AI glossary · Agents
Agent memory
In one sentence
Agent memory is the information an AI agent keeps so it can use it later: what has been said in the current conversation (short-term) and stored details from earlier conversations, such as a customer's preferences or history (long-term).
Also known as: AI agent memory · long-term memory · conversational memory · LLM memory
What is agent memory?
A language model has no memories. Each time it’s called, it only sees what you send it at that moment, inside its context window. If a customer writes today and again next week, the model has no idea it’s the same person. Memory is the layer an AI agent adds to fix that.
Put it to work
AI agents that look after your customers
From €49/mo + VAT
Voice, WhatsApp and email with the phone line included: agent, numbers and CRM in one fee.
See plans and pricingIt’s usually split into two types:
| Type | What it keeps | Where it lives | Example |
|---|---|---|---|
| Short-term | The current conversation | In the context sent to the model | ”You said earlier it’s for Saturday” |
| Long-term | Useful details across conversations | Database, CRM or vector store | ”This customer always books with Laura” |
Related to this term
UTIO agents
- Available 24/7 in 26 languages
- Set it up yourself for €0: built from your website
- 14-day free trial, no card needed
From €49/mo + VAT
See plans and pricingHow it works
- During the conversation, the agent sends the earlier messages back to the model so it keeps the thread. Because every message costs tokens, long conversations get summarised.
- At the end, it decides what’s worth keeping: a contact detail, a preference, an open commitment.
- Next time, it identifies the person (by phone number or email) and retrieves what was stored. If there’s a lot, it fetches only what’s relevant, using techniques similar to RAG.
- It injects that into the context before replying. Done well, the customer simply notices that they’re recognised; done badly, the agent brings up details that are out of date or irrelevant.
Memory is what lets agentic AI chain steps together without losing sight of the goal, and lets several agents coordinated through orchestration work from the same information.
Example
A bike shop in Edinburgh handles customers on WhatsApp. In March, a customer asks about a service and mentions that it’s a road bike and he can only drop it in on Mondays. In June he writes again. The agent recognises his number, remembers the bike and goes straight to offering Monday slots, without asking everything again.
Agent memory at UTIO
At UTIO, what the agent remembers is built on the CRM: every contact has a timeline with their messages, calls, emails and appointments on a single record. That way the WhatsApp agent and the voice agent work from the same history.
Recordings, transcripts, agent memory and the CRM are stored in Spain. We sign a data processing agreement, and you can ask for an export or deletion whenever you need.
Frequently asked questions
Do language models remember conversations?
Not on their own. Every call to the model starts from scratch; it only 'knows' what's in its context window. Memory is added by the system around it, which stores information and feeds it back when it's needed.
What's the difference between memory and RAG?
RAG searches your general business documentation, such as price lists or manuals. Memory keeps what was learnt from each interaction, such as a customer preferring afternoons. Technically both can use the same database, but they do different jobs.
Is it legal for an agent to remember customer data?
Yes, under the same rules as any personal data: a clear purpose, telling the customer, retention limits and the right to erasure. GDPR applies just as it does to your CRM.
Related terms
- Context window
A context window is the maximum amount of text, measured in tokens, that a language model can take into account at once: instructions, conversation, documents and its own reply. Anything outside it, the model simply doesn't see.
- AI agent
An AI agent is a system that uses a language model to understand a request, decide which steps to take and carry them out with real tools, such as your calendar or CRM, until the task is done rather than just replying.
- RAG
RAG (retrieval-augmented generation) is a technique in which a language model first retrieves the relevant passages from your own documents and then writes its answer from them, instead of relying on what it memorised in training.
- Vector database
A vector database stores embeddings, lists of numbers that represent the meaning of content, and finds the ones closest to a query very quickly. It's the search component in most RAG systems.
- Embeddings
An embedding is a list of numbers that represents the meaning of a piece of text, an image or audio. Items with similar meaning have embeddings that sit close together, which lets you search by meaning rather than by exact words.
Updated on