METRIS™
Research
Our work is grounded in data, methods and reproducible evidence.
We combine real-world experience with statistical analysis, machine learning and simulation to understand how AI works in practice.
Innovation and research are in our DNA
Every method we use began as a research question. Our founder's work started in cancer and Alzheimer's research, where a finding only counted if someone else could check how it was reached, and that is still how we build. Each analysis is tested on public or simulated data before it touches a client's records, and the methods behind our findings are named, so they can be checked and rerun.
Our research continues in three areas:
- AI decisions in healthcare: how to show that a person truly made each AI-assisted decision.
- Analytics for AI oversight: anomaly detection, change-point detection and forecasting applied to how AI is actually used.
- People and AI working together: the theme of The AI-Human Synergy (2024), and of everything we build.
From the field
What we keep hearing from the field
More than 300 conversations with regulators, clinicians, researchers, business leaders and the people who use AI every day keep returning to the same few truths.
- Accountability for AI belongs to no single department. It shifts with the size of the organization, its contracts and the level of risk, and it shifts again when the AI is bought from a vendor rather than built in house.
- Small errors multiply quietly. When AI takes over steps like procedure coding, a mistake repeats across thousands of cases before anyone sees it. Human checks and sign-offs are what catch it.
- Clinical AI rarely fails for being wrong. It fails when using it takes a clinician longer than doing the work alone. Any proof of oversight has to cost people less time, not more.
- Regulated teams already drown in documentation. Engineers and reviewers spend hours recording every change by hand. Evidence should come from the work itself, not from more paperwork.
- Buyers judge AI by money and quality together. A health plan wants lower cost without lower quality, and a provider wants to grow without hiring more nurses. Showing that quality held is half the case.
Each of these points to the same need: analytics that show what AI is actually doing, drawn from the records an organization already keeps.
Publications and talks
- First-author abstract published in the Journal of Clinical Oncology (ASCO 2024 Annual Meeting)
- Author of The AI-Human Synergy (2024)
- Live trainings on measuring AI for O'Reilly Media
- Manuscript reviewer for PLOS ONE and Frontiers journals
METRIS™ Research
In development
We will publish what small businesses and startups actually pay for AI, how often its work is checked, and what changes when owners act. We will publish no numbers until they are defensible. Before anything is published:
- Only owners who agree to share, with no names.
- No group smaller than 10 businesses is ever shown.
- Every chart says who answered, how many, and what the numbers can and cannot tell you.
- Ranges and uncertainty are shown, not single numbers.