In Vietnam's agricultural-input distribution sector — where the customer base is hundreds of thousands of farmers asking, in local dialect, about crop diseases, growing seasons, and specialized products — Nông Nghiệp ADI took on a bet few peers would dare: putting an AI Agent at the center of its sales advisory process. And they chose Filum.ai as their partner. The first-month result: two specialized AI Agents took part in roughly 95% of conversations flowing into the fanpages, freeing the sales team from over 75% of repetitive workload.
About Nông Nghiệp ADI
Nông Nghiệp ADI operates in agricultural-input distribution and retail, with a network of dozens of branches across northern and southern Vietnam, serving a very distinct audience: Vietnamese farmers who need concrete advice on four core product groups — rice seed, corn seed, fertilizer, and crop-protection chemicals. Two main fanpages plus many branch fanpages are the first place farmers turn to with questions about cultivation techniques, diseases, or product choices for each season.
The challenge: When industry knowledge doesn't meet the right technology
Before Filum.ai, ADI already had a whole ecosystem of operational tools — a multichannel messaging platform, an internal ERP, loyalty software, a care hotline. But sales advisory on Facebook — the most direct channel between the business and farmers — was a particularly hard problem. According to Vietnam Investment Review (2024), Vietnam leads the region in business messaging, with at least a third of consumers chatting with businesses weekly — the stream of messages into ADI's fanpages each day is a pool of potential leads, but also an unsolved challenge.
Farmers don't ask in polished marketing language — they use local terms: "nematodes," "leaf-rolling worms," often with spelling errors; and messages pour in late at night, on weekends, on holidays. The previous-generation AI chatbot ADI had piloted got stuck exactly there: limited session memory, poor connection to product data, and above all — it couldn't understand "farmer-speak." ADI couldn't solve this just by hiring more fanpage staff.
The solution: A multi-agent architecture, with knowledge fed by the sales team itself
Two AI Agents for two product lines, instead of one all-purpose AI
Instead of building a single general AI Agent, Filum and ADI split the system into two specialized agents — one handling the main fanpage across the broad product range, one dedicated to the rice-seed fanpage. Each agent is tied to its own knowledge base, so a question about crop-protection chemicals is answered with chemical-specific expertise, not "diluted" by rice-seed data. All conversations and comments across multiple fanpages are also unified into a single operating interface via Filum Inbox.
The sales team teaches the AI, and the AI knows when to hand off
ADI's knowledge base isn't locked in the hands of the technical team. Sales admins in the North and in Ho Chi Minh City are empowered to update product knowledge directly — new commercial names, region-specific promotions, season-by-season disease-prevention techniques. The people closest to customers are the ones teaching the AI every day.
At the same time, the AI knows its limits: when a customer's need exceeds automated scope, it hands off to the sales team along with the full conversation context. The sales rep takes over a customer who has already been pre-advised — knowing exactly what they're interested in and what stage they're at — instead of starting from scratch.
First-month results
In the first month of operation, ADI's two AI Agents took part in roughly 95% of all conversations flowing into the fanpages. According to McKinsey (2024), GenAI has the potential to reduce human-required interactions in service industries by up to 50% — ADI didn't just reach but far exceeded that benchmark in the first month, in an industry where very few Vietnamese businesses have ever deployed specialized AI Agents.
- ~95% of conversations involved an AI Agent — right in the first month of operation (go-live 1 Oct 2025).
- Over 75% of staff workload optimized — the sales team no longer answers every repetitive question, focusing instead on situations that truly need a human: closing large orders, deep technical advice, strategic account care.
- Time-to-value measured in months, not quarters — from kickoff (25 Sep) to ~95% AI-assisted conversations within the first month.
"Filum's AI genuinely advises farmers — it understands context and picks the right product for each disease, each seed variety. Our staff now only step in for large orders and hard cases. For a niche industry like agriculture, an ordinary chatbot could never do this."
— Lê Thanh Hải, CEO, Nông Nghiệp ADI
What's next
ADI's journey with Filum doesn't stop at the first month. The next phase is being planned by both teams: expanding knowledge by season — adding cultivation know-how, new products, and disease-prevention techniques into the AI knowledge base, ready for the upcoming main crop season.
ADI's story shows that when specialized industry knowledge is continuously fed by the people closest to customers, it becomes a real competitive advantage.
Do you also run a customer base with its own language, looking for an AI architecture that truly understands your industry?
See more success stories at filum.ai/customers.
