AI for Energy and Utilities in Kenya
Faults reported properly and field teams sent to the right place.
What is energy?
AI for energy covers fault reporting and triage, customer meter and payment queries, field team dispatch and reconciliation, for utilities, distributed solar providers and independent operators.
Energy customers contact you about three things: no power, a meter or token problem, and a bill or payment question. The volume is concentrated and the questions repeat.
The operational cost sits in dispatch. A field team sent to the wrong place, or sent to a fault that was actually a token problem, costs a vehicle, a crew and a day.
Sorting a report properly before anybody is dispatched is where the money is, and it depends on getting usable information from a customer who is annoyed and in the dark.
Why triage comes before prediction
Duara works with distributors, solar home system providers, mini grid operators and independent power producers from Nairobi.
We prioritise triage over prediction. Predictive maintenance is the headline application in this sector and it needs sensor data most Kenyan operators do not have, while better fault triage saves real money from next week using information you already collect.
Who this is for
- Fault reports arrive with poor information
- Field teams are dispatched to the wrong problem
- Token and payment queries dominate contact
- Customer contact peaks during outages
Industries we serve
- Electricity distribution
- Solar home system providers
- Mini grid operators
- Independent power producers
- Fuel and gas distribution
- Water utilities
Problems this solves
These are triage and dispatch problems, which is where operational cost concentrates in this sector.
- Fault reports with too little detail. A customer reports no power and nobody establishes whether it is the supply, the meter or a token before a crew is sent.
- Crews sent to the wrong problem. A dispatch to a token issue costs a vehicle, a crew and a day, and the customer still has no power.
- Token and payment queries at volume. The same questions about tokens, balances and payments arrive constantly and consume the whole contact centre.
- Contact spiking during outages. Everybody calls at once, the queue collapses, and customers with unrelated genuine faults cannot get through.
What we deliver
Structured fault reporting that establishes what is actually wrong, dispatch informed by that triage, routine token and payment queries handled automatically, and outage communication that reduces the call spike.
Structured fault reporting over phone and message
Triage before dispatch to identify the real problem
Token, balance and payment queries handled automatically
Outage updates that reduce inbound contact
Field team dispatch with better information
Reporting on fault causes and repeat locations
Features
- Triage before dispatch. Establishing whether it is supply, meter or token before a crew is sent, since a wrong dispatch costs a vehicle and a day.
- Questions answered from your own systems. Token and balance queries answered from real account data rather than generic responses.
- Outage communication that reduces calls. Customers told about a known outage stop calling, which keeps the queue open for unrelated faults.
- Prediction only where data supports it. Predictive maintenance needs sensor data most operators do not have, and we say so rather than selling it.
- Works on basic phones. Fault reporting over calls and SMS, because a customer in the dark is not opening an app.
Benefits
- Fewer wasted dispatches. Triage before sending a crew is the largest single saving available in most distribution operations.
- Contact centre capacity recovered. Routine token and payment queries stop consuming staff who should be handling genuine faults.
- Outage queues stay usable. Proactive communication during an outage keeps lines open for the customers with unrelated problems.
- Repeat faults become visible. Cause reporting shows which locations and equipment keep failing rather than treating each report as new.
Technologies we use
- Structured fault intake. Collecting usable detail from a customer over call or message before dispatch.
- Triage logic. Distinguishing supply faults from meter and token issues using account and network data.
- Account query handling. Answering token, balance and payment questions from your own systems.
- Outage notification. Proactive messaging to affected customers to reduce inbound contact.
- Fault pattern reporting. Identifying repeat locations and equipment across reports.
Predictive maintenance is the application everybody asks for and the one most Kenyan operators cannot support, because it needs sensor data that does not exist. Triage uses information you already collect and pays back immediately.
How the process works, step by step
Energy projects usually take six to ten weeks, starting with whichever costs more, wasted dispatches or contact volume.
-
Step 1
Free operations review
We measure wasted dispatch rate and what proportion of contact is token and payment queries.
