AI for Automotive in Kenya
Finding the right part for a vehicle nobody has a record of.
What is automotive?
AI for automotive covers parts identification and matching, service history, job card accuracy and workshop scheduling for dealers, garages and parts suppliers, in a market dominated by imported used vehicles.
The Kenyan vehicle parc is mostly imported used vehicles, from several markets, in specifications that were never sold here. That single fact causes most of the sector operational problems.
A customer arrives with a vehicle whose exact variant nobody can confirm. The parts counter guesses, orders the wrong component, and the vehicle occupies a bay for another week while the correct one is found.
Around that sits the job card, which is where workshop money is made or lost. Work done and not recorded is unbilled, and parts fitted and not captured are simply given away.
Why an uncertain match stays uncertain
Duara works with motor dealers, independent garages and parts suppliers from Nairobi on parts identification, job cards and workshop scheduling.
We treat an uncertain parts match as uncertain. A confident wrong part identification costs the part, the labour, the bay time and the customer relationship, which is far more than the few minutes saved by not asking a clarifying question.
Who this is for
- Wrong parts are ordered regularly
- Vehicle variants are hard to confirm
- Job cards miss work and parts
- Workshop scheduling is done on a whiteboard
Industries we serve
- Motor dealers
- Independent garages
- Parts distributors and retailers
- Fleet maintenance operations
- Commercial vehicle workshops
- Tyre and service chains
Problems this solves
These trace back to a vehicle parc that was never documented for this market.
- Variants nobody can confirm. An imported used vehicle arrives in a specification never sold here, and the parts counter is guessing from a chassis number.
- Wrong parts ordered. The wrong component costs the part, the labour already spent, the bay for another week and frequently the customer.
- Job cards missing work. Labour done and parts fitted are recorded incompletely, and everything missing from the card was given away.
- Scheduling on a whiteboard. Bay and technician capacity are managed informally, so jobs are promised into time that was already committed.
What we deliver
Parts matched against vehicle identification with uncertainty flagged rather than guessed, job cards captured completely, service history retained per vehicle, and workshop capacity scheduled against real availability.
Parts matched against vehicle identification
Uncertain matches flagged rather than guessed
Job cards capturing all labour and parts
Service history retained per vehicle
Workshop scheduling against real capacity
Reporting on wrong part orders and their cost
Features
- Uncertainty is flagged, not resolved. A confident wrong part costs the component, the labour, the bay and the customer, which dwarfs the minute saved by not asking.
- Built for imported variants. The parc here is imported used vehicles in specifications never sold locally, and any system assuming a clean local catalogue is useless.
- Job cards captured completely. Work and parts missing from the card were given away, and that is where workshop margin quietly goes.
- History kept per vehicle. What was done last time, on that vehicle, which matters more when the manufacturer record does not cover this market.
- Capacity scheduled honestly. Bays and technicians committed once rather than promised twice, which is what makes a delivery time real.
Benefits
- Fewer wrong parts. Flagging uncertainty rather than guessing addresses the most expensive recurring error in the sector.
- Workshop margin recovered. Complete job cards stop labour and parts being given away without anybody deciding to.
- Vehicles leave sooner. Bay time lost to a wrong part is the largest avoidable delay in most workshops.
- Promises that hold. Scheduling against real capacity means a collection time is a commitment rather than an aspiration.
Technologies we use
- Vehicle identification. Establishing variant from chassis number, registration and physical inspection details.
- Parts matching. Matching components to the identified variant with confidence stated rather than assumed.
- Job card capture. Recording labour and parts as work happens rather than reconstructing at invoicing.
- Service history. Retaining what was done per vehicle across visits.
- Workshop scheduling. Allocating bays and technicians against real committed capacity.
The identification problem here is genuinely harder than in markets with a documented parc. Any parts system trained on a manufacturer catalogue will be confidently wrong about a vehicle that was never sold in this market.
How the process works, step by step
Automotive projects usually take five to nine weeks, starting with whichever costs more, wrong parts or unbilled work.
-
Step 1
Free workshop review
We measure your wrong part rate and estimate what is missing from job cards.
-
Step 2
Identification tested
Parts matching is tested against your own recent jobs, including the ones that went wrong.
