Lleverage, AI Workflow Automation
Lleverage: one front door for every use case
Founding Designer and Head of Design, building brand, product and design research from the ground up for an AI workflow automation platform used by non-technical teams.
- Scope
- Brand, product, onboarding, nav
- Role
- Founding Designer, Head of Design
- Company
- Lleverage
- Year
- Add year
- Status
- Live
Background
Every client, a different problem, one platform
Lleverage’s clients each show up with a genuinely different reason to automate a workflow, and almost none of them are technical. The platform needed workflow canvases, data tables, integrations and onboarding that could flex to every use case without ever feeling like it was built for someone else, all while the category itself, agentic AI workflows, was brand new to most of the people using it.
Process
How the work was phased
Discover
Add timing
Ran focus groups per pivot, understanding non-technical users meeting agentic AI for the first time.
Define
Add timing
Identified the need for one navigation model to serve every client’s use case.
Design
Add timing
Built workflow canvases, data tables, onboarding and the new navigation.
Deliver
Add timing
Migrated 98% of clients to the new agentic navigation model.
Research
Testing every pivot before it shipped
Every pivot got its own focus group before it shipped, since the audience was meeting agentic AI for the first time and had zero tolerance for confusion.
Add count
Focus groups run across product pivots
5
Core product areas redesigned (canvas, tables, integrations, onboarding, navigation)
98%
Clients migrated to the new navigation
Problems identified
What was actually going wrong
Every client, a different use case
No two clients came to Lleverage for the same reason, so a single navigation model had to flex to genuinely different workflows.
An audience new to the category
Almost none of Lleverage’s users were technical, and agentic AI workflows were a brand-new concept to most of them.
Parts built before the whole made sense
Workflow canvases, data tables, integrations and onboarding were each shipped individually, but they’d never been pulled into one coherent system.
Representative of feedback patterns described in research and retrospectives, written to reflect the theme, not verbatim quotes from named individuals.
“I’m not sure what this workflow tool is even supposed to do for me.”
New client feedback, paraphrased
“I check three menus before I find what I need.”
Client feedback, paraphrased
Who this was for
Two different users, two different needs
Non-technical operator
New to agentic AI
- Onboarding that assumes zero prior context
- Guardrails that prevent costly mistakes on day one
- Confidence the product works before they touch a real workflow
Power / admin user
Runs varied workflows
- Navigation flexible enough for a genuinely different use case each time
- Fast access to workflow canvases, data tables and integrations
- A system built to keep working as the product keeps changing
Key decisions
Problem, decision, outcome
Final design
What I designed
Information architecture
Agentic navigation
Rebuilt navigation around how people actually use an agent-first product, not around a traditional menu structure, so every client type could find their own path through it.
0→1 product
Workflow canvases & data tables
Designed the core building blocks clients use to define and run their automations, built to flex across genuinely different use cases.
Onboarding
A first-run experience for non-experts
Built onboarding that assumes no prior AI or automation knowledge, so day one doesn’t require a training session.
Future-proofing
Designing for what’s next
With adoption largely solved, the current challenge is building the same system to hold up as the product and category keep moving.
Edge cases
Constraints that shaped the decisions
Constraint
Every client shows up with a genuinely different use case
Response: tested each pivot with a real focus group before shipping it platform-wide, rather than assuming one workflow generalised.
Constraint
A non-technical audience meeting agentic AI for the first time
Response: built onboarding that assumes zero prior context, so day one doesn’t need a training session.
Constraint
A category that’s still being invented, including by us
Response: treated navigation as something to keep iterating, not something to solve once and leave.
Trade-offs
What I chose not to do
We didn’t ship one universal workflow template
Every client’s use case was different enough that forcing one template would have hurt adoption, not helped it.
We didn’t wait for a perfect navigation model before shipping
Each pivot went out as soon as it tested well, even knowing it would likely change again.
Outcome
Where adoption landed
Adoption
98%
Clients migrated to the new agentic navigation
Scope
5
Core product areas redesigned
Testing
Add count
Focus groups run across pivots
Ground-up design across brand, product and onboarding.
Testimonials
“Add a quote here from a colleague about working with you at Lleverage.”
Name, Title, Lleverage
“Add a second quote here, if you have one.”
Name, Title, Lleverage
Reflection
What I learned
Test the pivot, not just the launch
Every direction change deserved its own round of real user testing, not just the big launches.
Simplicity is a moving target
Navigation that works today won’t automatically work for what the product becomes next.
Skills demonstrated
More work