AGENTIC AI | ALTERNATIVE INVESTMENTS | 2026
From AI capability to product value
Research that turned product uncertainty into a clearer value proposition
ROLE
UX Research
TIMELINE
4 weeks
SCOPE
AI discovery &
product strategy
FROM POSSIBILITY TO PRIORITY
AI creates more possibilities. It also creates more product uncertainty.
As AI capabilities expand, product teams face a different challenge than they do with conventional SaaS platforms. The question is no longer simply whether a feature can be built.
Teams have to determine where AI meaningfully improves a user's workflow, which problems justify an AI solution, and what people will trust an intelligent system to do. The range of technically possible use cases can make these decisions harder rather than easier.
A product can search, summarize, recommend, generate, predict, communicate, or act on a user's behalf. But technical capability does not tell a team which of those behaviors matters enough to change how someone works. That makes early research particularly important.
WHAT TO KNOW BEFORE YOU BUILD
Before committing significant product and engineering resources, teams need evidence around a set of questions.
I explored this framework at iConnections while leading foundational research for Violet, the company's first agentic AI product for the alternative investments market.
THE CHALLENGE BEHIND THE OPPORTUNITY
iConnections connects institutional investors across a large private-capital ecosystem, representing $50T+ in capital.
The platform was already effective at bringing thousands of investors together. In a high-trust, relationship-driven industry, the challenge was not creating more connections. It was understanding which connections were actually worth making.
Across 100+ investor conversations, research showed that finding the right opportunities and people mattered more than having more options. That raised a more strategic question: how could Violet, an agentic AI copilot, help the platform better understand what investors were looking for, when it mattered, and who they should connect with?
AI IN A HIGH-TRUST ENVIRONMENT
Institutional investing is a high-stakes, highly relational environment.
Investors may commit significant capital for years. Decisions depend not only on quantitative data, but also on judgment, relationships, transparency, and confidence in the people behind an investment. An AI system that produced the wrong answer, or an answer users could not verify, could quickly lose credibility.
The research was designed to reduce uncertainty around what Violet should do, where it could add value, and what the team should prioritize.
FROM AI CAPBILITY TO DIRECTION
Violet copilot opened up many potential directions.
It could help investors: discover managers and allocators, research investment opportunities compare funds, prepare for meetings, search investment documents, generate target lists draft outreach, identify investment intent, recommend connections, eventually take more autonomous actions.
All of those were technically plausible. That was precisely the challenge. The team needed to distinguish between what AI could do and what was valuable enough to build.
RESEARCH QUESTIONS THAT SHAPED STRATEGY
Rather than testing a predefined set of features, I focused the research on the decisions that would shape the product direction.
As AI capabilities expand, product teams face a different challenge than they do with conventional software. The question is no longer simply whether a feature can be built.
Teams have to determine where AI meaningfully improves a user's workflow, which problems justify an AI solution, what people will trust an intelligent system to do, and how much autonomy they are willing to give it. The range of technically possible use cases can make these decisions harder rather than easier.
A product can search, summarize, recommend, generate, predict, communicate, or act on a user's behalf. But technical capability does not tell a team which of those behaviors matters enough to change how someone works. That makes early research particularly important.
DESIGNING RESEARCH FOR A NON-DETERMINISTIC PRODUCT
Traditional usability testing assumes that participants encounter roughly the same system behavior.
Violet did not behave that way. Two users could enter different prompts, receive different outputs, interpret the answers differently, and take entirely different paths through the experience.
A successful session was not one where the user completed a task. It was one where we could understand whether the AI had produced something valuable and credible enough to enter a real workflow.
To understand where Violet created real value, I combined several research methods.
I used qualitative user interviews to understand investor workflows and decision-making, moderated working sessions built around participants’ own problems, and live field research to observe how Violet performed in realistic conditions.
Together, these methods helped me compare what users said they needed with how they actually interacted with the product, and identify which use cases, trust requirements, and product gaps mattered most.
Beyond clicks and task completion, I studied:
THE BREAKTHROUGH OPPORTUNITY
The research narrowed where Violet could create the most meaningful value.
Rather than expanding the agent across every possible workflow, the strongest opportunity emerged around helping investors discover and evaluate relevant opportunities faster through natural-language interaction.
WHAT MOVED THE NEEDLE
I presented the findings to executive and product teams, helping refocus the conversation on the user problems most worth solving before further investment.
The work clarified where Violet had the strongest potential, what needed to be solved first, and which ideas were less likely to create meaningful user value. That helped reduce the risk of spending product and engineering time in the wrong directions. Instead, the team had stronger evidence to focus on the use cases that mattered most and the foundations required to make them work.
The research also surfaced a more important strategic advantage: Violet’s value was not simply the AI model itself, but the proprietary data and investor activity the platform could bring into the experience.
AI+UX RESEARCH LEARNINGS
AI research is often framed around a broad question: Do users trust AI? I think the more useful question is: What exactly are we asking people to trust it with?
Trust changes with the task, the quality of the underlying data, the consequences of inaccuracies, data transparency, and how much automation is needed. The same person may trust AI to explore options, use it more carefully with complex information, and still want to make the final decision themselves. Those boundaries are not just UX details. They help define the product itself.
That is why I see UX research as especially valuable before an AI use case becomes a roadmap commitment. It helps teams distinguish:
- What is possible from what is valuable
- What is useful from what is trustworthy
- Helpful assistance from unwanted autonomy
- A compelling demo from something people will actually adopt
That is the difference between building impressive AI and building something people will trust, adopt, and return to.