AI and automation

What is large language model?

Also known as: LLM

Published September 10, 2026 · Reviewed by the Fidiora team

Definition

A large language model is a machine learning system trained on very large text collections to predict likely continuations of text. That single capability supports summarising, classifying, drafting, and answering. It has no inherent knowledge of your product and no way to verify its own output.

Why it matters

Every AI support product is a language model plus a set of controls, and the controls are what differentiate them. Understanding that the model is a text engine rather than a knowledge base explains why grounding, permissions, and evaluation matter far more than which model a vendor licensed.

What to know

  • Models have a training cut-off, so anything about your product released after it is unknown to them.
  • They do not know when they are wrong, which is why external verification is a system requirement.
  • Context windows are finite, so long conversations and large documents need retrieval and summarisation.
  • The same model behaves very differently depending on instructions, retrieval quality, and guardrails.
  • Model choice is usually a smaller lever than content quality in a real support deployment.

An example

Example: two support tools use the same underlying model. One is grounded in the customer help centre with a refusal path, the other answers from general knowledge. The first is deployable and the second generates invented policies. The model was never the variable.

Common mistakes

  • Choosing a support vendor on which model they use
  • Assuming the model knows your product because it sounds like it does
  • Ignoring the training cut-off when answering questions about new releases
  • Treating model output as verified because it is well written
Where Fidiora fits

Fidiora treats the model as one component and puts the emphasis where deployments actually succeed or fail: grounding in your content, a clean escalation path, permission boundaries a CX team controls, and pricing that only charges when the whole system produced a real resolution.

See Pricing
FAQ

Questions

Does the choice of language model matter for support quality?
Less than most buyers expect. Content quality, retrieval, guardrails, and escalation design dominate the outcome. A strong system on a mid-tier model beats a weak system on a frontier model in nearly every support deployment.
Will an LLM know about my product?
Only what appeared in its training data, which for most companies is little and often outdated. Everything specific must be supplied at answer time through retrieval from your own sources.
Is my customer data used to train the model?
That depends entirely on the vendor contract, and it is a question you should get answered in writing. Fidiora does not use customer conversations or customer data to train third-party models.

Related terms

Retrieval augmented generation Retrieval augmented generation is a technique where a system searches a document collection for passages relevant to a q... Grounding Grounding is the practice of restricting an AI system to answer only from a defined set of verified sources, such as you... Hallucination A hallucination is an AI-generated statement that is fluent and confident but not supported by any real source. In custo... AI agent An AI support agent interprets a customer request in natural language, retrieves relevant information from a company kno...
Get Started

See Fidiora resolve a ticket in 60 seconds.

No credit card, no sales call required. Connect your docs and watch it work.

Start Free Book a 15-min Demo

No credit card · Live in under an hour