This article was based on the interview with Domo Chief Design Officer Chris Willis on speed vs. trust in product design by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
We all feel the pressure. The mandate to innovate, to integrate AI, to move faster, is no longer a gentle suggestion from the boardroom; it’s a roar. Every conference, every webinar, every vendor pitch is a fresh reminder that if you’re not sprinting, you’re already behind. This has created a frantic scramble to deploy agentic systems and AI-powered tools, often with the primary goal of being able to say, “Yes, we’re doing that.” The unspoken assumption is that speed is the ultimate competitive advantage, and any delay to perfect the foundation is a risk the business simply cannot afford.
But what if the real risk isn’t moving too slowly, but moving too quickly on a foundation of sand? In our haste to show progress, we often overlook a more fundamental and far more consequential element: trust. Not the trust our customers have in our brand—though that is certainly an outcome—but the trust our own teams have in our data. As we’ll explore with insights from Chris Willis, Chief Design Officer and Futurist at Domo, the current AI gold rush is exposing the cracks in our internal data foundations. The truth, unglamorous as it may be, is that AI doesn’t magically fix bad data; it simply inherits, and then confidently amplifies, its flaws. For marketing leaders, this means the difference between an AI investment that compounds value and one that quietly erodes it lies in the foundational work most are tempted to skip.
The Inheritance Problem: AI Gets Its Confidence From You
The conversation around AI and data often starts with a focus on external applications—customer-facing chatbots, personalization engines, and dynamic ad creative. Yet, the initial and more critical challenge is internal. For decades, organizations have operated with a certain level of data ambiguity. Different departments maintain their own versions of the truth, often reconciled in the eleventh hour before a quarterly business review. These “shadow spreadsheets” and unwritten rules have been a functional, if inefficient, part of business.
AI changes the stakes entirely. It doesn’t understand nuance or historical context; it takes the data it’s given as gospel. When that data is inconsistent, the AI becomes a powerful and articulate purveyor of misinformation. This isn’t a theoretical problem; it’s showing up in organizations that have rushed to implement AI on top of a shaky data infrastructure. The confidence of the AI’s output belies the uncertainty of its input, creating a dangerous illusion of accuracy.
“AI inherits whatever confidence you already have in your data. The challenge, of course, is that data may not be super confident, but it sounds really confident when you run it through some sort of agentic system.”
For marketing leaders, this is a critical warning. Imagine an AI agent tasked with optimizing campaign spend. If it’s fed conflicting data about customer lifetime value from the sales and finance departments, it won’t pause to ask for clarification. It will make a confident recommendation based on the data it happened to prioritize, potentially shifting millions of dollars in budget based on a flawed premise. The same goes for personalization; an AI personalizing user experiences based on incomplete or contradictory data isn’t creating a meaningful connection, it’s just executing a flawed strategy at an unprecedented scale. The first step in building a trustworthy AI strategy is to confront the uncomfortable truth about your own data culture.
You Can’t Engineer Trust, But You Can Engineer For It
Faced with this challenge, the natural inclination for many leaders is to demand a solution: “How do we engineer trust into our systems?” It’s a reasonable question, but it’s aimed at the wrong target. Trust isn’t a feature you can add or a line of code you can write. It’s an outcome, an emergent property of a system that is designed and operated with intention. Attempting to build “trust” directly is like trying to build “happiness”; you can’t assemble it from parts. Instead, you must build the conditions from which it can grow.
Willis proposes a more pragmatic and actionable framework. Rather than chasing the abstract concept of trust, leaders should focus on engineering for its core components: visibility, predictability, and control. These three pillars provide a tangible blueprint for creating systems that people can learn to rely on, not through blind faith, but through verifiable experience.
“Trust is not something you can engineer for easily. So, my approach, and the approach we take is, at Domo, is we engineer for three things. Visibility, predictability, like we just talked about, and control.”
Let’s translate this for a marketing organization.
