Deutsche Telekom · TSC
Internal AI Tools for Consulting Teams
Within Telekom Strategy Consulting, I worked in the Digitalization & IT practice, helping consultants and internal clients adopt digital and AI solutions.
As the sole UI/UX designer, I designed and prototyped several internal tools, often working from early requirements through to implementation.
TEAM
Full-Stack developer
Counselors (Freiburg Cancer Counseling Center)
Researchers (Fraunhofer ISI)
Lead professor (Hochschule Heilbronn).
MY ROLE
Product design
INDUSTRY
HealthTech (psycho-oncological care)
PRODUCT TYPE
B2B2C
TIMELINE
12 months ( mini job )
From mid-2026 I was embedded within TSC, designing and building four AI-powered internal tools. Each product ran on a custom implementation of Telekom’s One Design System — with a push to make them genuinely good-looking within those constraints.
Uncorporate Scout
Employee feedback intelligence. Structured intake, AI classification, human-in-the-loop review and prioritization.
Agent Hub
AI tool directory for TSC’s consulting team. Custom ODS reimplementation, from-scratch component system.
Pitch Maker / T-Deck
AI-powered presentation generator. Flow design and frontend contributions across editor and export.
Video Maker
Figma concept only — not developed.
01
Uncorporate Scout
WHAT IT IS
Uncorporate Scout ingests employee questionnaire responses and turns them into structured, reviewable clusters. Employees submit feedback via a campaign — the system normalizes, scores, and classifies each entry against a fixed taxonomy — then surfaces results for human review, prioritization, and decision tracking.
MY CONTRIBUTION
I designed the UX and implemented the full frontend. The pipeline concept — replacing a manual Excel-based review process — was part of my contribution. The AI classification logic and backend were built by the developer, who merged from my branch.
DESIGN DECISIONS
Confidence as a first-class signal
The entries table surfaces AI confidence score as a primary sort axis, with tier color-coding. Reviewers can see at a glance where the AI is certain and where it needs attention — making the system auditable rather than opaque.
Human-in-the-loop prioritization
The scoring dashboard presents entries across multiple dimensions — but never replaces human judgment. Reviewers can override, re-score, and re-cluster. AI structures and proposes; the final decision is always human.
AI transparency by design
The cluster view surfaces model version, taxonomy version, and prompt version inline — so operators can trace any classification result back to the exact AI state that produced it. Deliberate design for long-term maintainability.
[ Screenshot placeholder — Uncorporate Scout UI ]
02
Agent Hub
WHAT IT IS
A landing page and directory for TSC’s internal AI tools — a single place where consulting teams can find and access the AI agents available to them. I built the entire product from scratch: design, component system, and implementation.
THE CHALLENGE
ODS is functional but conservative. The brief called for something visually stronger — a product that felt genuinely well-crafted within design system constraints. I implemented all ODS tokens faithfully, then pushed visual quality through layout, text hierarchy, and interaction detail. No external ODS dependency: every component is a local reimplementation.
[ Screenshot placeholder — Agent Hub UI ]
03
Pitch Maker / T-Deck
WHAT IT IS
An AI-powered presentation generator. Users input a brief, AI generates a storyline and slide content, then iterate in an editor before exporting. My contributions focused on the full flow design and specific frontend UI components throughout the product.
FLOW
Intake
→
Storyline
→
Editor
→
Export
KEY CONTRIBUTIONS
Drag-to-reorder on slide thumbnails (dnd-kit, optimistic update), export as a modal overlay wired from editor and My Decks, a reusable EmptyState component across all empty views, ODS icon badges replacing emoji in template cards, and a deliberate choice to disable the Back button from the editor — avoiding a destructive trap that would wipe per-slide AI edits.
[ Screenshot placeholder — Pitch Maker / T-Deck UI ]
REFLECTION
Working across four AI products simultaneously reinforced something I already believed: the hardest problems in AI tool design are not about the AI. They are about the humans interacting with it. How much do you show? When do you intervene? How do you make a model’s uncertainty legible without overwhelming the reviewer? These are the questions this body of work kept returning to.