How the work looks
One mobile app with three AI agents 
A proposal generated and saved to Salesforce 
Photo diagnosis of a crop disease with matching products 
The AI agronomist answering a field question
The Challenge
Field consultants needed roughly 40 minutes to diagnose crop issues, search a catalogue of more than 20,000 products, choose a suitable treatment, and prepare a quotation. The work happened across separate tools, making every recommendation slow to produce and difficult to analyse at network level.
What We Built
DeView built a three-agent system on Claude via Vertex AI, Salesforce, and Flutter. A consultant photographs the affected crop; the system identifies the likely issue, ranks suitable catalogue products, and writes a draft quotation directly into Salesforce for human approval. The same event trail powers anonymised vendor analytics.
AI Capabilities
- 01
Multimodal Field Diagnosis
Crop photographs and field context are analysed together to produce a diagnosis and ranked treatment recommendations.
- 02
Catalogue-Grounded Recommendations
Every recommendation is matched against the live 20,000-product catalogue, so consultants receive products they can actually quote.
- 03
Salesforce-Native Quote Drafting
The diagnosis becomes a structured quotation in Salesforce, including alternatives and substitution rules, ready for consultant sign-off.