Founding Designer - Building a B2B, AI product from 0-1
Led to a $5m seed funding round, attracted design partners, landing our first customers, 80% move to book 2nd meeting after demo
My Role
- Product Research
- Product Design
- Business Strategy
- Product Strategy
Timeline
2022 - Present
Team
Company size
9 people
Context
Korl's AI integrates deep context about customers' business, product usage, and key account data, generating branded, customized presentations
Problem
Most customer presentations are templated, rushed, and disconnected from the customer, the strategy, and the outcome.
Customer-facing teams (CS, Sales, Marketing, Execs) spend hours gathering data, organizing insights, and building decks - only to create something generic or misaligned. That creates risk and friction and it costs deals and renewals.
Research
Talking to 40+ GTM professionals and the people they sell to
I needed to validate whether GTM teams actually needed better communication tools—and what would make and what would make their customers (the buyers) respond.
Designed a dual-sided research framework
Activities:
Interviewed both GTM teams (our users) and their customers (buyers)revealing the critical disconnect between what teams prioritize and what actually drives purchases.
Defined what we needed to learn before drafting scripts—prevented confirmation bias
Activities:
Draft and share "what do we need to learn" and "how are we going to learn it". This helps prevent confirmation bias.
Established research rigor and guardrails
Activities:
Drafted research timeline, interview scripts, and how we would synthesize and prioritize the research.
Built synthesis frameworks to identify patterns vs. outliers
Activities:
Setup a process for synthesizing research and sharing with the team.
Designed a dual-sided research framework
Activities:
Interviewed both GTM teams (our users) and their customers (buyers)revealing the critical disconnect between what teams prioritize and what actually drives purchases.
Defined what we needed to learn before drafting scripts—prevented confirmation bias
Activities:
Draft and share "what do we need to learn" and "how are we going to learn it". This helps prevent confirmation bias.
Established research rigor and guardrails
Activities:
Drafted research timeline, interview scripts, and how we would synthesize and prioritize the research.
Built synthesis frameworks to identify patterns vs. outliers
Activities:
Setup a process for synthesizing research and sharing with the team.
Research Takeaways
From Research to Product Vision
Based on 40+ interviews, I identified the core problem: CS teams spend about 8 hours preparing for each customer meeting, manually gathering data from 2-5 tools and translating product updates into customer-relevant language and building presentations.
I synthesized this into a design hypothesis:"If we connect to customer data sources and use AI to generate contextual presentations, we can reduce prep time by 80% while increasing personalization."
This became our north star for design decisions.
Product data is messy and scattered across different tools
Product and GTM teams speak different languages
Customization takes hours per presentations
CSMs can't scale to all customer meetings
Key Insights
- GTM teams optimized for volume and speed.
- Buyers valued relevance and personalization.
This gap became Korl's core differentiator.
Impact
- Validated the hypothesis and product direction.
- Identified lack of personalization at scale as the #1 barrier — became our primary value prop.
- Established research framework the team still uses for ongoing validation.
"My objective would be to reduce the time that it takes my team to have quality conversations with customers so they can have more of them"
Solution
Personalized, on-brand presentations with real business context, not generic templates.
I designed an AI presentation creation tool that builds the right content for the right conversation - helping CS and Sales teams communicate strategically, and communicate with confidence.
Data Integration & Mapping
Design: A data mapping system that uses AI to auto-map fields and values from other tools to Korl's data model and provided contextual triggers to edit when needed.
Validation: "With the context of product tools, KorlAI achieved 90% accuracy of fields and values mapped, and near total reduction in manual effort".
Content Generation & Translation
Design: An AI translation layer that augments and generates key customer-facing content like feature names, descriptions, and impact statements.
Validation: "Translates 100% of features, ~1 hour saved per presentation. total consistency of messaging".
Personalized feature selection
Design: A system that auto-selects the most relevant features for the presentation based on the customer's business goals, personas, and the type of presentation.
Validation: "increased accuracy of feature selection".
Presentation Assembly
Design: A presentation builder tool that auto-assembles presentations based on the customer's business goals, personas, value prop and market positioning.
Validation: "95% reduction of manual effort, time to create presentation reduced from hours to seconds".
Usability testing & Iterations
Prototyping with Figma MCP and Claude Code
I created several prototypes for different workflows, using Figma MCP and Claude Code to rapidly prototype and test with users. This is the fastest and most realistic way to test the product before we spend resources to build it. User testing with realistic prototypes taught us a lot about what to built and saved time over traditional prototyping.
Iteration #1
Data management with variables
Problem Discovered: During regular product usage, I noticed users were editing content that Korl populated in presentations. This breaks sync with the data source and the 'single source of truth' we created in the Feature Detail pages.
Solution: Create variables for all tracked data in Korl, populate content as variables in the presentation - allowing full control of content from anywhere in the application without breaking the 'single source of truth'.
Design Tradeoff: Increased product complexity. Alternative was to limit to source data only (simpler but less useful). I chose flexibility because power users were already editing content in the presentation.
Validation: Breaking data sync occurences reduced, and users are able to edit content in the presentation without breaking the sync with the data source.
Iteration #2
Auto-apply customer branding
Problem Discovered: During testing and customer demos I heard repeatedly that generated presentations didn't match brand guidelines. Colors were close but not exact, fonts were similar but wrong weights, and placement of elements was off.
Solution: Customers upload a PDF of a current presentation and KorlAI crawls the PDF to understand the brand and applies the colors, fonts, and placement of elements to the presentation.
Design Tradeoff: Small increase in setup time for customers, but a big increase in brand consistency and alignment.
Validation: Time spent edting brand elements in the presentation is reduced, and the presentation is more consistent with the customer's brand.
Iteration #3
Proactive Presentation Generation and Delivery
Problem Discovered: When talking to customers and propsects I realized that the more we can do to automate presentation creation the more we can free up their time to focus on expanding accounts and closing deals.
Solution: I designed a KorlAI agent that proactively generates and delivers presentations for review via email or Slack without any manual intervention.
Iteration #4
Speaker Notes Generation
Problem Discovered: We heard from customers and prospects that consistency of messaging was an issue across ther CS teams. We found that customers rarely create speaker notes.
Solution: An agent that auto-generates speaker notes for the presentation based on the presentation content and the customer's business goals, personas, value prop and market positioning.
Validation: 100% coverage of speaker notes, and consistent messaging.
Other Foundational
Design Work
Brand Identity
I defined and created the company's brand identity, translating the concepts of trust and thought leadership into a visual language.
Content Design
Once I understood our core value proposition and our target personas, I was able to design our messaging, and create marketing assets and social media posts.
Design System
I started from day one creating components in Figma underpinned by a design token system to speed up design work. I worked closely with engineering to understand the best code component system to use so I could represent that in Figma for a strong connection between Design and Engineering. This laid the foundation for a growing team, and ensures consistency and efficiency in building the UI.
Impact
$5m seed funding round, attracted several cohorts of design partners, 0-1 launch with paying customers, 80% move to book 2nd meeting after demo
Key Learnings
When designing effective B2B, AI tools - context is everything.
It's not enough to rely on a chat interface and the knowledge of an LLM to generate content that provides real value. You must provide it with the underlying business context and allow the AI's analytical power to contextually surface and infer important data. The user should always maintain control of the AI's output.
See more work
Lightstep
Leading design of Deployments and Root Cause Analysis tools
Sisu Data
Managing a team designing complex ML Data Analytics tools