28/08/2026
The first thing Madam Zainab checked was not her daughter's temperature. It was the network bars on her phone.
First, One bar.
and Then none.
She lifted the phone toward the window.
One bar again.
Then it disappeared.
Her daughter, Amina, was lying on a thin mattress beside her.
Hot.
Weak.
Quiet in a way that made her mother uncomfortable. Children are rarely that quiet unless sleep has defeated them or something is wrong.
Madam Zainab placed the back of her hand on Amina's forehead again.
Still hot.
Outside, evening was settling over the small community. NEPA had taken the light since afternoon. Nobody asked when it would return.
You do not ask NEPA questions you already know nobody can answer.😁
The small generator belonging to the chemist down the road had also gone silent.
Fuel was expensive.
The nearest primary health centre was several kilometres away. And the health worker who usually stayed there was not always available at night.
Madam Zainab picked up her phone again. An old Android device. The screen had cracks near one corner.
Battery: 11%.
Data balance: almost finished.
She opened the health application her nephew in Abuja had installed for her months earlier.
White screen.
A spinning circle.
Then: No Internet Connection.
She tried again.
Nothing.
Off data.
On data.
Airplane mode.
Off again.
Held the phone toward the window.
Nothing.
And there was something almost insulting about it. Somewhere in Lagos, Abuja, London, California or Singapore, artificial intelligence could write reports, analyse images, translate languages and answer complicated questions within seconds.
But inside Madam Zainab's room, all that intelligence had just been defeated by three words:
No Internet Connection.
Amina coughed.
“Mummy.”
“Yes, my baby.”
“My body is paining me.”
Madam Zainab's stomach tightened.
She did not need artificial intelligence to give her the name of her daughter's disease.
She did not need a 20-page medical report.
She needed somebody medically trustworthy to help her answer one question: Should I wait until morning, or should I start looking for help now?
Is her breathing normal?
Is there something dangerous I should be watching for?
Is this getting serious?
That was all.
But the doctor was far away.
The internet was gone.
The battery was dying.
And the beautiful healthcare application sitting on her phone had become nothing more than a colourful icon.
This is one of the uncomfortable truths about digital health in Africa.
Sometimes we design technology for the Africa we wish existed.
The smartphone we wish everybody owned. The 4G connection we wish covered every village. The electricity we wish never disappeared. The data bundle we wish everybody could afford.
Then we call the product innovative. But innovation that disappears when the network disappears has not yet met the people who need innovation most.
Madam Zainab does not care how powerful the AI model behind an application is.
She cares whether it can help Amina tonight.
Her neighbour Musa heard what was happening and came over.
He looked at Amina.
“Have you called anybody?”
Madam Zainab shook her head.
“Which network I wan use call?”
Musa brought out his own phone.
Not a smartphone.
No touchscreen.
No expensive applications.
Just buttons.
The kind of phone many technology presentations have quietly forgotten exists. He pressed a health short code he had been shown during a community programme.
Something like: *347 #
A menu appeared.
1. Child is sick
2. Pregnancy concern
3. Adult illness
4. Medicine question
5. Emergency
Madam Zainab looked at the small phone.
“This one?”
Musa smiled.
“Na phone wey the work be phone.”
Now imagine that behind that ordinary menu sat something far more intelligent.
Not just a list of pre-written messages.
Not:
“Press 1 for fever.”
“Press 2 for headache.”
Imagine Madam Zainab could send: My daughter hot since afternoon. She weak and no dey eat.
No video call.
No 200MB application.
No fancy interface.
No 4G.
Just text.
The message reaches a healthcare triage system. Behind the scenes, an AI helps interpret what she has written, even if she mixes English, Pidgin or familiar local expressions. But the system does not send Madam Zainab six paragraphs of medical language.
It asks the important questions.
Can Amina breathe normally?
Is she conscious?
Can she drink?
Has she had convulsions?
Is she getting rapidly worse?
