How I think about AI’s impact on design

Leadership | 7 minute read

OVERVIEW

AI excels at production, helps with exploration, but still requires human judgement with framing problems. Design leaders need to invest in strategic capability, not just automation.

A framework for exploring AI in design

Since the first large language models arrived, I've wrestled with AI in two ways: bringing an AI assistant to market at NetDocuments and understanding how AI reshapes design leadership itself.

Every design leader I know is asking how to use AI but suggest we also need to ask what parts of design work AI actually changes and what that means for how we lead.

I’ll use framing from a former Fidelity Investments colleague, Jen Cardello, who broke the design research into three parts:

  • Right Problem: Are we solving the right problem? (intent, framing, business alignment)

  • Right Solution: Are we building the right solution? (exploration, validation)

  • Done Right: Are we executing well? (production, quality, delivery)

AI's impact differs dramatically at each phase. AI excels during 'done right,' helps with 'right solution,' but still requires human judgment with 'right problem.' Understanding this nuance separates design leaders who optimize for efficiency from those who build strategic capability.

A NOTE ON TIMING

AI is evolving faster than any technology I've integrated into product design teams. These observations originate from hands-on experience bringing AI products to market at NetDocuments and integrating AI tools into design workflows. Capabilities will continue to rapidly evolve, but I’m betting the strategic questions for design leaders will endure.

Right problem: Where AI requires human judgement

The ‘right problem’ stage is the least familiar and most difficult to many organizations. These are questions designers should be asking at this stage:

  • Are we solving the right problem for the business?

  • How do we balance user needs with business constraints?

  • Which problems move the needle of revenue, retention, or other strategic imperatives?

  • What are we choosing not to solve?

This is where business strategy meets design thinking which requires a careful consideration of product context, stakeholder needs, technical constraints, and market dynamics. Understanding that AI makes predictions based on patterns in your data, human judgement is needed when deciding which predictions matter for you.

At NetDocuments, I used ChatGPT to analyze a long list of customer ideas. It was fantastic at converging, categorizing, and ranking ideas, but missed the target when prompted which idea would deliver the greatest impact. ChatGPT lacked broader context. It didn’t know that we had a strategic prospect hammering on a nonexistent feature they wanted before signing, or that a past architectural decision made executing an idea extremely difficult, or that a competitor was generating buzz with a new capability we lacked. Harvard Business Review research on AI strategy confirms this pattern: organizations that lead with AI rather than strategy risk compromising their competitive position.

Understanding this distinction between what AI does well and what requires human judgment has also shaped my approach to building design teams. When hiring, I look for business curiosity, not just user empathy. I structure teams to give designers direct access to strategic discussions to develop context. I develop designers by teaching them to articulate business impact alongside user value. I instill that getting the right problem solved matters more than getting the solution perfect.

Right solution: Where AI helps (and hurts)

Once you've framed the right problem, the next challenge is finding the right solution. This is where AI's benefits and risks both intensify. Traditionally, this phase takes days to weeks depending on problem complexity. AI has compressed this cycle to hours by generating more ideas, faster. If you're already embracing fail fast, learn fast, this generative leap is incredibly powerful: generate options quickly, evaluate against user needs and business goals, choose the best direction to pursue.

At NetDocuments, we had a file-upload UI that lacked queuing. Using the AI-assisted design platform Uizard, I generated 30 variations in minutes, including directions I wouldn't have considered. After 30 minutes, I had 70+ options. The volume and speed of ideas was impressive, but therein lies a hidden cost.

In a Wharton study, teams using AI generated far fewer unique ideas compared to human-only groups. When AI trains on existing solutions, it optimizes for patterns already proven to work. Teams using the same models and prompts converge on similar answers. University of Exeter research showed AI increased usefulness scores but decreased variety scores. You get competent solutions faster, but breakthrough thinking requires human direction.

I saw this at NetDocuments. Upon closer inspection, those 70+ Uizard variations collapsed into three identifiable themes containing slight variations. For example, a button would be placed at above the data grid in one variation, then at the bottom in the next - a low-value variation. Still, evaluating each option against business context took time, even when dismissing options at a glance. Volume without discretion leads to the same solutions everyone else ships, just faster.

The Uizard experiment illustrates that design teams need to harness AI's speed without sacrificing originality. As a result, I've evolved how I develop designers to address this. Before opening AI tools, designers must define what they're testing and what success looks like. During reviews, I ask how many unique directions they explored, not how many total variations they generated. When designers show me dozens of AI options, I push them to collapse those into the three or four meaningful differences worth evaluating. Even then, I’ll challenge the rest of the team to propose alternative, novel solutions. The goal is to train designers to recognize when generating more options overwhelm their ability to use good judgement, while fostering originality.

