MIKE OICHERMAN
PRODUCT DESIGN · RESEARCH · AI-ASSISTED PROTOTYPING & BUILD
COLLECTIONS
Back
An AI-generated collection builder: it drafts, structures, and organizes content into a document you can then rearrange and edit.
Role Product Designer Consultant
Timeline Q4 2025 - Q1 2026
Company Guidde
The Problem
Enterprise teams accumulate knowledge across dozens (or hundreds) of individual guiddes, but that knowledge stays scattered, there was no way to combine multiple guiddes, videos, and outside resources into a single structured, scalable learning experience. Onboarding a new hire or ramping a team on a new tool meant hunting down and manually sequencing content. Collections gives teams two paths to solve this: describe a goal and let AI assemble a structured collection automatically, or build one manually with full control.
Goals
I was brought on as a consultant to design Collections from the ground up. The client had already completed preliminary research to identify product-market fit; I was responsible for designing the MVP, researching and testing on prototypes, and creating the new interface and components needed to support an LLM-driven experience; all while staying consistent with Guidde's existing design language.
OBJECTIVE 1 Design an MVP for Collections, from initial flows through testable prototypes.
OBJECTIVE 2 Introduce new interface patterns and components to support an LLM-powered creation experience, without breaking from Guidde's existing design system.
Research & Findings
GoalsValidate the assumptions taken in the MVP while exploring additional features and potential gaps in the experience.
Process
Existing research review Collected and synthesized information already gathered by CS and client-side managers on user needs and product fit.
10x Internal Interviews Conducted with key stakeholders to pressure-test MVP assumptions and surface gaps not covered by the client's preliminary research.
Competitive Analysis Investigated how competitors differentiate themselves and identified potential feature gaps in our offering.
Key Findings
The existing "playlist" solution didn't meet user needs for organizing more complex guiddes; a flat sequence of videos wasn't enough structure for real learning paths.LLM capabilities could significantly reduce the time it takes to generate a collection, compared to manually assembling and sequencing content.The experience around creating and editing a guidde more broadly could benefit from an update, collections surfaced this as an adjacent opportunity, not just a standalone feature gap.
Prototype
PlanningWith MVP assumptions and research findings in hand, I moved into prototyping two distinct flows: the creation side and the consumption side, to test them independently before merging into a single cohesive product
Prototype CreationI developed two high-fidelity prototypes: one focused on the create flow (how users generate a collection) and one focused on the consume flow (how end users ingest and navigate a finished collection).
How users generate a document Testing an AI-generated structure against a more manual, block-based approach.
How end users consume and navigate the finished collection A part of the experience easy to overlook when most of the design attention goes to creation.
Testing & Validation
Method15x live user interviews, conducted with a pool of various guidde.com users interacting with the prototypes and sharing their thoughts and feedback.
Key Findings & Iterations
Generating from scratch is hit or miss. Users wanted the ability to provide external context, resources and existing guiddes, rather than generating a collection from a prompt alone.
A natural fork in preference emerged. Some users preferred generating collections via LLM prompting, while others preferred a more manual, block-based node approach; the product needed to support both rather than force one path.
Segment-level control, not whole-document control. Users wanted to adjust specific areas of a generated collection without re-prompting and regenerating the whole document.
Consumption is an overlooked success metric. While most of the design emphasis naturally went toward the generation experience, how well users actually consume and navigate a finished collection is just as strong a signal of success, and needed equal design attention.
Design System
Foundation Following the existing Guidde design system the team and I have adopted a block-node framework which I have designed, expanded and maintained.
Coverage Beyond the base block node framework, I defined and created the UX around interacting with LLM interfaces across the whole product.
Live Product View Product