Future capability
Image recognition
Use a photo to help identify visible pump components or fault indicators. This would require model selection, testing and clear safety thresholds.
Concept demonstration · Funded pilot required
AI-assisted support for frontline water operations & maintenance
Could the same human-in-the-loop pattern behind AgriConnect help frontline water staff find trusted procedures, learn faster, and know when to escalate?
Problem
When a handpump fails, Joseph may not have immediate access to diagnostic expertise, technical documentation, or the district office.
Partner type
- Rural WASH implementing organisations
- District Water Offices / O&M teams
- Handpump maintenance programmes
Data available
- National handpump O&M manual
- District asset register
- Spare-parts catalogue and local supplier network
- Approved safety and escalation procedures
Important: this is a design concept. The water O&M assistant has not yet been built or validated.
Illustrative interaction
Show the idea, without pretending the product already exists
The example below demonstrates how a future assistant could retrieve the right source, ask for missing context, and stop when human judgement is required.
WaterConnect · concept interaction
Illustrative only
Joseph · Pump H-023The pump is not delivering water. What should I check?
Looking for an approved procedure…
Concept system response
I can help you find the relevant procedure, but I need to confirm a few details first.
- Confirm the pump model or asset ID.
- Describe the observed fault or symptom.
- Confirm the site is safe to inspect.
Source basis: approved O&M manual + asset register + escalation rules
Second use case
The same knowledge base could support rapid upskilling
For new or less experienced staff, the assistant could make existing institutional knowledge easier to navigate at the moment of need.
New technician
“I’m new to this pump. What should I learn before working on it?”
The system would not replace formal technical training. It could provide a structured refresher based on approved materials.
Potential learning path
1Understand the asset and its main components
2Open the relevant approved O&M section
3Review required tools, safety and competency prerequisites
4Know which tasks are authorised and which require escalation
5Complete a short knowledge check or supervisor review
Future development
What would require additional product work
These capabilities are plausible extensions, but they are not available in a WaterConnect O&M proof of concept today.
Future capability
Asset context
Combine the question with asset type, location, maintenance history and previous failures from the utility register.
Future capability
Voice & local language
Allow staff to ask questions by voice where typing is difficult, with language support tested for the operating context.
Future capability
Maintenance feedback loop
Record what was done, whether the fault was resolved, and what knowledge gaps recur across technicians and locations.
The proposition
Make trusted institutional knowledge easier for frontline staff to use.
AgriConnect provides a working pattern for AI-supported advisory with people in the loop. Water operations and maintenance is a direction to explore with WASH partners, using their own approved content, workflows and safeguards.
Trusted sourcesUse approved organisational knowledge
Human validationKnow where AI stops
Context firstDesign with the utility and frontline user
Learn from useUse recurring questions to improve training
Prepared by Akvo for discussion. Concept demonstration only. WaterConnect O&M is not yet developed and requires a funded pilot for design, development, testing and validation. This is not an official UNICEF-branded product.