Then it returns only what matters next.
Short.
Clear.
Within the limits of SMS.
And, importantly, designed to support triage and referral, not pretend that a text message has suddenly become a doctor.
Something like: If Amina is struggling to breathe, unconscious, convulsing, unable to drink, or rapidly worsening, seek urgent medical help. Reply 1 to continue.
A large amount of computer reasoning might happen behind that message. But Madam Zainab receives one small cup of useful information from the bucket.
That is the idea behind semantic compression.
Do the heavy processing elsewhere.
Send the person only what matters.
Let's Call this concept: SabiNexa AI.
This innovation is not simply putting AI behind SMS. This innovation is accepting scarcity as part of the design.
Weak data.
Old phones.
Short batteries.
Expensive internet.
Pidgin.
Mixed languages.
Poor connectivity.
Because people should not have to learn how technology likes to speak before technology can understand how they speak.
But there is still a problem.
A big one.
USSD and SMS may work without mobile data. But they still require some mobile network signal. So, what happens when there are no network bars at all?
Then even the cleverest SMS system has nowhere to send its message. And that means we have to change the question again.
What if the intelligence did not always need to travel across the internet? What if some of it lived inside the device?
Let's say, the following morning, Madam Zainab manages to get Amina to the small health post.
Hauwa, a community health worker, is there.
She carries a modest Android phone.
Nothing special from the outside. But inside it is a small AI model that was downloaded when the device previously had connectivity. The model can perform specific approved tasks locally.
No internet required at the moment of use.
Hauwa assesses Amina.
She follows the normal clinical process.
She performs the appropriate rapid test. Then, where the tool has been clinically validated for that purpose, she uses the phone camera to help interpret the test strip. The image does not need to travel hundreds of kilometres to some computer in a distant city.
The phone processes it locally.
Offline.
Hauwa still makes the clinical decision within her training and established protocols.
The AI is supporting her.
Not replacing her.
The record is saved.
Later, when connectivity becomes available, the information can synchronize with the wider health system.
Let's call this basic idea of SabiNexa: EdgeMed. “Edge AI” sounds like something meant for engineers until I explain it simply.
Most internet AI works like sending your question to a very intelligent person who lives far away. If the road to that person closes, you cannot reach them. Edge AI is like teaching someone in your own house enough to perform certain tasks.
Now, even when the road closes, some intelligence remains with you.
Suddenly: offline does not have to mean useless.
But then somebody asks a better question. A question we should have asked before celebrating.
Hauwa's phone has collected Amina's information. It has collected several other patient records too.
There is no network yet.
Everything is waiting for synchronization.
So what happens if the phone disappears?
Not the network.
The phone itself.
Three weeks later, Hauwa travels with another health worker to an outreach community.
They finish late. On their way back, armed men stop vehicles on the road.
People scatter.
Phones are taken.
Bags disappear.
In the confusion, Hauwa loses her device. Whether it was stolen, dropped while running, or lost somewhere along the road no longer matters.
The phone is gone. And suddenly the question is no longer: “Can EdgeMed work without internet?”
The question is: “Can the healthcare system survive without Hauwa's phone?” Because if twelve patients' medical records existed only inside that one device, we have simply created another fragile system.
We solved the internet problem, and created a phone problem.
This is where innovation has to become wiser. Because technology built for difficult environments must expect difficult things to happen.
Phones are stolen.
Phones fall inside water.
Motorcycles crash.
Batteries swell.
Screens break.
Workers get transferred.
People forget passwords.
Communities experience violence.
And in some places, health workers themselves face kidnapping, attacks and death while travelling through insecure areas. You cannot call a system resilient if one missing phone can erase the healthcare history of an entire outreach visit.
So EdgeMed WOULD need another principle: No vulnerable phone should ever become the only memory of a patient's healthcare encounter.
Imagine that before Hauwa leaves the health post, her phone automatically talks to a small box installed there.
The box is not glamorous.
