And why the future of health engagement won’t look like the past.
If you sit long enough with community health workers, caregivers, and health facility staff, the real economics of public health reveal themselves quickly. It’s in the tiny frictions no one budgets for: a mother with a sick child who texts a nurse at 9:47 PM because the clinic is closed, a community health worker hurrying between households, or a father who says quietly, “Hata sijui kama nitafika kesho. Naweza pata dawa imeisha.”
These are the decisions – the micro-moments – that quietly shape population health. And every one of them has always carried a cost.
More questions meant more staff.
More follow-up meant more time.
More communication meant more budget.
More data collection meant more burden on community health workers.
But something is changing.
Not because technology is new. Technology is always new.
But because for the first time, the economics of digital health have shifted.
What We’re Learning From the Field
At DPE, we build like learners: listen deeply, test small, improve fast.
And across counties, households, and programs, one lesson keeps resurfacing:
Most barriers in public health are not only clinical. They’re economic.
Economic for families.
Economic for health workers.
Economic for health systems.
Families face the cost of time, transport, uncertainty.
Community health workers face the cost of workload, documentation, follow-up, inadequate support.
Counties face the cost of staffing, campaigns, and digitization.
Health systems face the cost of scaling anything at all.
This is the context where agentic AI enters – not as a gadget, not as a chatbot, but as a new kind of infrastructural layer that can finally bend the cost curve.
Agentic AI simply means AI that doesn’t just answer. It acts.
It can plan, follow steps, make decisions within guardrails, and support a family or community health worker through a task from start to finish.
It’s the difference between a tool that waits for instructions and a system that helps carry the load.
A health worker in Kilifi Said Something We Haven’t Forgotten
“Sasa mkiweza kusaidia hawa wazazi usiku, mimi nitaweza kusupport watu wangu mchana.”
(“If you can help caregivers at night, I can better support my people during the day.”)
This is the economic heart of agentic AI: every solved micro-task lowers pressure on the system.
Every automated interaction reduces human burden.
Every nudge delivered at midnight prevents a crisis at 10 AM.
A New Framework for Public Health Economics
The Three Levers of Agentic AI
- Leaner Interactions
Tasks like answering questions, checking danger signs, and sending reminders become instantaneous and cost-light. - Lighter Workloads
Health workers only intervene where they’re needed – because agents triage, classify, and route early signals. - Lower Marginal Costs
Once the infrastructure is built, every additional household becomes cheaper to serve, not more expensive.
This is the opposite of how health systems normally scale.
Why Traditional Digital Systems Hit a Ceiling
Most digital health tools grow more expensive over time: more staff, more licenses, more devices, more reporting, more parallel workflows.
Agentic AI works differently. It can:
- Understand intent
- Interpret symptoms
- Personalize behaviour change messages
- Classify community feedback
- Support community workflows
- Escalate danger signs
- Document interactions
- Learn from every message
And it can do these things for thousands of people at once – day, night, weekends, holidays – without adding another human shift. This shifts the core economic unit of digital health from: people to tasks and outcomes.
Agentic AI replaces rigid annual planning with real-time resource allocation. Health systems no longer need to pre-pay for fixed capacity or staff-heavy campaigns – they can scale exactly when demand emerges. Instead of budgeting for “messages sent,” they can budget for work actually done: triages completed, follow-ups resolved, risks escalated, adherence nudges delivered. It is the shift from planning for capacity to paying for impact.
What This Means for Health Budgets (and everyone)
Fixed Costs (High upfront, predictable)
- Engagement infrastructure
- Localizing AI models
- Health interoperability pathways & integrations
- Safety built in
- Multi-language support
- Household support
Variable Costs (Decline quickly as scale grows)
- AI Model inference
- Personalized routing
- Follow-up sequences
- CLM classification
- Compute during outbreaks
This flips the traditional ratio:
Instead of 80% variable cost, 20% fixed…
Agentic AI becomes 80% fixed, 20% variable.
Meaning:
The more people the system serves, the cheaper each interaction becomes.
And guess what? Mobile money had the same curve.
The Real Value: Turning Engagement Into Outcomes
Agentic AI enables things we previously treated as “too expensive to scale”:
1. Hyper-responsive support
A mother can ask, “Is this a danger sign?” at 11 PM, and get the right guidance instantly.
2. Community intelligence without burden
Every incoming message becomes structured signal data for health system accountability, all automated.
3. Faster, earlier detection
If hundreds of caregivers mention similar symptoms or stock challenges, the pattern is visible on day one, not month two.
4. Health worker empowerment
Agents help with documentation, reminders, and escalation logic, freeing health workers to spend more time in real conversations.
5. Cost avoidance
Early nudges prevent expensive emergencies, late presentations, and unnecessary facility visits.
Agentic systems don’t just scale, they learn. Every message improves future personalization. Every symptom report sharpens triage. Every health case enriches community support.
Traditional ICT systems are cost centres. Agentic systems are learning assets. In most systems, usage increases cost. In agentic systems, usage increases value. This is what we all need to design for.
Why This Matters Now
Africa’s public health systems face simultaneous pressures:
- workforce shortages
- climate-health related shocks
- disease complexity
- funding fluctuations
- growing populations
- rising expectations from citizens
Agentic AI isn’t here to replace anyone. It’s here to relieve the pressure – to make care more reachable, more predictable, more human.
A future where:
Every household is within reach.
Every health worker is supported.
Every community voice is heard.
Every health system becomes adaptive, not reactive.
That future is possible, not because AI is powerful, but because agentic AI finally realigns the economics of care with the needs of people.
At DPE, we are building toward this future deliberately.
Our work is organized around three layers that make agentic public health possible in the real world:
(1) The Engagement Layer, where families, caregivers, and health workers interact in the simplest ways they already know – WhatsApp, SMS, and voice. Our InfoAFYA applications are based around this structure.
(2) The Intelligence Layer, where agents understand intent, plan tasks, route triage, support health workers, and turn every message into structured insight.
(3) The Infrastructure Layer, our Interch™ stack – the ontologies, health integrations, safety guardrails, and national data pathways that make this system reliable, governed, and interoperable.
These layers work together to create something Kenya has never had before: a health engagement system that is always on, always learning, and always lowering the cost of reaching people who were previously the hardest to reach.
This is the agentic future we’re building – one where every household can be supported without adding more burden to the people who already give so much.
An Invitation
If you work in county health, at a facility, in a community-based organisation, in a school, or run health programs – and you’re curious about what this economic shift could unlock, let’s talk.
We’d love to learn from your experience, share what we’re seeing on the ground, and build the next chapter of health engagement together.