UX Researcher

Angelina
Demirtzoglou

I'm a UX Researcher specializing in AI products, with a background in cognitive science and communication. I run research from early discovery through launch, helping teams build generative AI features that are intuitive.

Get in touch LinkedIn ↗

📍 Copenhagen, Denmark

Angelina Demirtzoglou

Research work

Portfolio

Corporate research

Microsoft Nov 2024 – Aug 2025 · UX Researcher (Maternity cover)
Business Central·Sales Order Agent 2 studies · assessing usability and generated content
Evaluative Study ›
Method

2 moderated usability testing studies, 7 participants (partners & customers) per study, affinity mapping.

My role

Sole and lead researcher.

Outcome
Terminology refined
System language clarified to reduce interpretation effort and build user trust
Discoverability improved
Key functionality surfaced more prominently in the UI
Tools

UserTesting

Business Central·Autofill 3 studies · AI feature for auto-filling customer card data
Evaluative Study ›
Method

3 moderated usability testing studies, 5–6 sessions each, industry participants, affinity mapping.

My role

Sole and lead researcher.

Outcome
Study 1
Discoverability fixed
UI prominence improved for hidden elements
Study 2
Terminology aligned
Language matched platform conventions
Study 3
Design validated ✓
Refinements confirmed — ready for rollout
Tools

UserTesting

Business Central·Summarize 2 studies · AI summary feature across customer card, sales order & more
Evaluative Study ›
Method

2 moderated usability testing studies, 7 partner interviews per study, affinity mapping.

My role

Lead researcher.

Outcome
Customisation scope defined
Determined how tailored the Summarize experience should be for partners
KPI preferences mapped
Priority KPIs identified, setting up a follow-up prioritisation study
Tools

UserTesting, UserInterviews

Business Central·AI GTM Value Propositions Mixed methods · understanding how partners communicate AI feature value
›
Method

7 semi-structured interviews with partners, used to develop surveys for both partners and users; affinity mapping; descriptive statistics; opportunity scoring.

My role

Supported senior UX researcher — recruited BC partners, moderated interviews, collected survey data, analysed data and created visualisations.

Outcome
Finding
Partner communication gaps
Identified how partners struggle to convey AI value to their customers
Impact
Product roadmap shaped
Insights fed directly into how Microsoft supports partner go-to-market
Tools

UserTesting, UserInterviews, Excel

Persona Study·Sales 7 interviews · identifying opportunities for AI features in the Sales landscape
Strategic Discovery Using Jobs-To-Be-Done ›
Method

7 moderated interviews, affinity mapping, Jobs To Be Done framework.

My role

Sole and lead researcher.

Outcome
Insights to guide ideation of new AI features for Sales in Business Central
Informed how Microsoft can better support partners' customers through new AI capabilities
Tools

UserInterviews, UserTesting, Excel, FigJam

Microsoft Nov 2022 – Jan 2024 · Product Design Researcher (student worker)
Business Central·Accounts Payable Benchmark Study Unmoderated usability testing · payment tasks & vendor workflows
Evaluative Study ›
Method

Unmoderated usability testing; descriptive statistics, SAT scores, affinity mapping, data visualisation.

My role

Lead researcher.

Outcome
Identified
Usability gaps
Design improvement opportunities surfaced across key AP tasks
Result
Tangible product changes
Design process moved forward with concrete improvements implemented
Tools

UserTesting, Excel

Business Central·In-product Interview Evaluating a role-query feature shown on first login
Evaluative Study ›
Method

Unmoderated usability testing; qualitative data only; affinity mapping.

My role

Sole researcher and lead.

Outcome
Usability friction identified
First-login design elements caused uncertainty around why certain information was being requested
UI made transparent
Design team removed the problematic elements entirely
Tools

Figma (prototypes), UserTesting, Excel

Onboarding Survey Understanding user experiences during Business Central onboarding
Survey ›
Method

Survey; mixed methods; affinity mapping; descriptive statistics.

My role

Responsible for thematic analysis, descriptive data analysis, and visualisations.

Outcome
Users needed clearer documentation on data integration
Finding directly informed improvements to onboarding resources and support materials
Tools

Excel

Business Central·Search in Company Data Assessing usability of the search feature for new and existing users
Evaluative Study ›
Method

Combination of moderated and unmoderated usability testing; affinity mapping.

My role

Lead researcher.

Outcome
Problem
Terminology confusion
New users struggled with system-specific language in search
Solution
Design changes made
Language updated to be more intuitive for new users
Tools

UserTesting, UserInterviews

Business Central·Search in Company Data (Follow-Up) Evaluating impact of changes made after the first study
Evaluative Study ›
Method

Unmoderated usability testing with non-existing users; affinity mapping.

