Project Summary
AI Branding Translator
Using Natural Language Processing, we built a product to translate any text into branded corporate language of a company. The translator aims to make communication towards customers or business partner more consistent while reducing workload of editorial writers and marketing.
Company
Vodafone Germany
Center of Excellence AI
Team Composition
Lead Product Designer (Me)
1 Frontend Engineer
PM
1 Data Scientists
1 Data Engineers
Backend Engineer
Agile Coach
Key Contributions
Stakeholder Interviews
User Research
Lo- & HiFi Designs
User Testing & QA
Vision Crafting
Feature Prioritization
Market & Competitor Research
Project overview
0 to MVP
Demand Intake
Product & Problem Discovery
Team Kickoff
User Research & Strategy
MVP Scope & Roadmap
Iterative Prototyping
MVP Deployment
Post-Launch Testing & Feedback
As this project was developed under NDA, I am showing simplified designs I custom made for this showcase.
Final Designs
Branded translation
Users are enabled to write or paste a text into the input field to instantly translate it to corporate, branding ready copy.
Translation insights
Users can select parts of the translation text in order to receive reasoning behind the proposed phrasing. This helps understanding the output text, builds trust and transparency.
Onboarding flow
First time users complete an onboarding flow accelerating time to first value, building reasonable expectations towards the product and reducing cognitive load.
Desktop designs available in private meeting.
Product & Problem Discovery
Overview
When: Very beginning of the project
Who: PM, Me (and technical roles where needed)
Goals:
Understand user journeys
Identify target groups
Understand user pain points, needs and goals
Increase stakeholder backing
Clarify business opportunity
My approach
Stakeholder Interviews
User Interviews (general & contextual inquiries)
Data analysis
Market analysis
Outcomes
Problem brief (user goals, problems, needs..)
Defined user groups
Hypotheses
Early success criteria
Business case refinments
Key Insights
Editorial writers have more copy requests to check than capacity
Users regularly just skip alignment with the editorial team due to time constraints
The planned concept would really help out users and save the company money
From the first problem discovery
Inconsistent brand communication across departments was weakening brand trust and slowing down time-to-market
Inconsistent tone and terminology
The company had multiple sub brands with distinct tone which made it even harder to apply the rules consistently
Branded writing was something that could be automated well but no technical fundament existed, yet
Business Impact: operational inefficiency, slower campaigns, delayed launches and less trust in brand consistency across touchpoints
From the first problem discovery
Non-Writers (customer support, marketing, ...) had to make sure any consumer facing copy is brand compliant
They are not trained on branding rules
Their time should be spent on other tasks
Check with editorial writer team is mandatory for consumer facing copy
Business impact: slower execution, inconsistent brand voice, insufficient use of recourses
Editorial writers at the company have a high work load causing bottlenecks and inconsistent branding
Checking every piece of content often requires multiple approvals
They cant keep up with review requests or branding checks in busy times
Their time could be spent on other tasks
Business impact: bottlenecks, skipped reviews, inconsistent brand voice, insufficient use of recourses
User Research
Approach
We knew at this point that first assumed user problems are existent and they connect to existing business problems. In order to understand how we need to shape our product and MVP in order to achieve a sufficient user adoption to reach our business goals, I created and executed a detailed user research plan.
Overview
When: During early ideation & prioritization
Who interviewed: Me
Interviewed stakeholders: Users within selected primary & secondary target groups
Goals:
Understand user journey, goals, pain points and needs more in detail
Identify necessary MVP scope for users to actually use the product (nice to haves vs must haves)
Understand expectations of users
My approach
Contextual inquiries (think aloud)
Prototype walkthroughs
Outcomes
MVP feature set
User personas
Prioritization insights
Design requirements
User flows
Refined user journey maps
Key Insights
Users find quality of translation most important
Ability to report errors, give feedback is highly regarded
Integration of a confidence score is crucial for adoption
Finding past translations is much needed for users' workflows
Product & Design Strategy
Within this phase, mostly my PM and me defined general boundaries for the product's development that make it measurable and tangible.
Hypotheses
In order to make our work testable, I initiated to build hypotheses. This way, we had clear assumptions about user problems. These can be tested and understood easily by every stakeholder.
Hypothesis 1
If we give employees an easy to use tool for brand compliant writing, messaging will become more consistent, because they won’t have to rely on difficult to follow guidelines
Hypothesis 2
If employees can self assure brand compliance, we are able to save resources of editorial writers, because fewer manual reviews will be needed
Hypothesis 3
If employees can confirm brand compliance using a specialized tool, various teams will reduce their reliance on expert editors and brand teams, because no reviews will be needed for standard content.
Success Critera
To make our assumptions measurable, I always define success criteria as early as meaningful. This builds clarity on what problem severity matters most to users and business. Also, this builds an early foundation for learning and adapting as you go.
Time saved per use
Highly important in order to measure user impact and generated business value.
Number of completed translations per user
In order to identify adoption and repeat usage
Reported trust score
In order to understand users relation of trust to adoption: how good does the trust score need to be for x% adoption?
Frequency of “rewrite” feature usage
To understand whether initial translations were sufficient. Also, to understand how high confidence scores need to be to not be rewritten.
MVP Scope
Understanding user and business context, it was time to bring those in alignment in order to generate tangible outcome for both. Based on our discovery and product strategy, we defined an MVP feature set as a team. I made sure to discuss feasibility as early as possible with our engineers and enable every team member to emphasize with our users and their problem spaces.
Translation of an input text into a branded version · Core feature
Display of a confidence score · Build trust
Feedback functionality · Build trust
Translation reasoning · Transparency
Translation history · Core workflow feature
Time saved per use
Simple & quick onboarding · trust, transparency & expectation management
Testing
Testing Overview
When: Before & after MVP release
Who interviewed: 2 student designers I mentored & me
Interviewed stakeholders: Users within selected primary & secondary target groups
Testing Goals
Pre Release
Test functional usability
Catch low-hanging usability issues and bugs
Post MVP Release
Are users completing the core task we built the MVP for?
Where do they drop off or get confused?
Do we feel like the product solves their problem?
My approach
Pre Release
Lightweight usability testing
Post MVP Release
Usability testing
surveys
Analytical analysis
Key Insights
Users find quality of translation most important
Ability to report errors, give feedback is highly regarded
Integration of a confidence score is crucial for adoption
Finding past translations is much needed for users' workflows
Impact & Reflection
Impact
Any numbers are left out of this section due to NDA. Available for discussion in private meeting.
We measured significant time saved by our primary and secondary users when writing branded copy.
This led to large savings of operational costs and more effective use of operational capacity.
Feedback showed that users felt comfortable using the tool with a very high trust score.
Measured through our quantitative user research and feedback feature
The translated texts were compliant with the official branding rules in almost all cases
While here and there mistakes and bugs occurred, safety features like the compliance ranking helped users identify errors before sending those texts out
The editorial writer team reported less requests as well as higher quality requests, saving operational costs.
While here and there mistakes and bugs occurred, safety features like the compliance ranking helped users identify errors before sending those texts out
Reflection
The more technical a product is, the more collaboration between different perspectives is needed for a shared directioned significant time saved by our primary and secondary users when writing branded copy.
Building for AI demands more transparency and control measures than for non-AI applications.
As much as we evolve, user evolve, too and we got to adapt.
Thanks for reading. Let's connect!
Maxim Barwinkewitsch
Product Designer · London
2025 Maxim Barwinkewitsch
