What we do

AI strategy for SMEs: four areas, one goal.

Less effort, more confidence.

You are not buying AI. You are buying less administration, fewer errors, simpler processes, better visibility and confidence in how you use AI. That is what all four areas are about.

AI Governance & Foundations

Tooling · Security · Data · Governance · Regulation

Most companies already use ChatGPT, Copilot, Claude or other AI tools. What is often unclear is which tools are appropriate, which data may be used and which rules are needed. We help companies adopt AI in a controlled and responsible way.

Tool selection

Which AI tools fit the company and the tasks at hand?

Data protection & security

Which data may be used? Where is it processed? Which risks arise?

Governance

What may employees do? Which tools are approved? Who carries which responsibility?

Regulatory requirements

Which regulatory requirements actually apply to the company?

Velox-KAi does not provide legal advice. Legal questions are covered together with a specialised legal advisor.

The Velox-KAi framework

Four stages. What comes out is something that holds.

One page that holds beats twenty pages nobody reads. Eight questions across four stages, each with a tangible result. We work through them together, fitted to your business.

  1. Take stock

    What is already running: officially, unofficially, and what the vendor quietly switched on.

    • 01 Tools
    • 04 Security

    ResultA one-page inventory. Usually longer than expected.

  2. Set the rules

    Three data classes, one accountable person, one sentence on disclosure.

    • 02 Data
    • 05 Rules
    • 06 Responsibilities

    ResultThe AI usage policy: one page, six sections.

  3. Approve

    Every tool through the decision tree: open, with a condition, never. The condition is written down.

    • 03 Data protection
    • 07 Regulation

    ResultThe approved list with traffic lights. No “it depends”.

  4. Review

    A date, not “regularly”. What could go wrong is named, and how large the damage would be.

    • 08 Risks

    ResultAn entry in the calendar. And someone who keeps it.

Stage 3 · Approve: what it looks like

Can this go into AI? Five questions, one answer.

Take a real case from your day, such as an email, a quote or a meeting note, and click through. You end up in one of three classes.

Question 1 · Content

Does it contain personal data or business secrets?

Names, addresses, salaries, customer numbers, costings, contracts.

Free to use

General text, research, public content: any approved tool.

Question 2 · Sensitivity

Is it sensitive personal data or data under confidentiality?

Health, religion, debt enforcement, individual salaries; anything a client entrusted to you in confidence.

Never without a contract and approval

Only in an environment with a data processing agreement, and only after approval by the accountable person.

Question 3 · Account

Does the tool run on a business account, with a contract, without training on your data?

A private free account is shadow AI, even if it is the same product.

Never on a private account

Set up the business account first, then start again from the top.

Question 4 · Location

Is it documented that processing takes place in Switzerland or the EU?

Not “servers in Europe”: the processing itself, in writing from the vendor.

Only with a condition

Condition: cross-border disclosure covered in the contract and the relevant account setting checked. Otherwise: no.

Question 5 · Customer contact

Does a customer see the result, or does the AI talk to customers directly?

Website chat, automatic replies, quotes that go out as they are.

Approved, with disclosure

Customers learn that AI is involved before they reply. One sentence is enough; it has to be there.

Approved

Add the tool to the approved list, with the date of the next review. Done.

The three classes
  • Free: any approved tool.
  • Only with a condition: the condition is written down.
  • Never: not without contract, approval, business account.

No 20-person business keeps more than three classes. Three are enough, if someone has written them down.

Simplified presentation, not legal advice. We go through the full framework with you, question by question. Discuss the framework

AI & Operational Efficiency

Automation · Custom Applications · Software Replacement

Together we look at where the business loses time and money today, and how that work can be simplified.

Automation

Let recurring work run on its own.

  • Process emails
  • Read documents
  • Transfer information
  • Prepare quotes
  • Coordinate appointments
  • Produce reports
  • Detect and route tasks

Custom applications

Lightweight applications that fit the company exactly. No unnecessary features, no oversized IT projects.

Software replacement

Replace expensive, complicated or ill-fitting software with simpler, tailored solutions.

This is what sets us apart: we don't sell licences. If standard software doesn't fit, we replace it.

See examples

AI is not only about saving time. It makes new things possible.

AI enables products and services that used to be too expensive, too laborious or technically out of reach. We help companies spot these opportunities, test them quickly and turn them into working offerings.

New digital services

Offer existing clients additional value.

New products and apps

Make existing know-how digitally available and scalable.

New routes to customers

Build useful digital applications that reach and win prospects.

New business models

Use existing know-how, data or services in new ways.

Discuss an idea with us

AI Transformation

Strategy · Leadership · Training · Change · Ongoing Advisory

Employees need to understand new tools. Leaders need to take decisions. Ways of working change. And someone has to make sure the first experiments turn into something lasting.

