Build Agents Faster + Smarter

With Generative AI promising intelligent agents in enterprise software there is a significant acceleration in the desire of businesses to automate processes and augment, or even replace, human decision-making. As companies race to implement agents with their promise of speed, scale, and efficiency, there's a risk of prioritizing problems based on their ability to be solved with this technology, rather than on their impact on the business. I am reminded of the underlying difference in how designers and engineers react to problems: an engineer’s reflex is to solve the problem; a designer’s is to understand it.

Problems in the enterprise—like heterogeneous or misaligned workflows, low user adoption, or fractured customer experiences, are fundamentally human in nature. They stem from mismatches in the expectations between systems and people. This is where design’s reflexive desire to understand the problem demonstrates its value, not as a service to engineering, but as a strategic function to determine what agents should be deployed, why they should exist, and how their success should be measured.

Given they are so very human, solving problems requires more than the effective application of a technology.

While engineers focus on solving the problem, designers seek first to understand the problem—deeply and holistically. That understanding starts with an objective appreciation of the work system: the complex interplay between people, tools, and the organization. Designers investigate the interdependencies across systems and stakeholders; surfacing not just how work is supposed to happen, but how it actually happens, roles & responsibilities, informal norms for communication and shared situational awareness, power structures and social dynamics, and of course legacy tools. These insights allow designers to define clear boundaries for intervention: what is in scope, what lies beyond it, and how to best gracefully integrate solutions to augment, enhance, and empower rather than disrupt, distract, and frustrate.

Too often agents are described based on the promise of their capabilities with little consideration to their boundaries. Likewise, their approach to the problem, reasoning, planning, etc, are not transparent, making it difficult to assess if the user is using the correct agent (assuming they have a say), if the agent is solving the right problem, or its solving the problem it in the right way. This can also result in the inability for the human operators to taking time action to correct any errors. Taken in collective, this prohibits create clear benchmarks to assess if the agent(s) are performing as expected.

If Engineers are bridge builders, then Designers are city planners.

Before considering if an agent is the answer, you should first ask; What is actually broken or needs fixing? Whose is impacted? What behavior(s) are we hoping to enable, change or reinforce? What is the experience required to make that happen? Without answering these questions, any investment in building agents is at risk; no one can afford expensive novelties that don’t deliver customer value.

When used strategically, design functions as a translator between a business’ ambition and their customer’s lived experience. When enterprises set out to deploy agents, they often begin with goals like reducing costs, improving responsiveness, or increasing throughput. These are all valid objectives, but they’re insufficient.

Design reframes these goals by identifying the human connections, a well-placed agent doesn't just automate a step—it alleviates a burden, amplifies a decision, restores clarity or empowers creative problem solving. Successfully understanding the problem, includes knowing where agents should (and should not) intervene. Rather than scattering automation across the organization, designers help identify the key moments where human experience and system logic intersect—moments of handoff, insight, escalation, or exception. These inflection points become the design surface for agents: areas where their intervention can meaningfully reshape the outcome, not just speed it up. Design maps these interactions across roles and journeys, aligning them with the company's broader vision for operational excellence and customer engagement. This shift in framing the strategy based on the problem, not the technology, changes both how a business plans their automations and defines success.

Prototyping is where this strategic framing becomes tangible. Prototypes have long served as tools of inquiry, allowing teams to poke at assumptions, navigating the ambiguity and complexity that characterize enterprise-scale problems. They expose interdependencies between users, workflows, and technologies—allowing stakeholders to assess desirability, feasibility, and viability in parallel. Crucially, Designers are now using GenAI development tools to build functioning code-driven prototypes in hours or days, making experimentation even faster and more cost efficient. Teams can now surface and resolve issues early before diving into costly development, thereby reducing risk and decreasing time to deployment.

Designing Metrics that Matter

When it comes to measuring success, traditional KPIs—like task completion rates or average handle time—fall short. They capture motion, not meaning. While engineering might measure performance in terms of latency or model accuracy, design adds the human lens: comprehension, confidence, perceived helpfulness, and emotional tone. Design introduces experience-based metrics: trust in the agent’s output, time-to-understanding, reduction in errors caused by confusion, and even the emotional tone of interactions.

These are not soft metrics; they are leading indicators of adoption and sustained value, of whether agents are actually being used, trusted, and delivering value. An agent that is fast but confusing will be bypassed. One that is accurate but context-blind will be mistrusted. Experience metrics give enterprises an early signal of misalignment between intent and impact. Design makes these dynamics visible—and addressable. Design also plays a critical role in defining the right success metrics, helping avoid the trap of optimizing for what’s easy to measure rather than what truly matters.

Organizations need to rethink how they value design. It’s not about wireframes or interfaces. It’s about shaping the questions that guide investment.

Deliver effective results, enterprise’s AI strategies need to be intentionally designed from the start, not just engineered. That means ensuring that the human problem to solved in well understood in the initial scoping of an agent capabilities. An anthropological-like understanding is required to optimize the data selection and model training decisions, and in the definitions governance frameworks that ensure responsible behavior. It also means partnering with engineering to set constraints that prioritize user well-being, not just system throughput. Design doesn't just refine what’s already been built—it determines what’s worth building in the first place.

This shift requires organizations to rethink how they value design. It’s not about wireframes or interfaces. It’s about shaping the questions that guide investment. As enterprises deploy agents that learn, adapt, and act on behalf of humans, they must also invest in understanding human complexity. Designers are not just decorators of AI—they are its conscience, its cartographers, and its co-pilots.

Summary

Problems are fundamentally human. Nature doesn’t have “problems”; Landslides, floods, fires, only become problems if they disrupt humanity’s plans. More often than not in the enterprise problems are defined by the ability to monetize their solution, and the bigger the economics, the greater the pressure to deliver an easy solution. However, given problems are so very human, solving them requires more than the effective application of a technology. It requires a holistic understanding of the problem and the interdependencies between people, processes, and social systems. The cost of discovering unanticipated interdependencies grows exponentially the further you go into the development cycle—either in rework or descoping your solution. Design is an extremely cost-effective investment to ensure that the right solution is being built from the get go.

In the end, agents are only as useful as the problems they are trusted to solve. And trust is not a byproduct of technical correctness—it is earned through clarity, reliability, and empathy. These are design concerns. For enterprises to truly succeed with agents, design must not follow the strategy. It must be part of its definition.