Find the signal
Discover
Spot repeatable opportunities across your creative creation teams - design, research and copywriting
AI Product & Experience Strategist
Designing for people Thinking in data Building for AI
Everything I do begins with human behaviour: I use research and data science to shape product strategy and design human-centred AI solutions.
What I offer
Sounds familiar
-
Returns
“We cannot show what our AI spend returned”
Pilots ran, but nobody measured the workflow before or after.
-
Adoption
“The tool is live and the teams still work the old way”
The workflow around the tool never changed.
-
Scaling
“It worked in one team. How will the others react?”
Adoption can be designed and anticipated: synthetic users grounded in your real data show the reactions first, and a trend agent turns the findings into formats teams engage with.
And that is where AI programmes stallNot because the technology fails, but because use cases, workflows and people were never designed together
Redesigning how people and AI work together
Understanding how work happens today is the starting point for designing what comes next. Research, operational data, and service design reveal where workflows can improve—and where AI can create meaningful value.
How it works
Understand the work
Map workflows, responsibilities, decisions, and dependencies across teams.
Find the patterns
Combine research and operational data to identify bottlenecks, rework, and opportunities.
Redesign the workflow
Define how people, processes, and AI agents should work together.
Validate the approach
Prioritise interventions, establish evaluation criteria, and define a path to implementation.
What you have afterwards
A practical blueprint for improving how your teams work, supported by evidence and a clear path toward AI-enabled delivery.
- Workflow evidence
Current-state workflow, behavioural patterns, bottlenecks, and dependencies
- AI opportunities
Prioritised interventions based on value, feasibility, and risk
- Future-state blueprint
Redesigned workflows, agent responsibilities, and human decision points
- Validation roadmap
Pilot scope, evaluation criteria, success metrics, and technical requirements
What this looks like in practice
Automotive
Synthetic Users for GEO
Starting situation
People now ask language models which car to buy. Customer research was abundant, but not in a form an agentic framework could use.
Result
An agentic framework that turns qualitative research into brand-neutral synthetic personas and a prompt catalogue by journey phase, used to audit brand visibility across AI engines.
Explore caseRetail
Feedback Intelligence
Starting situation
Thousands of customer reviews with praise and problems in the same sentence. Sentiment analysis flattened them to "neutral", and the fixable issues stayed buried.
Result
A single-pass GenAI decision layer that says what to fix, how urgent it is and who should act, on the same day.
Explore case
AI Product & Experience Strategist
Caro Milkez
During eight years at Accenture Song, I led experience and transformation work across automotive and enterprise products, and initiated research into synthetic users and agentic workflows.
Based in Munich, working with European and global clients. Happiest in the mountains, or with an espresso on the rocks.
- Accenture SongEight years, automotive and enterprise
- CertifiedCSPO Product Owner and Salesforce Business Analyst
- 15+ yearsProduct, UX and Design Leadership
- Data ScienceApplied AI and Data Science Recommendation systems
Ready to move from AI pilots to AI in operations?
When real value, redesigned work and the people who own it fit together, scaling stops being a leap of faith.
We check together whether a workflow assessment makes sense for your organisation.