KhulaSense fuses satellite imagery, drone data, AI and IoT to forecast informal settlement growth — giving municipalities months of lead time to plan infrastructure, not years of catch-up.
South Africa is already one of the continent's most urbanised nations, and researchers expect that trend to keep climbing for the next two decades. Yet today's planning still leans on manual surveys, community reports and annual inspections — methods that routinely run more than a year behind what is actually happening on the ground.
When detection lags by a year or more, consequences compound:
KhulaSense continuously analyses satellite and drone imagery alongside historical growth, weather, population density, existing infrastructure, road access and utility consumption — then tells you what to build, where, and in what order.
Identifies new informal settlements and boundary expansion as they happen.
Converts structure density and imagery into reliable population estimates per zone.
Surfaces fire, flood and health risk zones so emergency planning can get ahead.
Forecasts water, electricity, sanitation and waste demand tied to growth projections.
Recommends road and emergency access routes into growing zones.
Prioritises capital investment by urgency, population impact and cost.
Turns raw model output into a short, readable brief any council member can act on — no data science degree required.
Re-scans every zone on a rolling schedule, so forecasts get sharper with every pass instead of going stale.
KhulaSense pairs artificial intelligence that reads imagery and forecasts growth with a network of physical sensors that check that forecast against reality. Everything runs on Amazon Web Services (AWS), and OpenAI's language models turn the technical output into plain English a municipal official can read in under two minutes.
Computer vision models scan every new satellite and drone image and compare it with everything captured before, so KhulaSense can tell the difference between an empty field last month and forty new rooftops this month. Time-series models then take that pattern and project it forward, the same way a weather model projects a storm's path from where it's been moving.
Small, low-power sensors sit on water lines, transformers and drainage points inside monitored zones. They don't predict anything — they simply report what is actually happening right now, so the AI forecast can be checked, corrected and trusted rather than taken on faith.
Here is the same pipeline described as a stack, from raw picture to funded infrastructure.
Sentinel satellite passes, scheduled drone flights and the live IoT network all generate raw images and readings around the clock.
Every image and sensor reading lands in Amazon Web Services, where purpose-built AWS tools store it, train on it and react to it automatically.
Once the AWS models have produced coordinates, risk scores and demand figures, OpenAI's language models read that structured output and write it up the way a good analyst would — in a short, plain-English brief, e.g. "Zone 14 needs two new water points within eight months." Officials can also simply ask a question in normal language, such as "which zones need urgent electricity upgrades this quarter?", and get a direct answer.
The finished brief appears on the live dashboard, ranked by urgency, so planning and finance teams can commit budget before a shortage becomes an emergency.
Think of it like this: satellites and drones are the eyes, taking a photo of every growing neighbourhood every few days. AWS is the brain's memory and muscle — it stores every photo and runs the maths that spots what changed. OpenAI is the brain's voice — it takes all that maths and explains it in plain words, the way a colleague would explain a report to you over coffee instead of handing you a spreadsheet. And the IoT sensors are like a second pair of hands on the ground, double-checking that what the eyes saw from the sky is really happening down on the street.
A visual walkthrough of the product, and what each stage actually means for a municipal planner.
Satellites, drones and ground sensors continuously watch monitored zones, the way a security camera watches a building.
AWS-hosted AI models compare each new image to the last, spot new structures, and project where growth is heading next.
IoT sensors confirm the prediction against real conditions, then OpenAI writes it up in a short, plain-English brief.
The municipality budgets and builds against a ranked, dated list — then the loop starts again on the next imagery pass.
Imagine KhulaSense as a weather forecaster — but instead of predicting rain, it predicts where people will need water pipes, electricity lines and roads next. It watches neighbourhoods from above the way a weather satellite watches clouds. It thinks about what it sees using artificial intelligence hosted on AWS. It explains what it found in plain language using OpenAI, so anyone can read it in under two minutes. And then the municipality acts on that information months before a problem happens, instead of reacting to it years after the fact.
For teams who want the full technical sequence behind each of the four stages above.
Khula means "grow" in isiZulu and isiXhosa. KhulaSense was built on a simple premise: instead of documenting settlements after they form, give municipal teams a reliable forecast of what's coming, early enough to budget, plan and build for it. KhulaSense is designed, built and operated by YENOUS (Pty) Ltd, a South African private company registered in Gauteng, so the product is backed by a locally accountable engineering and support team rather than an offshore vendor.
Continuous Sentinel satellite and drone imagery capture settlement footprints as they change, week over week.
Computer vision and time-series modelling convert raw imagery into forward-looking growth and demand forecasts.
Municipal teams receive prioritised, budgeted infrastructure recommendations inside a single live dashboard.
South Africa continues to experience rapid urbanisation, and municipal planning often struggles to keep pace with informal settlement growth. Research from UN-Habitat, the CSIR, academic journals and current news reporting all point at the same gap. Below are seven real, current sources, and what KhulaSense does about each issue they raise.
UN-Habitat figures cited in this 2026 review put South Africa's urban population at 67% as of 2023, on track to reach 80% by 2050 — well above the global average — even after three decades of housing policy reform.
ISS Africa reports that although South Africa has cut the share of people living in informal settlements by 7% since 2000, roughly 1.4 million more people now live in them, because overall population growth outpaced that improvement.
CSIR researchers demonstrated over a decade ago that high-resolution satellite imagery could classify informal settlement structures in Soweto in areas where no other reliable planning data existed at all.
A 2025 study on Slovo Park, Johannesburg, compared machine learning algorithms for classifying informal dwelling structures from drone imagery, confirming that fine-grained, structure-level detection is achievable at scale.
A 2025 Nature study mapped every street block across sub-Saharan Africa using high-resolution building and street data, showing a consistent gradient: infrastructure access drops the further a settlement sits from a formal city centre.
Covering the 2026 State of the Nation Address, this piece reports that President Ramaphosa named water the country's single most pressing issue, backing a new National Water Crisis Committee with R156 billion in funding, against a backdrop of roughly 47% non-revenue water loss nationally.
At a 2026 South African Human Rights Commission inquiry, Midvaal Local Municipality reported supplying 47 informal settlements by water tanker while carrying an estimated R1 billion backlog on reservoir infrastructure alone.
Most existing tools — GIS, satellite classification, drone surveys — are built to detect settlements that already exist. Very few connect that detection to a forward-looking, budgeted infrastructure plan.
Every tier includes the core AI forecasting engine, dashboard access and standard support.
Tell us about your municipality and we'll prepare a demonstration using imagery and infrastructure data relevant to your region.