Perhaps it is the size of a small router.
Solar powered.
Battery backed.
It does not even need the internet to do its first job.
Call it the: Edge Health Vault.
When Hauwa's phone comes within range, the new patient records are copied into the vault. But not like copying family photographs to a flash drive. The information is locked using strong encryption.
To anybody without permission, it should look like meaningless digital noise. The vault now keeps one protected copy.
Hauwa's phone carries another.
If her phone is lost later that day, the health centre has not forgotten the people she treated.
The device disappeared.
The healthcare memory did not. And what if Hauwa spends three days away from the health post?
Then perhaps the phones carried by authorised health workers can help one another. Imagine Hauwa meets another community health worker during an outreach. Their devices quietly recognize each other.
Hauwa's phone creates an encrypted backup of the day's records. The second phone stores the sealed package.
That worker cannot open Hauwa's patients' information. It is like carrying somebody else's locked envelope.
Later, whichever authorised device reaches connectivity first can deliver the encrypted information to the central system.
Think about what just happened.
There was no internet.
There was no telecommunications tower nearby.
But information still moved. The people temporarily became part of the network.
We already understand this idea in Africa. When formal systems fail, human networks carry things.
“Give this parcel to your brother when you reach Enugu.”
“Please take this message to Mama when you enter town.”
“Help me send this document with the driver going to Abuja.”
The technology is simply borrowing OUR old human idea: If the road is unavailable now, protect the message and let the next trustworthy journey carry it forward.
But backing up the information solves only half the problem. What about the phone now sitting in the hands of whoever stole it?
Amina's health information is personal. So are the records of every other patient Hauwa treated. A stolen device should not become an open medical filing cabinet. The information stored inside EdgeMed should therefore be encrypted from the beginning.
Not after theft.
Before theft.
Patient names, test results, photographs and clinical notes should not sit casually inside ordinary folders.
The application itself should lock.
The data should be encrypted.
And where appropriate, the patient's identity can be separated from the clinical information.
Instead of a thief seeing: Amina Yusuf — Age 7 — Village X — medical result — mother's phone number the stolen device may contain protected patient codes whose identities can only be reconstructed by an authorised health system. And after information has safely synchronized, the phone does not need to carry years of everybody's medical history.
Old sensitive data can be removed from the field device according to proper retention rules.
The phone should carry what Hauwa needs. Not become a travelling warehouse of an entire village's secrets.
Remote wiping can also help. If the stolen phone ever connects again, the health system could revoke its access and erase protected application data. But we should not fool ourselves.
What if the phone never reconnects? Then there is no magic button in Abuja that can travel through thin air and wipe it. So security cannot depend on what we hope to do after the robbery.
The protection must already exist before the robbery.
That is what being proactive means. Then there is an even more important issue.
We have spent all this time protecting the phone.
What about OUR dear Hauwa? 😊
Because Hauwa is more valuable than every device and every database in the system. A digital health programme operating in an insecure environment cannot pretend worker safety is somebody else's problem.
Imagine Hauwa is expected back from an outreach by 6 p.m.
At 6:30, nobody has heard from her.
At 7 p.m., still nothing.
The system should not wait until the following afternoon before somebody asks: “Has anybody seen Hauwa?”
Where local conditions permit, field programmes can include safety check-ins.
Known travel plans.
Emergency contacts.
Escalation procedures.
Team travel where necessary.
And discreet location or distress features where they genuinely improve safety and do not create new risks.
Technology cannot defeat banditry.
An application cannot promise a health worker that nothing bad will happen. But good system design can at least refuse to behave as though the possibility does not exist.
This changes EdgeMed again.
It is no longer simply: offline-first.
It becomes: survivable.
A survivable health system assumes the electricity can fail.
The network can disappear.
The battery can die.
The phone can break.
The phone can be stolen.
The health worker may lose access to the device.
The internet may remain unavailable for days. And sometimes the person carrying the technology may face danger themselves.
