Andreas AhoniemiHead of Digital, Gullström & Co
SvenskaBook a call

AI in product and site

AI in the product means embedding language models into a site, a storefront or an internal system: search that understands questions rather than keywords, an assistant that answers from your own documents, or content generated from structured data. The pattern underneath is usually RAG, where the model retrieves your sources before it answers, and that retrieval step is what keeps answers tied to reality.

How the flow runs

  1. The question

    The customer writes in their words, not your keywords.

  2. The retrieval

    Vector and keyword search pull out your passages.

  3. The evidence

    Only the retrieved passages are passed to the model.

  4. The answer

    The model composes it and says where it came from.

  5. The log

    Question, evidence and answer are kept so deviations surface.

Search that understands the question

Keyword search matches words. Vector search matches meaning, so a customer typing "what do I do if my parcel never arrived" lands on the delivery terms even when those exact words are absent. For a deep catalogue it is often the single largest conversion gain available.

Assistants that stay on source

An assistant answering freely will invent things. An assistant restricted to retrieved documents, citing where each answer came from, can be trusted. The difference is in how the system is built, not in which model is used.

Cost and control

Model calls are billed per token and costs run away quickly. Flows are built with caching, smaller models for the simple steps, and per-user ceilings, so spend is predictable before traffic grows rather than after.

Start with an hour

Bring a process that grates. We walk it together and you get a straight assessment of whether automating it is worth doing.

Book a call