Focus area

Agentic AI: from chatbot to a colleague that actually gets the job done

An ordinary chatbot answers. An AI agent acts: it plans a goal in several steps, uses your systems — email, CRM, calendar, files — and works its way through the task until it is done. I build agents that take the repetitive work off your team’s hands, with a human at the wheel where it matters.

Ordinary chatbot
  • Only answers what you type — and forgets everything when the window closes.
  • Cannot look things up in your data or take actions in your systems.
  • You still have to do all the practical work afterwards: copy, send, create, update.
  • One question, one answer — no understanding of a task with several steps.
AI agent
  • Is given a goal and plans the steps needed to reach it on its own.
  • Uses tools and integrations: reads data, sends emails, creates cases, updates the CRM.
  • Keeps working on its own — and only asks for help when there is genuine doubt or a decision to make.
  • Remembers context throughout the task and can run around the clock.

How an AI agent works

1

Goal

The agent is given a concrete task — “answer this customer enquiry”, “qualify this lead”, “reconcile this invoice” — in clear, well-defined language.

2

Plan

It breaks the goal down into steps and decides which tools and data to use along the way. The plan can be displayed so you can follow along.

3

Action

The agent carries out the steps: looks things up in your systems, calls APIs, writes drafts, creates and updates records — one step at a time.

4

Control

For sensitive or costly actions the agent stops and asks for approval. You decide what it may do on its own and what requires a human.

5

Learning

Results are logged and the agent is adjusted over time based on feedback — so it gets sharper at your specific tasks and your way of doing things.

What the agents I build typically do

Agents aren’t a single product but a way of building. Some of the most valuable places to put them to work:

Customer service agent

Reads the enquiry, finds the answer in your knowledge base and systems, writes a draft — or answers the simple cases itself and escalates the rest.

Lead and sales agent

Qualifies incoming leads, enriches them with data, creates them in the CRM and prepares the next step for the salesperson.

Document and case agent

Pulls information from documents, fills in templates, reconciles supporting documents and puts the result in the right place.

Operations and monitoring agent

Keeps an eye on systems, data or inboxes, reacts to events and notifies you — or resolves things itself according to fixed rules.

Building blocks and tools

Tool use / function callingMCP integrationsRAG on your own dataHuman-in-the-loopGuardrails & policiesMulti-agent orchestrationEvent triggersFull logging

The agents can be built on leading models (Anthropic, OpenAI, Google) or on local models where data has to stay in-house — see AI & GDPR.

What it typically means

Round-the-clock operation

Agents work through the night and at weekends too — tasks are done by the time you get in

Fewer clicks

The repetitive copy-paste work disappears, so the team spends its time on what actually requires people

Control retained

You set the boundaries: what the agent may do on its own, and what always needs human approval

Agentic AI is powerful, but it needs sensible boundaries — I start small with one well-defined area, prove the value and expand from there.

Shall I build your first AI agent?

Together we find the best place to start — one concrete, well-defined task — and demonstrate the value before I scale up.

Book a free AI discovery call →