AI for SMEs · Weesp, Amsterdam & Het Gooi

Work smarter with AI, without client data leaving the building.

For accountants, notaries, law firms, healthcare, HR and manufacturers. We put an AI assistant in your office that knows your own documents, train your team and build AI agents that take work off your hands. Fixed price per phase, and you own the custom code.

30 minutes · no obligation · reply within one business day

Runs offline in your own buildingAI training for teams of 5–15Fixed price per phaseYou own the code and dataOur own workshop in Weesp
Architecture

What In-House AI looks like

In-House AI architectureStaff use a local AI server (Mac mini, Mac Studio or GPU server) running a language model and RAG over the company’s own documents, inside the office network. The connection to cloud AI is closed. YOUR OFFICE · LOCAL NETWORK Legallaptop · browserFinancelaptop · browserCustomer servicelaptop · browser Local AI serverMac · GPU server LLM RAG Documentscase filesmail archiveNAS · SharePoint connection closed Cloud AI
Staff use a local AI server (Mac mini, Mac Studio or GPU server) running a language model and RAG over the company’s own documents, inside the office network. The connection to cloud AI is closed.
Use cases

How AI fits your business today

A few scenarios we see often. Each one starts small, with a fixed price and a measurable goal.

01 Customer service

Customer service agent for email and WhatsApp

Problem
Your team answers the same questions about delivery times, order status and returns every day.
Solution
An agent reads incoming messages, looks up the order status in your system and drafts a reply. Hard cases go to a colleague.
LLMHelpdesk integrationWhatsApp API
Measurable goalFaster response times and less manual work per ticket
02 Back office

Automatic invoice and document processing

Problem
Purchase invoices, packing slips and forms are retyped by hand.
Solution
AI reads PDFs and photos, extracts the fields, checks them against purchase orders and stages them in your accounting system.
Document AIOCRAccounting integration
Measurable goalLess retyping and fewer errors
03 Sales

Quotes in minutes instead of hours

Problem
Writing quotes takes time: finding specs, calculating prices, writing the text.
Solution
An assistant turns a request and your price lists into a draft quote in your house style. The salesperson reviews and sends it.
LLMCRMTemplating
Measurable goalLead time from request to quote
04 Operations

Predictive maintenance from sensor data

Problem
Machines, vehicles or installations fail without warning.
Solution
We read sensor data (CAN, Modbus, IoT), spot abnormal patterns and warn you before something breaks. We already do this for EV batteries.
TelemetryAnomaly detectionEdge
Measurable goalLess unplanned downtime
Why Voltsoft

What most AI agencies don’t do

Commitments you get in writing on every project.

Fixed price, measurable goal

Every phase has a fixed price and an agreed KPI. No open-ended projects, no surprises.

You own it

All custom code, models and data are yours, running on your accounts or servers. No lock-in.

Private & EU-first

Sensitive data stays in the EU or on your own hardware, with a GDPR and AI Act check on every project.

From chip to cloud

We also build firmware, sensors and IoT, so AI works in machines, vehicles and on site.

We stay after launch

Monitoring, evaluation and improvements. Many agencies stop at handover. We don’t.

R&D that fits WBSO

We work as an R&D studio and document the technical novelty, so your projects can fit the Dutch WBSO scheme.

How we work

Start small, prove it fast, then scale

Every phase has a fixed price and a clear outcome. After each phase, you decide whether we continue.

± 2 weeks

AI scan

Map processes, estimate opportunities, pick one pilot with a clear goal.

4–6 weeks

Pilot

A working prototype on real data, tested by your own team.

6–12 weeks

Build

Make it production-ready: integrations, security, permissions, tests and docs.

Ongoing

Run

Monitor, measure and improve. Cancellable monthly.

Portfolio

Software already running in the real world

From a kitchen configurator with 29,259 variants to source-checked RAG and medical sensors. Built and maintained by our own team.

Illustration: ReportAnalyzer · Paper Analyzer
AI · RAG

ReportAnalyzer · Paper Analyzer

A local-first RAG platform for analysing annual reports, papers and long documents. Answers quote verbatim, with page numbers. An AnswerGuard checks every quote against the source and refuses to answer rather than make something up. Facts such as headcount and sustainability goals are extracted automatically.

  • Page-cited quotes
  • Quotes checked against source
  • 18 ADRs
RAGDoclingPostgreSQLEvals
Illustration: Voltron AI agents
AI agents in production

Voltron AI agents

Agents that help run Voltron’s business: a chat agent, a WhatsApp agent that reads messages and drafts replies (sent only after approval), and an MCP gateway that safely connects Claude to business systems.

ClaudeMCPWhatsAppHuman-in-the-loop
Eliton Compositions
Client project · Configurator

Eliton Compositions

A cloud configurator for Eliton S.A.’s kitchens and cabinets. Dealers and sales staff compose complete kitchens with live B2B and retail pricing across 29,259 individually priced variants. Includes a 2D editor with real textures, specialist quote approval and a dealer portal.

  • 29,259 variants
  • 332 finishes
  • 14 material groups
FlexFlowB2B portalData migrationReact
One group, two specialisms

Voltsoft and Voltron: under the same umbrella

The same team and the same workshop in Weesp, each with its own focus.

VoltsoftSoftware & Applied AI

AI agents, automation, private AI, custom software, apps, firmware and R&D for businesses in any sector.

All services →
VoltronEVs · ECUs · Workshop

Repair, diagnostics and conversion of electric vehicles and boats, ECUs, battery tests and parts.

Visit e-voltron.nl ↗
FAQ

Questions about AI in your business

How much does it cost to implement AI in my business?

It depends on the use case, but we always work in phases with a fixed price. You start with a short AI scan or go straight to a 4–6 week pilot. You only invest in the production version once the pilot hits its goal.

Is AI worth it for small and mid-sized companies?

Yes. In SMEs, where a few people cover many different tasks, automating repetitive work pays off quickly. We start from a concrete problem, not from the technology.

What’s the difference between a chatbot and an AI agent?

A chatbot answers questions. An AI agent also does the work: looking things up in your systems, drafting documents, updating records. Where needed, a human approves the result first.

Will my data go to OpenAI or other US companies?

Only if you want it to. We can run open-weight models on your own server or in a European data centre, so your data never leaves the EU.

Who owns the software when it’s done?

You do. Code, configuration, prompts and trained models become your property and run on your accounts or servers. You’re never tied to us.

How fast can a first version be running?

A working prototype on real data is usually ready within 4–6 weeks. Simple automations can be faster.

Do you work in English?

Yes. We work in English and Dutch, and we build multilingual AI solutions too.

See it for yourself: AI with the cable unplugged

In a free demo, a local AI server answers questions about real documents while the network cable is unplugged. At our workshop in Weesp or at your office. You get honest advice, even if AI doesn’t pay off for you (yet).