Building a visual workflow builder that lets businesses automate repetitive tasks using AI by connecting tools, services, and logic without writing a single line of code.
MY ROLE
As part of a cross-functional product team, I contributed across the full product process from research and competitive analysis, through problem framing, wireframing, prototyping, usability testing, and design system development.
I worked closely with engineers throughout, participating in weekly stand-ups and design reviews to ensure feasibility and maintain consistency between designed intent and built output.
My primary focus was the core workflow builder experience, specifically how users initiate, build, and configure workflows from scratch or from templates, and how the system guides them through each step without requiring external support.
Company: Mantium
Sector: Artificial Intelligence · SaaS · Cloud Platform
Methods: Research · Stakeholder Interviews · Problem Framing · Competitive Analysis · Wireframing & Prototyping · Usability Testing · Cross-functional Collaboration · Design Strategy
Tools: Figma · GitLab · Notion
Year: 2022
PROJECT OVERVIEW
WHAT IS MANTIUM ABOUT?
Mantium is a global cloud platform built to simplify the development and management of AI applications at scale. With a mission to democratize AI, Mantium enables teams to design, test, deploy, and share AI-powered automations using tools like the AI Builder and One-Click Deploy.
THE PROBLEM
Repetitive, manual tasks such as pulling reports, routing queries, and handling follow-ups cost businesses thousands of hours each year. The challenge isn’t the lack of AI capability, but the difficulty of turning that potential into usable workflows.
Building and deploying these solutions still requires technical expertise and engineering support, resources most teams don’t have readily available. As a result, automation ideas get stuck in backlogs or never get built.
The tools that exist don't bridge that gap either: they’re either too limited to scale or too complex for most teams to adopt without dedicated support. This creates a clear gap between what AI can do and what teams can actually execute, a gap that Mantium set out to close.
THE SOLUTION
To address this gap, we built a workflow builder that enables any team to design, build, and deploy AI-powered automations independently, without writing code or relying on engineering.
The experience combines a structured template library for quick starts, a visual canvas for custom workflows, and real-time guidance at every step, balancing ease of use with advanced control.
As a result, teams that once waited weeks for engineering support can now build, test, and deploy automations in a fraction of the time, turning ideas into real business outcomes without the bottleneck.
DISCOVERY & INSIGHTS
Before designing anything, we needed to understand what already existed and where it was falling short. Our research focused on three areas: defining who we were building for, understanding what they needed, and understanding what was stopping teams from automating the work they already knew could be automated.
UNDERSTANDING THE MARKET
During the research phase, we set out to validate our assumptions rather than design around them. We focused on:
-
Defining our primary user groups and understanding their specific goals and operational constraints.
-
Analyzing how the AI Builder could compete with and improve on existing solutions in the market.
-
Mapping the gap between what users needed from a workflow tool and what existing platforms were actually delivering.
COMPETITIVE ANALYSIS
We conducted a structured analysis of four direct and indirect competitors, Zapier, n8n.io, Runalloy, and Workflow86, to identify the gap Mantium was positioned to fill: a middle ground between ease of use and advanced capability. Each platform was evaluated across three key dimensions:
-
Ease of getting started for a first-time user.
-
Depth of capability for users who needed advanced control.
-
Clarity of guidance when users encountered complexity.
DEFINING THE PROBLEM
PROBLEM STATEMENT & GOALS
Teams across every industry have repetitive, time-consuming tasks that AI could automate, but building those automations has always required technical resources most teams don't have on demand.
How might we build a workflow tool that gets any team from problem to working automation as quickly as possible, without limiting what experienced users can build?
PRODUCT GOALS
The goal was to close the gap between AI's potential and a team's ability to use it. To get any team, regardless of technical background, from problem to working automation faster than any existing solution could.
STRUCTURING THE EXPERIENCE
Before any interface decisions were made, we mapped the full workflow builder experience as a system, clarifying entry and exit points, decision paths, and the moments where users would need guidance or could get stuck. This map became the foundation for every decision that followed: what users see first, what they can do next, and what happens when something goes wrong.
FROM FLOW TO WIREFRAME
Low-fidelity wireframes were developed to test the logical structure of the experience before investing in visual design.
