We start small and easy: with the email and the simple, recurring questions your team answers every day — order status, warranty, returns, appointments. Those get handled right away, so your people have time for the conversations that really matter. A full phone agent can come later, when you are ready — we have already proven that at an e-bike manufacturer. Smart assistants that know when a human needs to take over.
Recognise the pattern — day in, day out the same ten questions about status, warranty, returns or appointments? Then this works for you. It doesn't have to be complicated. We feel most at home in Dutch manufacturing, where we have proven the most, but the principle applies just as well beyond it:
Warranty and returns flow, dealer locator, battery and service appointments, order or repair status. Including integration with your dealer and workshop system.
Capacity and charging complaints, BMS readouts, warranty and return procedures, safety and certification questions, installation help. Verified answers from your own documentation and test data.
Webshop, wholesaler, installer or service provider — wherever the same questions about status, invoices, appointments or returns keep coming back, the same approach fits. We start with whatever costs the most time.
Under the hood it is always the same building block: a layer that wraps around your systems and automates whatever can be automated — whether that is invoices or answering the phone automatically.
The real product sits under the hood: a deterministic layer that wraps around your systems and reliably executes whatever you want to automate. The channels sit on top — we start easy with email and the simple, repetitive work, and expand towards phone as soon as you are ready. Anything that deserves attention is passed neatly to your team.
Our core: a dedicated server that wraps around your critical systems — ERP, webshop, customer service system, email. Strict, predictable execution on verified input only. With it we automate whatever can be automated, from an invoice to a warranty check by serial number. The channels above do the talking; this layer does the work.
Where we usually start: an agent that categorises incoming email, replies where possible, or escalates with full context — every email gets an owner or an answer.
The AI knows when a human is needed — and warm-transfers to the right colleague, with a short summary. The customer tells their story once; your colleague has the full context at hand.
The agent learns from your own manuals, FAQs and historical tickets. Answers are always in line with your policies, product information and tone of voice.
Everything is logged neatly in your existing CRM or ticketing system. Gorgias, Zendesk, HubSpot, Odoo — we connect it.
For those who are ready: a phone agent that sounds natural, asks follow-up questions when needed, and passes complex questions neatly to the right colleague. Not a required first step — but where it can grow. It is already live at an e-bike manufacturer.
The biggest pitfall of AI in customer service is the agent making things up. An invented tracking link, a fictional returns policy, a price that differs from what is in your systems — useless.
That is why we build two layers. A conversation layer where the agent talks and listens naturally. And beneath it a deterministic action layer that only executes real, verified actions through API integrations with your business tools.
The tracking code the agent shares comes straight from PostNL. The stock level it mentions comes from your inventory system. The ticket it promises is actually created in your CRM — otherwise the conversation only wraps up once that succeeds.
The paradox of modern customer service: the more you automate, the more room there is for real attention. Provided you do it in the right place.
The AI handles the simple, recurring work — whether it is an email or a phone call — and hands over as soon as it gets complex. The customer reaches a human the moment it matters, instead of after seven minutes on hold or three days waiting for a reply.
The simple, repetitive questions are handled instantly. That leaves more time for the customer with a complex issue, the question that needs real expertise, and the kind of conversation your people were hired for in the first place.
When a conversation is handed over, your colleague gets a summary, customer details, previous orders — everything. The customer tells their story once.
Ten questions a day or a thousand — your service stays personal where it should be, and efficient where it can be. Your team stays calm, your inbox stays under control.
It starts with advice, not code. First we dig into your customer contact and your processes: where the friction is, what you can automate and what is better kept human. Sometimes that advice alone is the biggest win; often it is followed by a solution we build together with your team — their knowledge as the starting point. Short iterations instead of a six-month project, step-by-step rollout instead of all-or-nothing.
We talk to your team, look at your processes and systems, and see where time disappears. You get honest advice: what AI delivers here, what it should not do, and what we keep human. Even if the conclusion is: don't automate just yet.
The agent learns from your knowledge base, your tone of voice and your past conversations. Short iterations with test calls, your team gives feedback, we refine until it's right.
First on one line or segment, with your team as a safety net. What works well we scale up; what falters we adjust. Monthly reviews together.
Fully transparent, fully transferable. You get access to the code, the configuration and the servers it runs on. We remain your technical partner — for as long as you need us.
Your own environment — owned by you. Your own server, your own code repository, your own integrations, tailored to your tools and processes.
We handle the complex maintenance — model updates, monitoring, security patches. Your team gets access to an internal tool to adjust conversation flows, answers and escalation rules themselves.
At any time you can move management to your own team, or to another party. Documentation, source code and knowledge transfer are ready. We cooperate fully with the handover.
For PHATFOUR (e-bike manufacturer) we built a multi-stage voice agent with five specialised flows: support — automatically split between warranty and workshop service — shipping, sales, dealer and a catch-all for everything else. The router listens, classifies within seconds, and handles simple matters (order status, opening hours, warranty terms) directly.
Beneath the conversation layer sits a deterministic action layer that makes real API calls to Odoo (frame number and warranty checks), Gorgias (tickets, tags and email macros) and WhatsApp for customer communication. Only verified data from live systems — what the agent says comes straight from your systems.
After every call a Gorgias ticket is created automatically, with B2C or B2B routing already filled in based on whether a consumer or a dealer is calling. At the same time the customer receives a WhatsApp message; as soon as they reply with their email address, the system picks the right one of seven email templates and sends the confirmation straight away — service appointment, warranty claim, sales follow-up or dealer report.
For anything that needs a human — a hesitant buyer, a complex warranty claim, a technical problem that requires product knowledge — there is a warm transfer to the right colleague, with full context. Or a WhatsApp callback when the customer would rather be called back later.
Fill in the form and we will schedule a no-obligation introductory call at a time that suits you. We listen to your situation and give you an honest picture of what AI can realistically deliver for your business.
