Baconco: Connecting product data to customer value with Agentic AI

A product catalogue can contain detailed information and still leave customers searching for answers. Product facts may sit in one system, agricultural guidance on a website and specialist knowledge in technical documents. The business has the information, but the customer has to connect it.

For Baconco, a Vietnamese producer and distributor of fertilizers, chemicals and packaging, R Digital’s work connects these layers: product information management, a WordPress and WooCommerce website, consent and analytics infrastructure, business intelligence and now a custom bilingual Farmer Advisor AI.

The latest AI delivery builds on that foundation. It gives farmers and distributors a conversational way to reach the product and knowledge sources Baconco maintains.

Connect the value chain around the customer’s question

A useful digital value chain connects the work of maintaining information with the experience of finding and understanding it. At Baconco, that means carrying product knowledge from AesirX PIM into the website and making relevant information accessible through the advisor.

Each layer has a purpose. PIM organizes product facts. WordPress publishes selected content. Approved documents hold specialist material. The AI advisor uses connected tools and retrieval to bring relevant knowledge into a conversation in Vietnamese or English.

Start with product data the business already maintains

Product information loses value when teams must copy it into multiple tools and keep every version aligned. A website update and a separate chatbot document update become two jobs, with the risk that one is missed.

AesirX PIM gives Baconco a maintained source of product information for its WordPress and WooCommerce experience. Purpose-built product lookup skills also connect the Farmer Advisor to PIM product facts and specifications. These are lookups against connected information, rather than reliance on product descriptions uploaded once and treated as static knowledge.

The business value starts upstream: maintaining product information in the system responsible for it. The same investment in data quality can then support both catalogue publishing and AI-assisted discovery.

Put Agentic AI to work through tools and retrieval

R Digital developed the Farmer Advisor as a custom WordPress plugin. Its agentic capability is concrete: it can use purpose-built integrations and knowledge retrieval to access relevant business information for a visitor’s question.

• AesirX PIM: connected product lookups for current facts and specifications.
• WordPress: selected published content retrieved in the visitor’s language, with category controls and explicit post or page exclusions.
• Approved knowledge: DOCX and PDF material converted to Markdown, plus manually created Markdown or text sources. Only approved, current, non-expired revisions are eligible for public answers.

The present implementation uses the OpenAI Responses API. Administrators can configure the AI provider, model and reasoning effort, alongside skills and guardrails. In this delivery, the advisor’s role is information retrieval and customer assistance; it is not presented as an autonomous ordering system or a replacement for agronomic judgment.

Make knowledge useful at the moment of discovery

Customers should not need to understand the company’s information architecture to use its knowledge. They need a clear way to ask a question and inspect the information behind an answer.

The Farmer Advisor opens from an AI star icon on Baconco’s website. It offers Vietnamese and English, suggested starting questions and an expanded view. Relevant answers can include excerpts from verified or approved sources, giving visitors material to inspect rather than only a generated response.

For example, the Farmer Advisor interface shows a visitor starting with “Find NPK products.” The customer-facing value is easier access to relevant product information within the website journey.

See the Farmer Advisor on Baconco’s website → Open the AI star icon in the lower-right corner.

Keep the knowledge chain governed

Connecting more information makes source control more important. Baconco’s administrators choose which article categories are in scope, can exclude individual posts or pages and manage language-specific retrieval. Drafts are excluded. Uploading a technical document does not automatically make it eligible for public answers: approval, revision status and expiry matter.

These controls let the people responsible for product and technical knowledge maintain it in their own systems. WordPress administration also includes AI configuration, PIM integration, skills and guardrails, conversations and diagnostics, providing an operational home for the advisor.

Source excerpts support review, but they do not guarantee that an AI response is correct or suitable for a particular field or crop. Agricultural guidance requires human judgment and attention to the product label.

Privacy needs the same precision. Baconco’s first-party analytics and consent infrastructure concern measurement and visitor choices. The AI chat separately discloses that messages may be processed by an AI provider. A connected platform still needs clear boundaries for each service and its data processing.

Build for value that lasts beyond the first AI conversation

The broader Baconco delivery matters because the advisor is connected to an existing operational platform. WordPress and WooCommerce provide the website experience; PIM supports consistent product publishing; first-party analytics and BI support measurement and business insight; consent management handles visitor choices.

The practical value is a more maintainable path from knowledge to assistance: fewer duplicate product documents to manage, clearer control over eligible sources and another way for farmers and distributors to discover information in their own language.

Organizations considering a similar solution should evaluate source freshness, retrieval relevance, the handling of unanswered questions, bilingual usability and the effort required to maintain knowledge. These are useful evaluation criteria, not reported Baconco performance results.

The lesson from the Baconco case is to build the connections that make AI useful. Start with the customer’s question, identify the trusted systems that can help answer it, and give the advisor well-scoped tools, approved knowledge and clear operating boundaries.