AI agent development
Custom AI agents
that know where to stop.
AI agents for one job at a time: following up leads, answering routine email, keeping the CRM current. Unusual cases go to a person.
How it runs.
Choose a step to see what it reads, what it does and where a person decides.
Starts whenDone
An enquiry arrives
- What it does
- A new request comes in through the website form or email. The agent picks it up within a minute.
- Works in
Gmail
- Example
- “We need a quote for 40 pallets a month to Tallinn.”
ReadsDone
Looks up the company in the CRM
- What it does
- It checks whether the company is already a customer, who owns the account and what was said before.
- Works in
HubSpot
- Example
- New company · owner: sales team · no earlier deals
AI stepDone
Creates the record, drafts the reply
- What it does
- It creates the CRM record with the details filled in and drafts a follow-up that asks only what is missing.
- Works in
Gmail
HubSpot
- Example
- Record created · draft asks for pallet size and start date
CheckDone
Within its rules?
- What it does
- The agent only acts inside the rules you agreed: which tools it may use, what it may send alone and what it must hand over.
- Example
- Tools: CRM, email ✓ · price quoted: no · sending alone: not allowed
If unusualSent to a person
A price question goes to a person
- What it does
- Anything outside the rules, such as a custom price or a complaint, goes to a named person with the context attached.
- Works in
Slack
- Example
- “Can you beat €2,100?” → sales owner on Slack
Your decisionWaiting for you
The sales owner approves
- What it does
- The draft waits in the owner's inbox. One click sends it; any edit is kept so the agent improves.
- Works in
Gmail
- Example
- Draft ready 09:12 · waiting for the sales owner
ResultNext
Sent, CRM up to date
- What it does
- The reply goes out from the owner's own address, and the CRM shows what was sent and when.
- Works in
Gmail
HubSpot
- Example
- Sent 09:20 · deal stage: qualified
Illustrative run with example data. Your tools and rules decide the detail.
More AI automation services
What you have after the first phase.
Scope and price are fixed before we build anything. A first phase usually takes 3–6 weeks.
- One agent live on one recurring task, working in your own email and CRM.
- Agreed rules for what it does alone and what waits for your approval.
- A record of every run, and unusual cases sent to a named person.
What an AI agent can take on.
- Sales
- A new enquiry gets a researched, drafted follow-up within minutes, waiting for the owner to send it.
- Lead generation
- Inbound leads are scored against your criteria and routed to the right person with context.
- Routine questions get a prepared reply; anything unusual goes to a person.
- CRM
- Contacts, deals and notes stay current without anyone retyping them.
What our agents do not do.
Autonomy is a setting, not a selling point. You decide how much of it each agent has.
- No unchecked sending
- Anything that leaves the business can wait for a person to approve it.
- No open-ended access
- An agent only reaches the tools and data its task needs.
- No silent failures
- Every run is recorded, and exceptions go to a named owner.
Common questions
What is the difference between an AI agent and a chatbot?
A chatbot answers questions in a conversation. An AI agent takes steps: it reads an email, looks up the customer, updates the CRM and prepares a reply. Many of the systems we build combine both.
Can an AI agent work with our CRM and email?
Yes, if they offer an API, and most do: HubSpot and other CRMs, Gmail, Microsoft 365, Slack, Google Sheets and many industry tools. We confirm what can connect during discovery.
How much does it cost to build a custom AI agent?
It depends on the task, the tools involved and how much needs review. After a short discovery we propose a fixed scope, timeline and price for a first agent, before any building starts.
Do you build agents in n8n or in code?
Both. Simple, well-structured workflows often fit n8n; agents that need more control, testing or custom logic are built in code. We recommend one during discovery and explain why.
Bring the task you would hand to an agent.
One recent example is enough. We will tell you honestly whether an agent is the right tool.
A 30-minute conversation. No technical brief needed.