A calm, simple explanation of how our AI makes recommendations — and why you can trust them.
Our recommendations are based on aggregated intelligence, not individual tracking.
Connectivity quality, transport availability, cost trends, safety indicators
Example: 'Nairobi connectivity is 75/100 and improving'
Where you're going, when you're arriving, what you need
Example: 'User is arriving in Nairobi tomorrow and needs data'
How well similar recommendations performed for similar situations
Example: '85% of arrivals found eSIM helpful'
Reliability, coverage, user satisfaction scores
Example: 'Zendit has 4.5/5 rating for Kenya coverage'
These are hard-coded prohibitions, not policies. Our system cannot access this data.
Real-time location tracking
Individual behavioral fingerprints
Personal browsing history
Device identifiers
Precise GPS coordinates
Social media activity
Our privacy enforcement is architectural — violations are impossible, not just prohibited.
Every recommendation follows a constitutional process:
Where are you? What's your travel state? What do you need?
What do our signals say about this city right now?
How strong is the match between your needs and available options?
Does this recommendation prioritize user value over revenue?
No feature grids. No option overload. One primary recommendation.
We show partner offers when they genuinely match your needs — not when they pay the most.
User value weight must be higher than revenue weight (enforced in code)
Maximum one partner offer per session (no spam)
Must match your city, timing, and actual needs
All partnerships disclosed transparently
Example: If you're arriving in Nairobi and need data, we might recommend an eSIM. If you're settled and researching, we won't interrupt with partner offers.
Privacy isn't a policy — it's enforced architecturally at every layer.
All data aggregated to city level or higher. No individual tracking.
No real-time surveillance. Minimum 24-hour delay on all data.
Minimum 20 data points required for any intelligence output.
Statistical noise added to prevent individual identification.
Constitutional guarantee: Our enforcement engine blocks privacy violations before they reach any database. Zero privacy incidents in 2 years of operation.
These are not aspirations — they're programmatically enforced rules.
Revenue weight can never exceed user value weight. Violations are blocked before reaching users.
Maximum one primary recommendation per session. No overwhelming with options.
Every recommendation includes why it's shown and confidence level.
Individual tracking architecturally impossible, not just prohibited.
Real data from our constitutional compliance system (updated daily).
Constitutional Compliance
→99.4%
Recommendations meeting all rules
Trust Violations
→0
Revenue > user value violations
Privacy Incidents
→0
All-time privacy breaches
Average Confidence
↑84%
Across all recommendations
Suppressed Outputs
→12
Blocked in last 30 days
Partner Frequency
→0.8
Offers per session (max: 1)
Full quarterly transparency reports available at africaai.app/privacy
Questions? Contact us