← Case StudiesDJ Delos Santos

№ 01, AI-NATIVE SUPPORT

Hey Uli

Turning Music Tribe's support from static forms into a chat-first assistant, designed directly with founder Uli Behringer and launched to the public.

COMPANY

Empower Tribe (Music Tribe)

DURATION

4 months (concept to dev handoff)

TEAM

Sole designer, working directly with founder Uli Behringer and engineering

MY ROLE

AI Product Designer (AI-assisted workflows)

THE PROBLEM

Support started with the customer guessing

Customers had to understand Music Tribe's internal support structure before they could get help, and every wrong turn cost the business time.

Music Tribe customers had to work through static forms and figure out how the company's support was organized before they could get help. A customer might need product registration, troubleshooting, a service center, billing help, or a real person, but it was on them to pick the right path and give enough detail. When information was missing or unclear, agents had to chase it down, which slowed every case, a cost that grew with the support volume of a global brand.

THE CHALLENGE

Not a chatbot bolted onto a form

I had to turn a founder's vision into a real product, not just bolt a chat widget onto the old form.

The job was to turn Uli's idea of a digital assistant into a product that actually worked. Hey Uli had to feel personal and conversational, understand what people meant from plain language, walk them through support tasks, know when to stop and hand off to a person, and give agents cleaner information, all without feeling like a generic bot stuck onto a form.

MY APPROACH & PROCESS

From a vision to a launch-ready product

01

Framing it as a real product, not a bolt-on

I turned the founder's vision of a digital Uli into a real product plan: the main way customers get support, not a chatbot floating on top of the old forms.

  • Set up Hey Uli as Music Tribe's main support experience, not an add-on
  • Mapped the support requests that cost the most: registration, troubleshooting, service center, billing, and handoffs to a person
  • Changed the starting point so customers just describe the problem instead of picking the right form
02

Designing it to read what customers mean

I built the assistant to figure out what the customer needs and gather details a little at a time, so people describe the problem in their own words while the system quietly collects what agents need.

  • Designed flows for registration, step-by-step troubleshooting, and service center help
  • Added quick replies, voice, and photo input to make asking easier
  • Decided exactly when the assistant handles it and when it hands off to a person
03

Prototyping fast with AI tools

I used AI tools to get from idea to a working, testable prototype quickly, and gave the assistant a real personality instead of a generic bot voice.

  • Built the working prototype in Figma Make to test the full support journey
  • Used ChatGPT, Gemini, and Nano Banana for scenarios, content, and visual direction
  • Shaped a practical, direct, sincere personality modeled on Uli
04

Handoff and public launch

After direct reviews with the founder, I refined the product and handed the final design to engineering in mid-March 2026. It launched publicly on the Music Tribe website the following month.

  • Defined when to hand off to a person and exactly what information agents get
  • Documented the customer flows and support rules for engineering
  • Delivered a launch-ready design that shipped to real customers after handoff

KEY SCREENS

Intent-first flows, end to end

Landing, An open, chat-first start, so customers describe their problem in plain words instead of picking a support category.
Guided Troubleshooting, The assistant asks the right questions, tells smart products from basic ones, and tries to fix the issue before creating a ticket.
Product Registration, A plain request becomes a guided registration, with details filled in automatically and confirmed step by step.
Resolution, Confirmation pages wrap things up with reference numbers, warranty details, appointment info, next steps, and a way to give feedback.

SIGNALS & FEEDBACK

Backed by the founder

Uli approved the direction and pushed to get it live quickly, a clear sign from the top that both the idea and the design were ready.

Built for agent efficiency

The assistant handles common issues itself and gives agents cleaner information, so the cases that reach a person arrive ready to work, not cold.

Grounded in real patterns

Research into other products and AI-built scenarios shaped the support categories, handoff points, and limits, instead of guessing at them.

DESIGNING THE AI'S BEHAVIOR

  • Let customers start in their own words instead of picking from internal support categories.
  • Work out what the customer needs first, then ask only for the details that case requires.
  • Try step-by-step troubleshooting before creating a ticket when the issue can likely be solved.
  • Hand off to a person when the issue is complex, unresolved, about billing or partners, or beyond what the assistant is sure about.
  • Pass the person the full history, what the issue is, what was already tried, the AI's read on it, and suggested next steps.

KEY INSIGHT

Designing an AI product isn't about adding a chatbot to a screen. The real work is deciding how the AI behaves, when it solves the problem itself, when it stops, and when it hands off to a person, because that behavior is what cuts support tickets and makes agents faster.

THE SOLUTION

A conversational support portal

I designed Hey Uli as a chat-based support assistant that replaced manual forms with guided help. Customers start naturally through chat, voice, or a photo. Hey Uli works out whether they need product registration, troubleshooting, a service center, or a person, then asks only for what that case needs. For common issues it walks through troubleshooting before creating a ticket; for anything complex or about billing or partners, it hands off to a person, giving the agent a fully packaged case.

WHAT THE AGENT RECEIVES ON ESCALATION

  • Full conversation history
  • Issue classification
  • Attempted troubleshooting
  • Priority context
  • AI analysis
  • Suggested responses

IMPACT & OUTCOMES

What shipped

Shipped to public production

Went from a founder's idea to a designed, prototyped, approved, and publicly launched support product on Music Tribe's site.

Forms became a conversation

Turned static support forms into a guided conversation, in step with Music Tribe's Industry 4.5/5.0 direction.

Agents get cleaner handoffs

Handoffs carry the full history, what was already tried, the AI's read, and suggested replies, cutting the back-and-forth that slows things down.

Effort shifted off the customer

Guided support means customers no longer have to figure out how the company is organized before they can get help.

WHAT I LEARNED

Reflections

  1. 01

    The real design work in an AI product is its behavior: the assistant has to know when to guide, when to solve, and when to hand off, and those lines are what keep trust and cut cost.

  2. 02

    Personality is a product decision. Hey Uli had to feel practical, sincere, and direct, not like a generic bot, for customers to trust it.

  3. 03

    AI can improve both sides of support at once: a lighter customer experience and a faster agent workflow behind it.

  4. 04

    AI tools speed up exploring ideas, but the designer still owns the judgment, trust, tone, and behavior that make it actually usable.

← All case studiesNEXT PROJECTFree MusicTribe →