Key decisions at this stage focused on:
-
Simplifying navigation so users could move through the workflow builder without losing context or their place in the flow.
-
Structuring the template library so users could find and implement relevant templates in the fewest possible steps.
-
Structuring the canvas layout so complex, multi-step workflows remained readable and manageable as they grew.
SOLUTIONS
Workflow Initiation
Two clear entry points were established at the start of the builder: "Explore Templates" for users who wanted a guided starting point and "Get Started" for users building from scratch.
This removed decision paralysis at the most critical drop-off point in the user journey, the moment a user arrives and doesn't know where to begin.
Pre-built Workflow Templates
A structured library of pre-built workflows organized by business use case, marketing, customer service, finance, and translation, with search and filtering by integration type.
The goal: get users to their first working automation as quickly as possible. This reduced the time between signing up and completing a first working workflow, the most critical drop-off point in any SaaS product adoption cycle.
Information Architecture & Navigation
Collapsible sidebar navigation reduced visual complexity during multi-step workflow creation. A contextual side panel surfaced configuration options for each workflow block, bringing guidance into the product at the exact moment users needed it, rather than directing them to external documentation.
Real-Time Validation System
As users configure each step, the system provides immediate feedback, green checkmarks for correctly completed steps, error indicators for missing parameters, and a running error count at the top of the interface.
This allowed users to identify and resolve issues independently, reducing support dependency and increasing confidence in completing complex workflows without help.
Active App Management
The apps overview page was restructured with a toggle distinguishing active from inactive apps and an error summary button providing quick visibility into outstanding workflow issues, reducing the time required to diagnose and resolve cross-app configuration problems.
USABILITY TESTING & VALIDATION
To evaluate the effectiveness of the workflow builder, usability sessions were conducted with five users, measuring how quickly they could complete their first workflow, how easily they discovered relevant templates, whether they could identify and resolve errors independently, and how confident they felt navigating the builder overall.
WORKFLOW EFFICIENCY:
Users navigated the workflow builder without guidance, confirming that the progressive disclosure approach was reducing complexity at the right moments.
TEMPLATE ACCESSIBILITY:
Users who started with a template reached a working workflow faster than those who started from scratch, validating the template library as the most effective entry point on the platform.
ERROR HANDLING EFFECTIVENESS:
The real-time validation system was the most positively received feature. Users reported feeling more confident completing workflows because they could see exactly where issues were and resolve them without external help.
ONE GAP IDENTIFIED:
Users building multi-step workflows with branching logic occasionally lost their place on the canvas, flagged as the highest priority for the next iteration.
DESIGN SYSTEM & DEVELOPER HANDOFF
To ensure what was built matched what was designed, a component-level design system was developed and shared with the engineering team.
Weekly stand-ups and design reviews created regular touchpoints to compare implementation progress against design specifications, catching discrepancies early and reducing rework during development.
This collaborative approach was particularly important for the real-time validation system and step-by-step guided interactions, where behavior logic required close alignment between design decisions and engineering implementation.
CONCLUSION
IMPACT
Although quantitative metrics were not formally tracked at launch, the AWS partnership and qualitative feedback from the team indicated meaningful progress across the platform.
Wider Product Accessibility
Teams across sectors were able to independently build and deploy AI workflows, eliminating a key bottleneck that previously required engineering support for every automation request.
Faster Time to First Value
Users starting from templates reached a working automation significantly faster than those building from scratch, reducing a major drop-off point in the onboarding journey.
Reduced Support Dependency
Real-time validation enabled users to identify and resolve workflow errors on their own, minimizing reliance on external support during the build process.
Improved User Confidence
Usability feedback showed that users felt more in control of complex workflows, with clearer visibility into issues and how to resolve them at each step.
Strategic Business Validation
Post-launch, the feature helped secure a partnership with Amazon Web Services, validating the product direction and reinforcing enterprise credibility.
KEY TAKEAWAY
The most important lesson from this project wasn't about AI or automation. It was about access.
The best tool in the world is only valuable if the right people can actually use it. The product challenge here was never "how do we make this powerful?" Mantium's engineering team had already solved that.
The challenge was, "How do we make this approachable without making it less powerful?"
Asking that question first before building anything was what made the difference.