- Visibility means your team can see what the AI actually did. When a machine learning model creates a new high-value audience segment, can your team easily see the data points and logic it used? Or is it a black box that spits out a list of names? True visibility allows for inspection and understanding, which is the first step toward confidence.
- Predictability means the same inputs produce the same, or at least directionally similar, outputs. If you run the same customer data through a segmentation model on Monday and again on Tuesday (with no new data added), you should expect a consistent result. An unpredictable system, where the AI “improvises” differently each time, is impossible to manage or scale.
- Control is the essential human backstop. If an AI-powered pricing engine suggests a promotion that wildly deviates from brand guidelines, can a human easily intervene, stop it, and adjust the parameters? Control isn’t about micromanaging the AI; it’s about having the final say and the ability to course-correct when a model drifts from your strategic goals.
By focusing on these three elements, you shift the conversation from the philosophical to the practical. You give your teams the tools they need to build a working relationship with the technology, one based on evidence rather than hype.
The Hidden Cost of Speed: The Verification Problem
One of the most seductive promises of generative AI is its ability to produce content and answers with breathtaking speed. With a simple prompt, we can generate a dozen landing page headlines, a social media campaign calendar, or a complex SQL query to analyze performance. The “effort to generate” has plummeted. However, this has created a new, and often overlooked, bottleneck: the “cost to verify.” Is the output accurate? Is it on-brand? Is it compliant? Is it even good?
This asymmetry between generation and verification is a critical trap. We’ve become so enamored with the speed of creation that we’ve underestimated the human effort required to ensure quality. In the past, the natural friction of the creative process—the time it took a copywriter to write, an analyst to code, a strategist to think—served as a built-in verification layer. Ideas were refined and vetted before they were ever produced. AI often inverts this process, giving us a “magic answer” first and leaving us with the laborious task of deconstructing it to see if it’s right.
“On your X axis, think of it as effort to generate something…On the vertical axis, think of that as your cost to verify…what you wanna do is you wanna think about, all right, let’s automate things that can be verified easily, and then things that require…are slower and more costly, those are where human judgment come in.”
This framework is a powerful lens for any marketing leader deciding what to automate. Automating the creation of meta descriptions based on page content? Relatively low cost to verify; a human can quickly scan them for sense and accuracy. A good candidate for automation. Asking an AI to write a public-facing statement on a sensitive brand issue? Extremely high cost to verify; it requires deep contextual understanding, legal review, and brand judgment. This is where human expertise remains paramount. The undocumented judgment of your experienced team members—the knowledge of what “feels right” for your brand and your customers—is the most expensive thing to replicate and the most dangerous thing to ignore. The smartest AI strategies aren’t about automating everything, but about deliberately choosing to automate the tasks that can be easily and affordably verified.
The Path to Agility is Paved With Unglamorous Work
The pressure to act on AI is real and justified. This technology represents a fundamental shift in how we will operate. However, the prevailing narrative that speed is the only metric that matters is a dangerous oversimplification. The path forward isn’t a mad dash toward “moonshot” projects built on shaky ground. It’s a more deliberate, focused approach that begins with the unglamorous, non-negotiable work of building a governed data foundation. As Willis’s insights reveal, the most successful organizations aren’t the ones launching the flashiest AI tools first; they are the ones methodically building internal trust in their systems by focusing on visibility, predictability, and control.
For marketing leaders, this means shifting the focus from lagging indicators like immediate ROI to leading indicators like internal adoption. Encourage your teams to start small, to build focused tools that solve a single, nagging problem, and then watch to see if people actually use them. That organic adoption is the clearest signal you’re on the right track. It proves you’re creating real value, not just chasing a trend. This approach requires patience, a quality in short supply right now, but it’s the only way to build a sustainable advantage. True agility, after all, isn’t just about moving fast. It’s about having a foundation so solid that you have the confidence to change, adapt, and move in the right direction, again and again.