That system therefore protects three things independently:
The patient.
The patient's information.
The health worker.
There is something else honesty requires us to admit. Suppose Hauwa sees a patient deep inside a remote settlement.
She creates the record. Two minutes later, before the phone has backed up to another trusted device or vault, it falls into a river and is destroyed.
Can artificial intelligence recover that record?
No.
There is no magic here.
If information existed in only one place and that place is destroyed before a copy is created, the information may be lost.
That is not an African problem.
It is not even an AI problem.
It is simply how information works. So the engineering challenge becomes reducing the period when there is only one copy.
From days
to hours.
From hours
to minutes wherever possible.
The goal is not pretending failure can never happen. The goal is making one failure less capable of destroying everything.
Now, let's return to Madam Zainab.
Her phone is still scratched.
NEPA still takes light.
The network still behaves like a visitor who sometimes comes without warning and leaves without saying goodbye.
The road to the larger town is still bad.
Hauwa still works in a health post that nobody would mistake for a hospital in London.
Nothing about the community has suddenly become futuristic. And yet something important has changed.
When basic network exists, Madam Zainab can reach SabiNexa through something as ordinary as SMS or USSD.
When internet data disappears, the service does not necessarily disappear with it.
When there is no network at all, Hauwa's device can still perform specific offline functions.
When electricity fails, solar charging and power banks keep essential devices alive.
When records cannot immediately reach the central system, they wait securely.
When another authorised device becomes available, encrypted information can move closer to safety.
When Hauwa returns to the health post, the Edge Health Vault keeps another protected copy.
If her phone disappears, the patients do not automatically disappear from the system with it.
If somebody steals the device, they should not inherit the village's medical secrets.
And if Hauwa herself does not return when expected, somebody should notice.
Now we are no longer designing a clever application. We are designing around real life.
That may be one of the biggest lessons for African innovation. We do not necessarily need smaller copies of every technology designed elsewhere. Sometimes we need systems built from their first line of code around the things people elsewhere may call “edge cases.”
No electricity.
No network.
Expensive data.
Cheap phones.
Broken roads.
Multiple languages.
Device theft.
Long distances.
Accidents.
Insecurity.
But for millions of Africans, these are not edge cases. They are simply the environment.
Offline is not an exception.
Sometimes offline is Tuesday. And losing a phone is not an impossible disaster scenario.
Sometimes it is what happened last month. So perhaps the future of African AI healthcare will not always look impressive on a conference screen.
Perhaps sometimes it will look like this:
A short code.
A text message.
A scratched feature phone.
A cheap Android device.
A tiny AI model living inside it.
A solar panel on the roof of a health post.
A small encrypted box nobody outside the clinic pays attention to.
Two health workers' phones exchanging sealed digital records when they come close.
A database waiting patiently for connectivity instead of demanding connectivity before doing anything useful.
And a mother who does not know what “Edge AI,” “semantic compression,” “encryption” or “store-and-forward architecture” means.
She does not need to.
What she understands is much more important:
Her daughter was sick.
She needed help.
The light was gone.
The data was almost finished.
The network disappeared.
And somehow, the healthcare did not disappear with it.
That is the standard.
Not technology that performs beautifully when everything around it works. But technology that still knows what to do when almost everything around it doesn't.
Because the real test of innovation is not whether it survives the demonstration.
It is whether it survives the village.
The darkness.
The bad road.
The empty data balance.
The missing network bar.
The stolen phone.
The broken screen.
The unexpected journey. And all the messy realities between a good idea and the human being whose life may eventually depend on it.
That is the kind of AI worth building.
This story is part of my UNITAR Empower Africa Programme: Japan–Africa Partnership for Youth and Women’s Economic Resilience project.
If you’re building for Africa, don’t just ask whether the technology works. Ask whether it can survive the reality of the people it is meant to serve. If this problem speaks to your work, I’d love to connect, exchange ideas, and explore what we can build together.
