Artificial intelligence
AI agents and process automation
An AI agent is software that carries out multi-step tasks across the tools you already use, such as reading an e-mail, working out the request, checking the ERP and drafting the reply. It is for teams that lose hours to repetitive admin. Work keeps moving without waiting for anyone, and a person approves the steps that have consequences, such as sending, paying or deleting.
Work that repeats every day
In many companies, someone spends the morning opening e-mails, working out what each one wants, looking the customer up in the CRM, checking stock in the ERP and replying. Each step is simple, but together they eat hours, and one distraction leaves a request unanswered. An AI agent works through that sequence on its own and leaves the final call to your team.
What an agent can do
- Triage of a shared inbox, with each request routed to the right person and a draft reply ready
- Orders received by e-mail or PDF entered into the ERP and confirmed to the customer once approved
- Follow-ups on proposals sent on the right day, with the context of the last conversation
- Weekly reports pulled together from the CRM, invoicing and spreadsheets
People on the steps that matter
A useful agent is not an unsupervised one. We agree with you where it can go ahead and where it must stop and ask for approval. At first almost everything goes past a person; as results prove themselves, your team decides what can run on its own. When an agent writes to customers directly, the message says it was generated by AI.
Properly connected to your tools
An agent is only as good as its connections to the software you use. Where a connection is missing, we build it as part of an integrations and APIs project. If the work starts with invoices and forms on paper or in PDF, the agent can use document intelligence. The stages of the work are set out in how we work.
What you get
- A map of your repetitive processes and a choice of which ones to automate first
- Agents that read e-mails and documents, sort requests and prepare the next step
- Connections to e-mail, CRM, ERP, invoicing, spreadsheets and other tools
- Human approval points before anything is sent, paid, changed or deleted
- A log of every action the agent takes, with its reasoning and the data it used
- Alerts when the agent meets a case it can't handle
Best for
- Teams sorting hundreds of e-mails a week by hand
- Distributors retyping orders that arrive by e-mail or PDF into the ERP
- Sales teams that lose track of following up on proposals and leads
- Offices that build the same monthly reports from several tools
Typical features and modules
Communication & leads
- SMS & e-mail notifications
Payments & sales
- Automatic invoicing
Integrations & automation
- CRM connection
- ERP / invoicing connection
- Spreadsheet import/export
- Calendar sync
- Custom API
AI add-ons
- Document reading
- AI transparency & logging
How we deliver it
Observing the current process, who does it, how long it takes and where mistakes happen
Designing the flow, the approval points and what the agent must never do alone
Building the agent with the minimum access it needs to each tool
A shadow period, where the agent proposes and your team compares it with what they would do
A gradual move to production, with action logs and regular reviews
Frequently asked questions
What's the difference between an AI agent and ordinary automation?
Ordinary automation follows fixed rules, such as "if an e-mail arrives with this subject, file it here". An AI agent can read free text, understand what is being asked and choose the next step within limits we set. The best solution often combines the two, and we use tools like n8n or Make when they are enough.
Can the agent make decisions on its own?
Only the ones you choose to leave to it. Actions with consequences, such as e-mailing a customer, issuing an invoice or changing records, go through human approval until your team trusts the results. Some always stay with a person.
What if the agent gets it wrong?
Every action is logged with its reasoning and the data used, so you can see what happened and put it right. Before going live, the agent runs in shadow mode, and when it isn't sure, it asks for help rather than guessing.
Will customers know when an agent is writing to them?
Yes. When an agent communicates with customers directly, the message says it was generated by an AI system. We do this for transparency, and since 2 August 2026 the EU AI Act has required it whenever someone is interacting directly with an AI. Drafts that a person reviews and sends go out as messages from your team.
Is our data safe?
Each agent only gets the access it needs, and we choose AI providers that sign a data processing agreement and whose terms rule out training on your data. For more sensitive data, the agent can use a model hosted in Europe or on your own servers.
How much does it cost?
Every project is quoted by hand from your requirements, so pricing is on request. The AI model has a usage cost that depends on the volume of tasks, and we explain it in the proposal. Request a quote
Shall we talk about your project?
Tell us what you have in mind and what you want to achieve.