-
Step 3
Uncertainty thresholds set
What gets flagged for a person is agreed, and we set it conservatively for parts.
-
Step 4
Job cards first or parts first
Whichever costs more goes first, and for most workshops it is the job card.
-
Step 5
Scheduling added
Capacity allocation follows once job data is being captured reliably.
Quality assurance and how we work
The commitments attached to systems in a workshop.
- Uncertain part matches are flagged. A confident wrong identification costs the part, the labour, the bay and the customer. The minute saved by not asking is not worth any of that.
- We test on jobs that went wrong. Matching proved on straightforward vehicles tells you nothing about the imported variant that caused last month expensive mistake.
- Job card capture does not slow the bay. A technician will not use something that costs them time, and a half completed card is worse than the paper one.
- We do not diagnose faults. Deciding what is wrong with a vehicle is a technician judgement, and a wrong diagnosis has safety consequences.
- Service history stays with the vehicle. It is your record and it should transfer with the vehicle if the customer moves, which is a decision for you rather than a lock in.
Frequently asked questions about automotive
Why is parts identification hard here?
Because the parc is mostly imported used vehicles in specifications never sold locally, so a manufacturer catalogue is confidently wrong about them.
What happens when a match is uncertain?
It is flagged for a person. A confident wrong part costs the component, the labour, the bay and often the customer.
Can it just pick the most likely part?
It could and we set it not to. The minute saved by not asking a clarifying question is worth far less than one wrong order.
How do you test the matching?
Against your own recent jobs including the ones that went wrong, since straightforward vehicles prove nothing about the difficult ones.
Where does workshop margin actually leak?
Job cards. Labour done and parts fitted that never reach the card were given away without anybody deciding to.
Will technicians use a digital job card?
Only if it does not slow the bay. A half completed card is worse than the paper one, so capture speed decides this.
Does it diagnose faults?
No. Deciding what is wrong with a vehicle is a technician judgement, and a wrong diagnosis has safety consequences.
Can it keep service history?
Yes, per vehicle, which matters more here because the manufacturer record frequently does not cover this market.
Does it help with scheduling?
Yes, against real bay and technician capacity, so a collection time is a commitment rather than an aspiration.
How long before parts matching improves?
Five to nine weeks, starting with whichever costs you more, wrong parts or unbilled job card work.
Does it work for a small garage?
Yes, and small garages often feel a wrong part hardest since one bay tied up is a larger share of capacity.
Can it work with our parts supplier catalogues?
Yes, and matching across several supplier catalogues is usually necessary given how parts are actually sourced here.
What about commercial vehicles and fleets?
Yes, and fleet operators gain most from service history since the same faults recur across similar vehicles.
Can it identify a vehicle from a photograph?
Partly, and we would be honest about the limits. Chassis number and registration remain more reliable than appearance.
Will it integrate with our workshop system?
Yes, feeding job cards and history into what you already run rather than creating a second record.
Does it handle parts pricing?
It applies your own pricing. Margin and discount decisions stay with your parts manager.
What if our job card data is poor today?
Then capture comes before anything else, since history and scheduling both depend on it being reliable.
Can customers see job progress?
Yes, which reduces the calls asking whether the vehicle is ready, and those are a large share of workshop contact.
Who owns the vehicle service history?
You do, and it should transfer with the vehicle if a customer moves. That is your decision rather than a lock in we impose.
What does an automotive project cost?
Bay count and parts volume determine it, quoted after the workshop review.
Glossary of terms
Garage and parts counter terms in plain language.
- Vehicle parc
- The population of vehicles in a market, which here is mostly imported and used.
- Variant
- The specific version of a model, which determines which parts fit.
- Job card
- The record of labour and parts on a job, and where workshop margin is won or lost.
- Bay time
- How long a vehicle occupies a workshop position, which a wrong part extends by days.
- Grey import
- A vehicle imported outside official channels, frequently in a specification never sold locally.
- Confidence flag
- Marking a parts match as uncertain so a person confirms before ordering.
Related services in Industries
Talk to us about automotive
Count the wrong parts ordered last quarter and what each cost in bay time. That figure is usually larger than expected.
Start your project