Product answers in plain languageDirect Sales Nutrition Company

An AI Chatbot Trained on Company Data

A nationwide direct sales nutrition company keeps its product knowledge in written documentation: product details, usage guidance, policies, and procedures that its product team relies on every day. The documentation was thorough, but it was spread across many files, and getting to a specific answer meant knowing where to look. New team members had the hardest time.

The company asked Insite to build an internal tool that lets employees ask questions in ordinary language and get answers drawn only from the company's own documentation, with each answer pointing back to where it came from.

Goals

  • Give the product team one place to ask product, policy, and procedure questions and get an answer in conversation rather than through a search.
  • Show the source behind every answer so employees can verify it before acting on it.
  • Keep the knowledge base in the company's hands, with updates published on the company's own schedule.
  • Track how the tool is used and how its answers are rated, so the documentation can improve over time.

Pain Points

  • Product knowledge lived across a large set of documents, so answering a question meant finding the right file and then the right section within it.
  • New hires had a full reference library to learn and no quick way to check a fact while working.
  • The company had no view into which questions came up most often or where its documentation fell short.

Solution

Insite built a web-based chat assistant that answers from the company's documentation and nothing else. Rather than pulling isolated fragments, the assistant has the full knowledge base in view for every question, which lets it connect information across documents and handle follow-up questions in context. Every response lists the documents and sections it drew from, and a thumbs up or down on each answer feeds a review queue for administrators.

Conversations are kept short by design. Each session runs a fixed number of exchanges before it resets, which keeps answers grounded in the documentation and stops a long conversation from drifting.

The knowledge base itself stays in the company's version-controlled documentation. An administrator publishes updates with a single sync, the previous version keeps serving until the new one is ready, and every answer records which version of the documentation produced it. Access is by request: employees register, an administrator approves them, and an admin panel covers user management, usage and cost reporting, and a searchable audit log of every question asked.

Results

The company's product team has documentation it can question directly. Answers arrive in conversation, cite their sources, and stay within the company's own material, so an employee can check a policy or a product detail without opening a file. Administrators can see what the team is asking, which answers were rated poorly, and which parts of the documentation need attention, and they can publish a correction to every user in one step. Every question and answer is logged against the version of the documentation that produced it, so the company has a complete record of what its team was told and when.