AI where accuracy matters
Who measures if your AI works?
We do. We train teams, test AI tools, and turn large amounts of text into clear answers — each one checked against human judgment.
Three Ways We Help
Each one ends the same way: results checked against human judgment, a written report you can hold us to, and one fixed price agreed before work begins.
Train Teams
Hands-on training that teaches your staff to use AI on their own work — and to catch what it gets wrong.
- Half-day, full-day, or two-day formats
- Built around your team’s own tasks and data
- No programming background needed
Test AI Tools
Before staff or the public rely on a chatbot or AI tool, we test whether it gives correct answers — and says “I don’t know” when it should.
- Test questions drawn from real use
- Pass/fail thresholds agreed before testing
- Independent report in plain language
Analyze Text
We use AI to read thousands of comments, reports, or tickets and tell you what they say — measured, not guessed.
- Models built for your question
- Findings with a full methods appendix
- One-time studies or ongoing tracking
Nonprofits, newsrooms, and schools pay 25% less. Larger or combined work is quoted as one fixed price — never an hourly meter.
Not sure which fits? Tell us what you’re trying to do and we’ll point you to the right starting point — or tell you honestly that you don’t need us.
How an Engagement Runs
Whether we’re training a team, testing a tool, or analyzing text, we treat AI output as a measurement — and measurement has a discipline. Every engagement follows the same four moves.
Step 1
Define the question
A working session to agree what you need to know — and what a good-enough answer looks like — before any work starts.
Step 2
Build the instrument
The test set, the coding scheme, the model, or the training plan — built for your question, with the right method for the job.
Step 3
Run at scale
The instrument runs on everything — thousands of documents or hundreds of test questions — consistently, item by item.
Step 4
Validate & deliver
We check the machine against human judgment, report where it fails, and deliver findings with a methods appendix — so the analysis survives a skeptical board, editor, or reviewer.
Why Rihlah
AI made it easy to produce answers — and just as easy to be confidently wrong. The difference between an answer and a measurement is method. That is what we bring.
Research-Grade Methods
Led by Dr. Mohammad Yousuf — an award-winning professor who teaches computational text analysis and whose research on misinformation, media, and language technology has been cited more than 600 times and published in the world’s top computer-science venues.
A Journalist’s Skepticism
Eight years in national and international newsrooms taught our founder to verify before he trusts. Every model output — including AI’s — is held to the standard he learned on the copy desk: nothing is true until it’s verified.
A Builder’s Hands
We don’t resell someone else’s dashboard. We build the pipelines, the models, and the test sets ourselves — end to end — and hand them over documented, so what we build keeps working for you.
A Note on AI
Modern AI is the most powerful text-reading technology ever built, and we use it daily. But reading fast is not the same as measuring well: language models drift, miss, and confidently invent. That’s why every Rihlah analysis and every AI test is validated against human judgment, reported with its limitations, and documented so it can be reproduced. Our founder researches these systems for a living and publishes on their failures as well as their strengths. When the answer matters — to a board, a funder, an editor, or a court — you deserve analysis that survives the question: how do you know?
Who We Work With
Any organization that relies on AI or on words — which is nearly every organization. If your decisions depend on what people wrote or said, or on what an AI tool tells them, we speak your language. For example:
Newsrooms & Media
Coverage and audience feedback analyzed, and newsroom teams trained to use AI without losing accuracy.
Education & Research
Open-ended surveys and program evaluations analyzed at scale, and faculty trained to use AI well.
Businesses
Chatbots and AI tools tested before customers rely on them; customer feedback turned into signals leadership can act on.
Nonprofits & Public Agencies
Public comments, community input, and constituent-facing AI tools — measured, tested, and reported defensibly.
Not on this list? The industry matters less than the pattern: text you collect but never analyze, or AI you use but have never tested. Tell us what you’re working on.
Funders, associations & program leads
We deliver analyses and training cohorts on behalf of foundations, associations, and program offices — one contract, many organizations served.
Multilingual corpora
Analysis doesn’t stop at English. We build and analyze corpora across languages — so international coverage, communities, and sources count too.
Based in Albuquerque — delivering across New Mexico, nationwide, and internationally.
Get Started
Five minutes, a few questions. Dr. Yousuf reads every submission personally and responds within two business days.