-
Step 2
Triage first
Fault sorting goes first where wasted dispatches are the larger cost, since it pays back immediately.
-
Step 3
Account queries added
Token and balance handling follows, which frees the contact centre for genuine faults.
-
Step 4
Outage messaging switched on
Proactive updates reduce the call spike that makes outages worse operationally.
-
Step 5
Pattern reporting
Repeat fault analysis follows once enough structured reports have accumulated.
Quality assurance and how we work
The commitments attached to systems affecting power supply and customer safety.
- Safety reports always reach a person. Anything involving a fallen line, a burning smell or an injury goes straight to staff without triage.
- We do not sell predictive maintenance without the data. It needs sensors most Kenyan operators do not have, and selling it anyway would take money for something that cannot work.
- Triage informs dispatch, it does not decide it. The crew decision stays with your control room, which knows things the system does not.
- Account answers come from your systems. A wrong token or balance answer damages trust more than no answer, so nothing is generated speculatively.
- We report wasted dispatches honestly. Including after implementation, because that figure is the honest measure of whether triage worked.
Frequently asked questions about energy
Can you do predictive maintenance?
Only where the sensor data exists, and for most Kenyan operators it does not. Selling it anyway would take money for something that cannot work.
What should we do instead?
Fault triage. It uses information you already collect and it saves real money from the first month.
How does triage save money?
By establishing whether a fault is supply, meter or token before a crew is sent. A wrong dispatch costs a vehicle, a crew and a day.
Does the system decide dispatch?
No. It informs the decision and your control room decides, since they know things the system does not.
What about safety reports?
Anything involving a fallen line, a burning smell or an injury goes straight to a person without triage.
Can it answer token queries?
Yes, from your own account systems. A wrong token answer damages trust more than no answer at all.
Will it work during an outage spike?
That is one of the main purposes. Telling affected customers about a known outage keeps the queue open for unrelated genuine faults.
Do customers need a smartphone?
No. Reporting works over calls and SMS, because a customer sitting in the dark is not opening an app.
Can it work for solar home systems?
Yes, and token and payment queries are usually the dominant contact type for those providers.
What about mini grid operators?
Yes, and smaller operators often gain most since they have the least contact centre capacity to absorb spikes.
How long before triage is running?
Six to ten weeks, starting with whichever costs you more, wasted dispatches or contact volume.
Does it integrate with our billing system?
Yes, since answering balance and token questions requires real account data rather than generated responses.
Can it identify repeat fault locations?
Yes, once structured reports accumulate, which usually shows a small number of locations generating most of the work.
Will it reduce our field team?
That is your decision. Most operators redirect crews toward the backlog rather than reducing, since triage reveals work that was queued.
Can customers report faults in Swahili?
Yes, and for fault reporting it matters, since a customer under stress will use their first language.
Can it handle meter reading?
Where you have the data feeds, yes, though physical reading is a separate operational question.
What if our fault data is poor?
Then structured intake is the first improvement, since better triage depends on better information at the point of report.
Does it help with revenue collection?
Indirectly, through payment queries and reminders. Disconnection decisions stay entirely with you.
Who can see customer account data?
Only your authorised staff. Account records stay in your systems and train no public AI model.
What does an energy project cost?
Customer numbers and contact volume determine it, quoted after the operations review.
Glossary of terms
Utility terms in plain language.
- Triage
- Establishing what is actually wrong before dispatching anybody.
- Token
- The prepaid credit code applied to a meter, and a frequent source of customer contact.
- Wasted dispatch
- Sending a crew to a problem that did not need them, costing a vehicle and a day.
- Outage
- A supply interruption, during which customer contact spikes sharply.
- Mini grid
- A local generation and distribution network serving a defined area.
- Predictive maintenance
- Anticipating equipment failure from sensor data, which requires sensors to exist.
Related services in Industries
Talk to us about energy
Work out what proportion of your dispatches last month turned out to be token or metering issues. That is the triage case.
Start your project