C
COCREATE
Adaptive AI Interviewing at Survey Scale
Company
CoCreate
My role
Founder · System Architect ·
Design Lead
Overview
Why learning speed, not research quality, was the real problem
Product teams weren’t failing to research users. They were failing to learn fast enough to inform decisions. I initially built a system to accelerate analysis of qualitative research data, but early validation showed that analysis wasn’t the bottleneck—learning speed was.
Cosmic Unity 3 N7
Basketball Shoes
1 Colours
US$170
Nike Benassi N7
Slides
1 Colours
US$35
Nike Sportswear Club Fleece N7
Pullover Hoodie
5 Colours
US$60
Nike Sportswear Club Fleece N7
Joggers
5 Colours
US$60
N7 Collection
Home
Shop
Favourites
Bag
Profile
9:41
All
Tops & T-Shirts
Hoodies & Pullover
9:41
What are the first things you notice in this?
There were some things that seemed a little off
I don’t know if this is clothes for women or if I can buy all sizes here?
Problem
Insights arrived after decisions hardened
As product velocity increased, insights consistently arrived after decisions hardened. Interviews delivered depth too late. Surveys delivered speed without understanding
Core Tension
Depth and scale broke down at product speed
Scaling research traditionally sacrifices adaptability, while preserving adaptability requires more researchers. Both paths broke down at speed.
The Trade-Off
Surveys = Scale
Interviews = Depth
Alternative?
None
Strategy
Redirected the system for faster analysis to adaptive interviewing
After validating that faster synthesis didn’t meaningfully change decision-making, I redirected the system to scale interview adaptability instead—redesigning how learning happens before decisions harden.
Previous product direction prioritized qualitative data analysis
Solution
AI-assisted interviews with human-controlled intent
An AI-assisted interviewing system where researchers define intent and boundaries, and AI conducts adaptive interviews within those guardrails—extending reach without replacing judgment.
External Validation
Accepted into a Carnegie Mellon Univeristy Incubator
The system concept and research direction were accepted into an external incubator with Carnegie Mellon University.
Risk Accepted
Optimized for trust over raw accuracy
We accepted the risk that adoption would hinge on trust rather than accuracy, and designed the system accordingly.
Outcome
Enabled learning before roadmap commitment
Validated that adaptive interviewing can deliver qualitative depth at survey scale, enabling teams to learn at the pace of product decisions. Teams could test positioning, messaging, and product direction before committing roadmap decisions.
Being able to test the interview chat before sharing built trust
My Role
Founder-level system design and execution
I owned the end-to-end design of the system — from problem framing and research methodology to interaction design, product strategy, and iteration.
I made foundational decisions about where automation should assist versus defer to human judgment, and how learning could adapt without undermining trust.
You mentioned you don’t really know what it is, could you elaborate on that?
It isn’t clear where on the homepage I am
supposed to find the checkout button
Adaptive interview prompts generated in real time, with researchers retaining control over intent and follow-up.
Impact
Clarified the limits and opportunities of adaptive learning systems
Demonstrated that adaptability mattered more than raw analysis speed
Identified trust and timing as the primary adoption constraints
Informed how AI systems must remain human-directed to influence decisions
Reflection
Learning Must Happen Before Commitment
This work reinforced that the hardest problems in product aren’t always solvable with features. Some require changing when teams learn, what they trust, and how much uncertainty they’re willing to hold before committing.
While the system didn’t reach broad adoption, it clarified the real constraints around adaptive learning — insights that continue to shape how I approach research, AI, and decision-making systems.
CEO @ RJ Ventures
Next Project
Buidling a Product Learning System That Turned Guesswork Into Confident Decisions
Go to Next Project
Take me back home.
