How we used AI to remove friction inside a complex B2B product
This case study is password protected
Role
Product Designer
Scope
AI features, UX, Interaction design
Type
B2B SaaS 500+ users
Context
Logic Builder is a no-code platform for designing automated customer communication flows without engineering support. Users connect nodes on a canvas to build automated customer support and communication workflows.
The product was originally built by engineers for internal use, but later evolved into a customer-facing B2B SaaS platform used by non-technical operations and support teams at large enterprise companies.
While working on the product, two things happened in parallel. The product grew — users created more flows, flows got bigger, and environments started holding hundreds of them. At the same time, our LLM team was rapidly expanding what was technically possible inside the product.
Across our work with the operations team, we kept seeing the same kinds of friction. Users struggled to start new flows from scratch. They opened large existing flows and couldn't tell what they did. They knew something was broken but didn't know how to fix it. They couldn't find the flow they needed in environments with 100+ of them.
These were not new problems. They had been with the product for a long time. What changed was that AI gave us a real way to solve them.
Not as one AI assistant, but as a set of tools designed to solve different user problems.
This case is about that work: how we approached AI as a set of small, focused tools matched to different moments where users got stuck.
My goal was to identify where AI could genuinely help users — and turn those moments into practical product solutions.
Key achievements
+30%
First flows created
during the first 7 days
-50%
Dependency on repeated onboarding calls
-35%
Operational onboarding load on the team
+30%
First flows created during the first 7 days
Based on operations team tracking and product analytics, 3 months post-launch.
What we learned before starting
Before deciding where AI could help, we mapped out what we already knew about users — from operations team feedback, walkthroughs, and day-to-day work with customers.
To structure this, we looked at several user scenarios — new users building from scratch, new users updating flows, experienced ops working at scale, experienced ops maintaining other people's work.
Each scenario had its own problems, but the same ones kept showing up.
Customer journey maps for different user-task combinations.
Learnings
The same four friction points kept showing up across all scenarios.
1
Creating flows took a lot of manual work
Every node configured by hand. Slow for new users, repetitive for experienced ones.
2
Understanding existing flows took time
Users often had to trace through dozens of nodes.
3
Fixing issues slowed people down
Small issues could take time to find and fix.
4
Finding the right flow wasn't always easy
Users often scrolled through long lists trying to find what they needed.
How we decided where AI could help
Once we understood the main friction points, we brought together design, engineering, and the LLM team to explore where AI could help.
The goal wasn't to add AI for the sake of it. We already knew the problems. The question was whether new AI capabilities could help solve them in a simpler or faster way.
We reviewed the biggest pain points — manual flow creation, hard-to-read existing flows, slow error fixing, and difficult discovery — generated ideas, discussed technical limitations, and prioritized the concepts that felt both valuable for users and realistic to build.
Impact vs Effort matrix mapping the AI ideas we considered — including some we didn't move forward with.
The matrix helped us identify which ideas offered the strongest combination of user value and technical feasibility. From there, we focused on four opportunities that balanced user impact, implementation effort, and long-term product value.
What we built
Each decision addresses a different version of the same problem — the fear of starting in a complex product.
1
Create Flow with AI
Start Faster
Generate workflow drafts from a simple prompt.
2
Explain the Flow
Flow Clarity
Turn complex workflows into easy-to-follow steps.
3
Fix Errors with AI
Faster Fixes
Spot problems and guide users through resolving them.
4
Smart Search
Find Faster
Find workflows by intent, not exact keywords.
1
1
Create Flow with AI
Starting a new flow from scratch was the hardest moment. New users faced an empty canvas and dozens of nodes without knowing where to begin. Experienced ops rebuilt similar flows over and over. The question wasn't whether AI should help — it was how much of the process it should handle.
Two approaches
Step-by-step — AI guides users through questions, building the flow one decision at a time.
AI Copilot guides users through multiple questions before generating a flow.
2
2
AI voice-guided onboarding
I broke onboarding into smaller contextual steps, so users got help when it became relevant instead of all at once. Users could choose how to start: guided tour, AI-assisted build, or self-exploration.
The first version used text-only tooltips, but users skipped them. The issue wasn’t the content — it was attention. So I added AI voice narration with a Mute toggle. The content stayed the same, but voice made users more likely to follow the onboarding instead of skipping through it.
Voice guidance wasn’t part of the original scope, and the client didn’t have budget for voice production. I tested AI voice tools, built a working prototype myself, and helped the team ship it without external production.
3
3
Build with AI Copilot as a starting path
Even after onboarding, the empty canvas was still a barrier. Users understood the system, but many still didn’t know what to build first.
The product already had an AI Copilot inside the builder, but it was treated like a secondary feature. I moved it earlier in the experience and made «Build with AI» one of the main starting options alongside «Take a tour» and «Explore on my own.»
Instead of starting from an empty canvas, users started with something they could edit and learn from.
4
4
Templates that customers build themselves
The request sounded simple: «add templates.» The real challenge was that every customer worked differently — different workflows, conversation logic, and brand voice. There was no single template that would work well for everyone.
Instead of creating templates for customers myself, I pushed for giving teams a way to create and save their own inside the workspace. One person builds a useful flow, the next person starts from it instead of starting from scratch.
It didn’t solve the first-time user problem immediately, but over time teams stopped facing an empty canvas every time they created a new flow. Knowledge created by one person became reusable for the rest of the team.
Producing the motion myself
Animated onboarding usually requires a separate production pipeline — motion designers, voice actors, editors. We didn’t have that setup, so I built the workflow myself using lightweight tooling and AI voice generation.
I prototyped the motion in Figma, combined it with AI voice and sound effects, then exported everything directly into the product and knowledge base.
The setup turned out faster than a traditional production process. When the UI changed, onboarding could be updated the same day instead of waiting for a full production cycle.
REFLECTIONS
The hardest part wasn’t designing onboarding screens. It was realizing the problem wasn’t a lack of explanations — the product exposed too much complexity too early.
Instead of layering more onboarding on top, I focused on reducing friction directly in the product: simplifying configuration, guiding attention, and helping users start with something editable instead of an empty canvas.
One thing I’d improve in the future is onboarding measurement. We didn’t have clear tracking for where users dropped off, so many decisions relied on observation and support feedback instead of behavioral data.
This wasn’t an onboarding project. It was a complexity reduction project applied where users meet the product for the first time.
Remove the human dependency.
Let the product do the teaching.
Collaboration: Operations team (weekly syncs, 3 collaborators). Engineering (active partners — feature scoping, implementation, and walking me through how the underlying systems worked). Hallway testing across departments. Design decisions: mine, in coordination with the design team.
CV