.png){kind=logo}
Before Adrian Aliu began setting product priorities for AI systems, he was thinking about cities, land use, and legal reasoning. His interests in urban planning and law trained him to work through problems shaped by rules and competing constraints. Today, as cofounder and Chief Product Officer at Proper Motion, Aliu applies that way of thinking to a different challenge: understanding when an AI agent technically completes a task but still leaves the customer dissatisfied.
“What interested me about law was the reasoning,” Aliu said. “You have to understand the rules, interpret the situation in front of you, and be precise about how you frame the problem. I found that the same habits became useful when I moved into AI product work.”
Aliu’s analytical interests surfaced early. At 20, he placed second in a legal writing competition intended for practicing lawyers, and he was admitted to Harvard Law School on a deferred basis while still an undergraduate. The deferral gave him room to explore the rapid development of generative AI instead of moving immediately into law.
“I had a window where I could take a real swing at something,” Aliu said. “AI was moving quickly, and I wanted to understand it by building products and working with customers rather than watching the change from the outside.”
At Proper Motion, the company is building continual learning for AI agents. Its platform is in private beta with ten customers, giving Aliu and the engineering team a group of active users whose agentic platforms expose problems that are difficult to identify through conventional monitoring alone.
Proper Motion also participated in a16z Speedrun and received over $500,000 in backing from the accelerator. Aliu directed development of the product demo submitted with the application and shaped much of the narrative and branding used to present the company and their platform.
Aliu’s responsibilities put him directly between customers and engineering. He conducts sales calls, studies how customers use Proper Motion, and brings what he learns into product-roadmap decisions. He has also worked from customers’ offices in a forward-deployed capacity, helping configure the platform around their agent improvement stack and seeing firsthand where an implementation creates friction.
“My job is not simply to collect requests and hand them to engineering,” Aliu said. “I am trying to understand what the customer is really reacting to, whether that points to a product problem, and what we should prioritize because of it.”
Proper Motion’s private beta gives Aliu access to adverse agent behaviors that do not always look like conventional software failures. A page may not break, an expected action may still happen, and no error message may appear. An agent can complete a task while still frustrating the person using it.
A customer may repeatedly clarify instructions before an agent finishes a workflow. An answer can be correct but unnecessarily long. In other cases, the agent may mishandle a task once and cause the customer to stop using that workflow altogether. Changes in tone, repeated corrections, or abandoned use can reveal dissatisfaction even when the software registers the interaction as successful.
“You can have an agent that technically succeeded and still created a bad customer experience,” Aliu said. “That is harder for an engineering team to see because there may be no error message telling them something went wrong.”
Proper Motion is being developed around those less obvious failures. The product focuses on nondeterministic agent failures and is designed to let teams analyze traces while limiting exposure of customer or user personally identifiable information. Proper Motion is also using reward models to examine how agent behavior connects with customer-success signals such as usage and churn, allowing teams to look past making technical improvements to their product for the sake of maximizing benchmark performance, instead focusing on changes that actually drive positive business outcomes.
Sales conversations reveal how companies describe their needs before and during implementation, while actual product behavior can show where users struggle once an agent is operating inside a live workflow. The Proper Motion platform Aliu is developing distinguishes the signal from the noise: deciding which agent behaviors justify a product change and which problems deserve priority on the roadmap.
“Customer language can tell you a lot about where the product needs to go,” Aliu said. “Sometimes the important part of a sales conversation is not the feature someone asks for. It is understanding the problem behind the request and deciding what that means for what we build.”
Proper Motion cofounder Rafit Jamil brings related research experience to that problem. He previously worked on applied research involving user reward models at OpenAI, focused on determining whether users were satisfied with ChatGPT responses based on behavioral signals. Proper Motion is applying these insights to vertical agentic AI products, where dissatisfaction may develop across several interactions rather than appearing as one clearly incorrect result.
Aliu’s contribution is centered on building a platform informed enough to decide what those signals should change. The private-beta customers give him a concrete group from which to learn what creates value, what causes friction, and which problems are significant enough to affect continued use in different verticals; with the eventual goal of being able to proactively anticipate the most common types of agent failure when a new customer in a familiar vertical is onboarded.
A technically successful interaction can still damage the customer experience, and a simple complaint can expose a deeper product issue. Proper Motion’s role is to identify that difference early enough for Proper Motion’s engineers to act on it.
“AI products create a different kind of ambiguity,” Aliu said. “The challenge is figuring out what the user’s behavior is telling you, then turning that into something the product and engineering teams can actually act on.”