Done Right: Where AI shines

Once you’ve framed the right problem and solution, you need to produce. This is where AI delivers on the productivity gains that make headlines.

McKinsey reports up to 70% faster cycle times for production tasks similar to this phase. At NetDocuments, AI-assisted design tools like Lovable and Bolt in combination with our design system got us to 60% “good” very fast with the designers delivering the critical 40% of detail and refinement in Figma. Using Claude Code, NetDocument engineers performed a series of visual checks between Figma renderings and live code to quickly identify and address inconsistencies which slashed the design QA burden on designers. We had similar experiments running for regression testing, design system consistency, and UI debugging, all designed to proactively address production friction. The opportunities during this phase seem endless.

With production work compressing, strategic work expands. The question for design leaders is how do you reinvest productivity gains?

I now incentivize designers to spend more time on problem framing and less on production. Junior designers used to advance by improving craft, now they progress by demonstrating business understanding. I evaluate them on whether they grasp how their work connects to revenue, retention, or strategic goals. Similarly, senior designers get promoted based on their ability to influence which problems the organization solves.

I've established new ways of working. Pairing a designer with an engineer when using AI-assisted coding tools improved both velocity and quality of output. I've also built safeguards against AI risks like hallucinations by requiring human review of all AI-generated outputs before they reach production.

Understanding where AI accelerates production and where human judgment stays essential has changed which skills I hire for, how I measure performance, and where I focus development efforts. Designers who thrive in this environment spend less time perfecting pixels and more time shaping strategy. Given this massive amount of change, design leaders who fail to redefine success risk leaving their teams unclear about their purpose and value.

How and why this matters now

Production work in the Done Right phase used to consume most of a designer's day. AI has reduced the time needed for delivery tasks, raising the stakes for aligning design activity to business impact in the Right Problem phase.

The decisions leaders make now about AI will define their design organization for years. Once you restructure teams, shift hiring criteria, and automate work, reversing course introduces cost and risk.

I've seen teams get this wrong. A colleague at a sales enablement platform shared they reduced their R&D team by roughly 20% after adopting AI tools, assuming production acceleration meant they needed fewer developers and designers. Six months later, they struggled with strategic alignment. Design focused almost exclusively on execution with little time framing the right problems to solve. Senior designers left, frustrated by their lack of influence and delivering 'me-too' designs. They're now rebuilding at higher cost.

Understanding where AI accelerates work and where human judgment remains essential has changed how I operate in three ways: who I hire, how I measure success, and how designers spend time.

Hire for judgement

Portfolio reviews now focus on problem selection alongside craft:

  • I look for cross-functional experience. Candidates who've worked closely with sales, finance, or support understand business context better than those siloed in design. I ask about their relationships with non-design stakeholders and listen for curiosity about business strategy

  • I show candidates AI-generated design options and ask "what's missing here?" Great answers identify similarities and explain how they'd push beyond obvious patterns

Measure differently

Performance reviews now incentivize strategic thinking:

  • I reduced the weight of "quality of visual execution" as a standalone metric while adding "influence on problem selection" and "alignment of work with business outcomes"

  • Junior designers progress by articulating how their work affects retention, revenue, or competitive position. Senior designers get promoted based on influencing organizational strategy

  • I track how designers allocate time across the three phases: right problem, right solution, done right, looking for a consistent shift left

Make strategic time

Designers spend more time thinking strategy. I estimate good teams now spend 40% of their time on problem framing and business context. Recalibrating expectations led to restructuring calendars to support that shift:

  • I insist each designer attend 1 cross-functional meeting weekly: revenue planning, customer advisory sessions, roadmap prioritization. This builds the context they need to apply judgment to the right problem phase of work

  • I rebalanced critique to focus on problem framing over pixel refinement. The group pressure-tests business assumptions, not just design execution

Tools referenced

Bolt: AI-assisted app development tool
Lovable: AI-assisted app development tool
Uizard: AI-powered UI design tool
Claude Code: AI-powered coding assistant

Takeaway

I've spent over 15 years building design teams that drive positive business outcomes. The key lesson is design activity must connect to business outcomes. AI makes that connection more critical, not less. AI represents a unique opportunity for entire design teams to operate in a more business savvy manner, not just design leadership. Organizations that recognize this now, using AI as a way to expand design’s business impact, will see durable success.

If you’re building that kind of organization and looking for design leadership that thinks this way, I’d welcome a conversation.

All Leadership Essays