have a responsibility to intentionally develop the cognitive muscles that AI can otherwise atrophy. Her framing was striking: if we don’t design AI adoption in ways that preserve and strengthen human judgment, we risk building a workforce that’s dependent on tools it can’t evaluate.

“It’s not AI for AI’s sake. We have to go back to what we said before—to build the skills around where are the thoughtful points in the process where people still show up.” — Rachel Brecht, Chief People Officer, Pittsburgh Penguins

Dr. Gregg connected this to her classroom experience at Point Park University, where she assigned students to design an application that solved an unmet need in sales and marketing—and three students independently produced the same application. The prompt had been too open-ended, allowing AI to converge on a generic answer rather than driving original thinking. Her response wasn’t to ban AI. It was to redesign the prompt so that students had to engage their own judgment before reaching for a tool.

The organizational parallel is direct.

If AI adoption is designed purely around efficiency—do this faster, automate that—it will tend to replace the thinking rather than augment it.

If it’s designed around human-centered outcomes—what judgment do we want people developing, and how does AI support that?—it builds capability rather than dependency.

This distinction is at the core of how HIKE2 approaches AI deployment: starting with the human, the role, and the outcome before specifying the technology.

Government as a Case Study in Responsible AI Adoption

Sara Innamorato offered one of the most concrete and instructive examples of the session: Allegheny County’s decade-long experience using predictive AI models in its Department of Human Services—the department responsible for mental and behavioral health, reentry programs, homelessness response, and child welfare for 1.2 million residents.

The models were designed to do something genuinely difficult: take limited government resources and direct them toward people most at risk of entering expensive, high-stakes systems—before a crisis occurred.

In child welfare, in homelessness, in behavioral health, the goal was early intervention rather than late response. And the data showed that the approach has worked: independent analysis found the models reduced racial disparities within the child welfare system, a result that Innamorato cited as both a validation of the approach and a reminder of what’s at stake when AI is applied to decisions that affect vulnerable people.

“These models had to be designed with intention, with a ton of community collaboration, with a tremendous amount of transparency. And all of this modeling is always checked by an expert. It is always human checked. It is never the answer for a person.” — Sara Innamorato, Allegheny County Executive

The key conditions that made this work: the models were co-designed with community input, subjected to ongoing public dialogue when concerns arose, and—critically—never treated as a substitute for human judgment. Every AI output is reviewed by a human expert before action is taken. The AI narrows the field and surfaces risk signals. The human makes the call.

For organizations in financial services, insurance, healthcare, or any sector where AI-informed decisions affect real people’s lives, this is the template. AI can dramatically improve the targeting and efficiency of human expertise. It cannot replace the accountability, the empathy, or the contextual judgment that high-stakes decisions require. Building that human checkpoint into the process isn’t just a compliance requirement—it’s what makes the system trustworthy enough to scale.

Innamorato also highlighted the public sector’s role in preparing the broader workforce for an AI-enabled economy—through partnerships with community colleges, workforce investment boards, and employers co-designing training programs that build AI literacy across sectors, not just in knowledge work. Her point: AI readiness is a regional economic development issue, not just an organizational one.

Adoption Is a Design Problem, Not a Rollout Problem

One of the panel’s sharpest debates was about AI adoption—specifically, why so many well-resourced organizations with thoughtful AI strategies still struggle to get their people actually using the tools.

Dr. Gregg’s answer, grounded in diffusion of innovation research, was blunt: you can’t just open the door and assume people will walk through it. Adoption requires a deliberate plan that addresses different people at different starting points—and that plan has to work both top-down and bottom-up simultaneously.

Five generations are currently active in the workforce—a first in history—and each arrives at AI with different reference points, different anxieties, and different intuitions about what the technology can and can’t do. A rollout designed for one generation will underserve the others.

Dr. Gregg’s recommendation was to meet people where they are, establish clear acceptable use policies that provide guardrails rather than prohibitions, and create space for questions and concerns before mandating adoption.

Brittany Mitlo offered a reframe that several panelists immediately adopted: instead of the tsunami metaphor—which creates panic and the instinct to flee or freeze—think about AI transformation the way we think about the Industrial Revolution in retrospect.

Looking back, the changes seem obvious and even inevitable. People who lived through them found ways to adapt, develop new skills, and find new roles. The transition was real and difficult, but it produced more opportunity, not less, for people who had the support and the skills to navigate it.

Her prescription for leaders: stop framing AI as something being done to your workforce, and start framing it as something you’re figuring out together. Curiosity and thoughtfulness—not urgency and pressure—are the conditions under which people actually change how they work. Organizations that create those conditions will have workforces that develop genuine AI capability. Organizations that mandate adoption through fear will have workforces that comply minimally and resist meaningfully.

Rachel Brecht added a forward-looking dimension: as AI agents become ubiquitous, every individual contributor will effectively become a manager—responsible for directing, evaluating, and quality-checking AI outputs as part of their daily workflow. Teaching people how to manage—whether the thing being managed is human or technological—is one of the most important workforce development investments organizations can make right now, and very few are making it deliberately.

Decision Authority Is Shifting—and That’s Mostly a Good Thing

The panel’s discussion of decision authority surfaced one of the more optimistic threads of the conversation. AI, it turns out, is not just reshaping which decisions get made—it’s reshaping who gets to make them.

Mitlo described seeing individual contributors at Duolingo gain new autonomy: AI tools let them think through problems independently, test approaches, and move forward without requiring the extensive sign-off processes that previously gated their work. The traditional deep hierarchy—where information flows up and decisions flow down—is beginning to dissolve, replaced by smaller, more agile decision-making units that combine human judgment with AI-assisted analysis.

Brecht noted that this shift, while broadly positive, carries its own risk: not everyone at every level of an organization has been equipped to make good decisions with greater autonomy. If organizations push decision authority down without simultaneously investing in the judgment and context-setting skills that good decisions require, they will get faster decisions that are not necessarily better ones.

The answer isn’t to slow down the distribution of decision authority—it’s to accelerate the investment in the capabilities that make distributed decision-making work. That means leadership development at levels of the organization that haven’t historically received it, clear frameworks for evaluating AI outputs, and the psychological safety to flag when something the AI produced doesn’t seem right. For HIKE2’s clients, this is a change management and organizational design challenge as much as a technology one—and the organizations getting it right are the ones treating it that way.

Watch the Full Session

The full Humans + AI panel covers additional ground on the generational divide in AI adoption, the specifics of Allegheny County’s AI task force and acceptable use framework, how Duolingo is evolving its early talent strategy in response to AI, and a rich audience Q&A that includes advice on building anonymous feedback mechanisms before rolling out AI training programs.

Your Workforce Is Your AI Strategy

Technology deployments succeed or fail based on the people who use them. The organizations that are getting the most out of AI aren’t necessarily the ones with the most sophisticated tools—they’re the ones that designed the human side of adoption as carefully as the technical side. That means understanding the current state of how people work, building for the full range of roles and experience levels, and creating the conditions for genuine behavior change rather than surface-level compliance.

Human-centered design is at the core of how HIKE2 approaches every AI engagement. If you’re thinking about how to bring your workforce along on this journey—or how to build the governance, change management, and skills development infrastructure that makes AI adoption stick—we’d love to talk.

Contact HIKE2 to start the conversation →