What we offer
  • Training
  • Workshops
  • Strategy
  • Leadership advisory
  • Rollout
  • Change support
  • Ongoing advisory

On request, we act as your external AI partner.

Not for a single workshop or project. We support companies over the longer term, pragmatically, and scaled to what is needed at the time.

Discuss working together
How we work

Start simple. Learn. Build on it.

  1. Understand

    Where does the business stand? Where are the problems and the opportunities?

  2. Prioritise

    What actually creates value? What is realistic? What is better left alone?

  3. Deliver

    Build a first working solution and test it with real users.

  4. Improve

    Measure, learn, and keep investing where real value emerges.

Start small.Prove the value.Scale what works.

AI Quick Check

In 60 seconds: where AI delivers most in your business.

5 short questions, no sign-up, no sales call. Just clarity on which of the four areas comes first for you.

Ready
01 AI Governance02 Operational Efficiency03 AI-enabled Growth04 AI Transformation
◷ ~60 seconds ⛨ Anonymous & free

Your result

  • Your biggest leverWhich of the four areas delivers first for you, and why.
  • First stepsThree concrete steps, no IT jargon.
  • The next areaWhat delivers most after that.

A first assessment, not an audit.

ERP check

Can your ERP be connected to AI?

Two answers, one honest verdict from our study on the Swiss SME ERP landscape 2026. The check runs entirely in your browser: nothing is sent, nothing is stored.

Where does it run?

Choose your ERP and the answer appears here.

30 minutes with Matthias Read the study (PDF, German) Study status: 22 September 2026.

Frequent questions

Do I have to replace my ERP?

No. An ERP stays in use for 17 years on average. The question is not which ERP, but how to automate the work around the one you have. The ERP remains the system where the numbers are right.

What does an AI connection actually do?

Capture documents from the inbox, turn enquiries into quotes, prepare dunning runs, answer questions about key figures and keep data in sync between ERP, CRM and portals.

What is an MCP server?

An open standard through which a language model such as Claude accesses a system securely. The MCP server defines what it may do: find customers, create a quote, read a balance. It is the socket between ERP and AI.

Does my data leave the building?

Anthropic processes requests to the language model outside Switzerland. Which data goes there can be controlled through the MCP server. That has to fit data protection law and your client contracts, and your clients should know where processing happens. Which data you may give to an AI at all is settled by our framework in five questions →

What may an AI assistant do in the ERP?

Only what it should. It should work with its own ERP user with limited rights and carry out payments, payroll runs or deletions only after confirmation. KLARA and Vertec build their connections that way, and that is how we set it up. Who approves what in your business is set by our framework →

Our system has no interface. Is anything possible?

Yes, through the document flow: email, PDF, export and import. It works with any system, just not in real time and without queries back to the ERP.

Questions

What executives ask us first.

01 Do we have to introduce new software?

No. We work with what you have: Outlook or Google Workspace, your accounting, your industry package. We build something new only where it is simpler than what exists.

02 Where is our data processed?

That depends on the tool, and it is exactly what we settle first: which data may go where, and which may not. Our website is hosted by Infomaniak in Switzerland; email and calendar run in Microsoft 365 with the EU data region. Which AI services your business uses is decided together, by purpose and data protection.

03 How quickly will we see something?

We start small: a first working solution in weeks, not months. Then we measure whether it delivers, and scale what works.

04 What does it cost?

A one-off setup fee plus an optional monthly advisory retainer. Both are spelled out in the proposal before you commit. No hidden costs.

05 Are we locked in?

No. Ongoing advisory is cancellable monthly. We earn the engagement every month.

06 Is this legal advice?

No. We tell you which data-protection and AI-regulation questions matter for your business. The legal assessment is done together with a specialised legal advisor.

Where could AI make the biggest difference in your business?

  • Perhaps your team spends too much time on administration.
  • Perhaps you pay for software that doesn't really fit.
  • Perhaps you have an idea for a new service.
  • Or you first want to know how to adopt AI safely.

Let's find out together.

30 minutes · no obligation · directly with Matthias

Pick a time (opens in a new window)

Opens Microsoft Bookings in a new window. Choose a free slot and book as a guest, no account needed.

Or email us directly at info@velox-kai.ch.

Or call directly: +41 79 470 55 84

Prefer to write?

Tell us briefly what it is about. Matthias replies within one working day with proposed times.

We use your details only to arrange the conversation. Privacy

Call Book a conversation
ERP check

Can your ERP be connected to AI?

Two answers, one honest verdict from our study on the Swiss SME ERP landscape 2026. The check runs entirely in your browser: nothing is sent, nothing is stored.

Where does it run?

Choose your ERP and the answer appears here.

30 minutes with Matthias Read the study (PDF, German) Study status: 22 September 2026.

Frequent questions about ERP connections →