Solution 02

AI agents and assistants for landscapes

Reliable, timely information for extension staff and farmers, with experts validating every answer.

The problem

Extension staff and farmers don't get reliable information in time to decide, and field teams spend too many hours writing reports instead of working with producers.

What we deliver

Conversational assistants

On WhatsApp or web, answering from expert-validated knowledge (RAG) and citing their sources.

Data agents

Query climate and field data to give recommendations grounded in the landscape.

Report automation

Draft reports from traceable data for the team to review and approve.

Human oversight

Review dashboards, conversation logs and regular quality evaluation.

Typical measurable outcomes

We agree indicators with you before we start. These are the most common:

↗Queries resolved
↗Response time
↗Expert-validated accuracy
↗Active users

Technology and methods

  • RAG
  • WhatsApp Business API
  • Evaluated LLMs
  • Climate data
  • Oversight dashboards

Related work

Challenge
Producers and extension staff in the watershed needed timely agronomic guidance, but in-person technical assistance can't cover every question.
Outcome
An always-available assistance channel, with answers traceable to their sources and an improvement cycle driven by expert review.
Live
Coffee Cloud dashboard with a map of Guatemala and indicators
Challenge
Coffee-farming families were making management decisions without access to climate information or timely technical assistance.
Outcome
An internationally recognized farmer-information model (MIT Technology Review Innovators Under 35, Cracking the Nut 2017).
AI agents and assistants for landscapes
~6,000coffee farmers reached
Challenge
Livestock producers kept few records, making it hard to improve productivity or demonstrate good practices.
Outcome
Farm records that support management decisions and technical assistance.
AI agents and assistants for landscapes

FAQ

What people usually ask

How is this different from a regular chatbot?

A generic chatbot answers with whatever it finds or generates. The assistants DAP designs answer with expert-validated knowledge, and every conversation is logged so it can be reviewed and improved.

Does it work on WhatsApp?

Yes. Assistants can run on WhatsApp or the web, depending on how farmers and field staff communicate. Somos Río Lempa, in El Salvador, is an example on WhatsApp.

How is answer quality controlled?

Answers can be traced back to their sources and experts review the conversations. That review feeds a continuous improvement cycle.

Next step

What outcome do you need to prove in the next 12 months?

Tell us about your challenge. In a first conversation we'll see whether a diagnostic makes sense for your organization.

Book a diagnostic →