Organics R&D AI

Organics’ cosmetics R&D team had its AI help split across separate assistants, one for stock and raw materials and another for sales formulation, each with its own history, while ingredient references, current stock and past formulas lived in different places. Checking whether a formula could be made from materials in stock, and at what cost, meant moving between tools. The brief: one AI assistant that can search materials, confirm stock, draft and cost a formula and research the market in a single conversation, while making clear what is confirmed, what is an estimate and what still needs the lab.

We replaced several disconnected AI assistants with one, and wrote down how it handles each kind of work as a versioned guide that names what it may use and the rules it has to follow. Then we followed one question from the moment a formulator types it to a formula someone on the team confirms.

1. 1. Ask: every kind of work in one chat

   Every request runs plan, act, check, then summarise. The screen shows the plan and what the assistant looked up, never the AI’s private reasoning. The five guides are loaded each time it runs, with a short backup set, so an incomplete setup never answers from half the guidance.

2. 2. Find materials: real stock kept apart from references

   The assistant searches five knowledge sets from the team’s own data by meaning, and by exact match for codes and specific fields. For stock it always reads the current stock record first, and only a match there counts as in stock; catalogue matches are references.

3. 3. Draft the formula: costed, scaled and versioned

   The assistant can create, find, open with comments and revise formulas. Anything the AI drafts stays a draft until a person confirms it, and each confirmation adds a new version to the history. Costing and scaling label which figures are estimates.

4. 4. Research the market: every answer has sources

   Market questions come back with links to their web sources, and when current information can’t be verified the assistant says so instead of making it up.

5. 5. Quality and access: measured, and for signed-in users only

   A test set scores how closely answers stick to their sources, how accurate and complete the retrieved information is and how fast replies come, plus tests that try to trick the AI into breaking its rules. Only signed-in users can use it, and the system always identifies who is acting from that sign-in, never from what is sent with the request.

R&D AI runs in production as one workspace for the team. The old assistant links redirect into it, and earlier stock and sales conversations carry on in the same history.

• The whole question in one chat

  Formulators and sales staff find a material, check it is in stock and cost a draft formula without switching systems.

• Clear about what is confirmed

  Confirmed stock is kept apart from catalogue references, estimates are labelled, and lab validation is always called out.

• Formulas with a history

  Every formula keeps its versions and comments, and nothing the AI drafts becomes final until someone confirms it.

• Market answers with sources

  Research questions come back with cited links, and the assistant says so when current information could not be verified.