Project highlights
Task
Redesign the AI workflow automation platform’s UX/UI and create a go-to-market MVP to attract investment and users
Team
1 Product Designer, 1 BA/PM, working directly with the client's CEO and CTO
Duration
~2 months (core product design + follow-on for landing page)
Scope
UX/UI redesign using an existing component library, chat-based agent creation, real-time run visibility, integration and trigger setup, Webflow landing, and animation concepts
About the client
Tandom is an AI-native automation company focused on the complex logistics workflows that traditional robotic process automation and low-code solutions can’t handle – from freight quoting to rate confirmations. Unlike typical automation platforms that require coding or flowcharts, it lets anyone create powerful AI agents by chatting with the system.
With a team that previously scaled logistics tech to tens of millions in ARR, Tandom is now focused on building the automation layer for a $10T global industry, eliminating the $1T in wasted productivity caused by manual, exception-heavy work.
What the client needed
The founders came to us with a very early MVP. Their idea was compelling: create agents through a natural chat, then watch and control step-by-step runs. But the user experience hadn’t yet caught up to the vision. The team needed our help to transform their initial concept into a product that felt as intelligent as the technology behind it:
- turn chat ideas into agents that semi-technical users can review and tweak;
-
show runs with real-time progress and easy error insights;
-
give flexible control: run manually, on a schedule, or from triggers like new emails;
-
simplify integrations, so connecting Gmail, Sheets, or portals feels safe;
-
deliver a professional marketing website that tells their unique story.
Challenges
- design had to stay flexible while working within the client’s existing component library (shadcn/ui);
- users ran real work across spreadsheets, portals, and email, so progress and errors had to be fully visible, not “black box” magic;
- the UX had to support automations even in tools with weak or no APIs;
- a small founder team meant focusing on what shipped, not reinvention.
Results
- a realistic, clickable MVP positioned the product for its upcoming beta launch;
- users could build automations from chat, then review and tweak them with editable, transparent steps;
- real-time run views helped teams catch failures fast and retry or adjust them quickly;
- connecting tools like Gmail and Sheets felt safer, with clear permissions and easy-to-manage access;
- a single control panel made it easy to set up schedules, triggers, and notification preferences.
Our approach .
Tandom brought us in early, when the MVP still felt like a proof of concept. We worked directly with the founders in short, high-trust loops, mapping gaps, refining flows, and designing fast without overcomplicating:
-
cadence: 1-2 weekly feedback loops with “show, decide, ship” rituals;
-
method: lo-fi → hi-fi → clickable prototype; short validation cycles with the client’s domain experts;
-
delivery: production-ready Figma files mapped to shadcn/ui, motion concepts and implementation notes;
-
principles: chat-first, privacy-by-default, step-by-step transparency, and aimed for minimal cognitive load.
Agent creation
Tandom's core idea is letting users create AI agents by chatting, with no flowcharts, no code. But the early flow made it hard to understand what the system would do or how to adjust it. We redesigned the flow to feel more transparent and editable for semi-technical ops and finance users:
-
every chat exchange generates a structured plan showing what the agent will do;
-
users can open any step to adjust inputs, outputs, or logic before running it;
-
guided integration prompts help users securely connect tools like Gmail or Meet;
-
built-in trigger setup lets agents run manually or on a defined schedule.
Run tracking
Teams needed to see what their automations were doing, not just whether they worked. Our goal was to design a transparent run view that shows exactly what’s happening at every step through:
-
real-time progress indicators for each action, with clear success and fail states;
-
visible error points with context to help users retry or correct quickly;
-
manual controls for stop, rerun, and history to keep every run easy to follow.
Settings
Agents could be triggered in different ways depending on the workflow. The updated settings view brought all options into one place and made scheduling easy to manage with:
-
unified options for manual, scheduled, external, and chained runs;
-
clear controls for time, repeat frequency, and timezone;
- simple fields for start and end dates.
Notifications
To keep users informed without digging into logs, agents now send in-app notifications when runs succeed, fail, or are about to start. Each message shows the agent name and trigger type, giving teams a quick overview of activity right inside the platform.
Integrations
Users could connect tools like Gmail, Sheets, and internal portals through short, guided steps. Each connection showed clear permission details and worked even for systems without APIs, using browser-based automation for a seamless setup.
Landing page
To help the team present Tandom to investors and early users, we created a Webflow landing page explaining the chat-to-agent concept through a short hero animation. The page outlined key product outcomes, highlighted real use cases, and gave the brand a clean, confident presence ahead of launch.
Marcelo Fujimoto
Co-founder & CEO at Tandom
They didn’t try to impose any of their pre-held beliefs around product on us. They were really open-minded and listened to us. In fact, most of their opinions, we did end up putting into the design, into other projects that we've worked on together.
we reply under 24 hours.
free
session