My role

Sole researcher and lead.

Outcome
Improvements confirmed
Post-change experience measurably better for new users
Clicks reduced
Further refinements guided to streamline the user flow
Tools

UserTesting

Persona Study·Finance Discovery that sparked the Payables Agent — now in production
Strategic Discovery Using Jobs-To-Be-Done ›
Method

Mixed methods; opportunity scoring; segmentation survey; quantitative data analysis; Jobs To Be Done.

My role

Co-lead — led discovery phase, created segmentation survey, analysed quantitative data in R and Excel, built visualisations.

Outcome
Sparked the creation of the Payables Agent for Dynamics 365 Business Central
Research defined the MVP and experience principles
✓ Deployed in production
Tools

R, Excel, FigJam

one.com Apr 2022 – Jul 2022 · UX Research Intern
Persona & User Journey Mapping Defining the user journey and refining personas for the WordPress product
Strategic Discovery ›
Method

Moderated interviews with existing customers; card sorting sessions; thematic analysis; stakeholder brainstorming (Brainwrite technique).

My role

Supported senior designers — recruited participants, moderated interviews and card sorting, conducted thematic analysis, facilitated brainstorming sessions.

Outcome
Refined personas
Clearer, more accurate user personas for the WordPress product
Journey map clarified
Improved empathy and shared understanding across the team
Tools

Hotjar, Maze (recruitment), DoveTail, Miro

Academic research

Master's Thesis

Cognitive Biases in AI: Exploring the Anchoring and Framing Effects in Large Language Models

As LLMs become integrated into decision-making — from financial choices to purchasing decisions — understanding their susceptibility to cognitive biases is critical. This research investigated whether LLMs exhibit human-like biases, specifically anchoring bias and framing effects, and explored the implications for trustworthy AI systems.

Aug 2023 – Dec 2023  ·  University of Copenhagen · Master's Thesis

›
Research question

Can user inputs cause LLM outputs to be influenced by cognitive biases, potentially leading to untrustworthy and unfair decision-making patterns?

Models tested

ChatGPT-3.5 Turbo, Llama 2, Mistral, Vicuna

Methodology
  • Anchoring bias: Low vs. High Anchor conditions across 5 experimental trials
  • Framing effect: Loss Frame vs. Gain Frame across 4 trials
  • Independent groups design; statistical significance testing vs. human baseline (N=63 recruited participants)
  • Python used for automated API prompt generation; R and Excel for statistical analysis
Key findings
  • Anchoring bias was statistically significant across all 5 tasks in all LLMs
  • Framing effect was significant in 2 of 4 tasks
  • Human responses showed notably higher variability than LLM outputs in most tasks
Discussion
  • AI Safety: demonstrates the need for bias-aware design in AI decision-support systems
  • User Awareness: users consulting LLMs for important decisions should be informed of potential bias amplification
  • Model Development: opportunities to improve LLM training to reduce cognitive bias susceptibility
  • Fairness & Trust: raises questions about AI-assisted decision-making when biases can be introduced through input framing
Tools
Python R Excel Statistical Analysis API Integration Data Visualisation
Experiment

Gamification: Examining Whether Presenting Tasks as Games Increases Engagement

Does simply framing an activity as a "game" rather than a "task" change how people engage with it? This experimental study investigated the psychological mechanisms behind gamification through controlled experimentation and mixed-methods analysis.

Feb 2022 – May 2022  ·  University of Copenhagen · MA Research Project

›
Research question

How effective is presenting a task as a game vs. a task, as measured by participant engagement and intrinsic motivation?

Design
  • Game group: task framed with game-like language and presentation
  • Task group: neutral, straightforward instructions
  • 19 participants total; 2 independent groups
Methodology
  • Lexical Decision Task administered via Gorilla Experimental Builder
  • Intrinsic Motivation Inventory (Likert scale) measuring interest, competence, and effort
  • Open-ended questionnaire responses coded in NVivo
  • Statistical analysis in R; visualisations in Excel
Key findings
  • Game group made significantly more attempts which means higher engagement
  • Game group showed subconscious enjoyment even when consciously unaware of it
  • Task group showed measurable absence of subconscious enjoyment
  • Qualitative analysis revealed distinct conceptual frameworks by framing condition
Discussion
  • Educational Design: reframing learning activities as games could enhance engagement without complex gamification systems
  • UX: interface designers can use linguistic framing to influence user perception and motivation
  • Workplace: task presentation strategies could improve engagement in repetitive activities
  • Psychology: evidence for subconscious processing of contextual framing in task engagement
Tools
Gorilla Experimental Builder R Excel NVivo Asana Statistical Analysis

Get in touch

Let's connect

Email angelinademirtzoglou@gmail.com
LinkedIn angelina-demirtzoglou