Which AI models are used for agents?

Agents are typically built on leading large language models with tool-use capabilities. The right model depends on the task, cost and data requirements; the workflow design and guardrails matter more than the model choice.

How long does a first AI agent take to launch?

A single, well-scoped workflow can usually be piloted in a few weeks, including integration and a supervised trial period.

Agentic AI for marketing and e-commerce teams: where to start

Every marketing team has been told AI will change how it works. Fewer have a clear idea of what to build first. Agentic AI, meaning AI that carries out tasks using tools rather than just writing text, is most useful when it is pointed at a specific, boring job.

Chatbot, automation or agent?

A chatbot answers questions from a script or a knowledge base
An automation moves data between systems by fixed rules
An agent decides which steps to take, uses tools to take them, and reports back

Agents are the right choice when the task needs judgement that a fixed rule cannot capture, but not so much that a person must make every call.

Good first use cases

First replies to leads

An agent answers new WhatsApp or web enquiries instantly, asks qualifying questions, and hands warm leads to sales with a summary.

Product content

An agent drafts titles, descriptions and attributes for each marketplace from one source of product data, and a person approves them.

Weekly reporting

An agent pulls numbers from ad platforms, marketplaces and the webstore and writes a short summary of what changed and why.

Five rules for a safe rollout

Start with one workflow and a baseline: how long it takes and how often it goes wrong today
Give the agent only the access that workflow needs
Require human approval for anything customer-facing or irreversible at first
Log every action and review a sample every week
Expand to the next workflow only when the first one is reliably better