schnell.digital AI Kit

schnell.digital AI Kit · The AI platform for mid-sized businesses

You describe the task. The AI Kit builds the workflow that anyone can read.

You describe a process in your own words. The AI Kit connects it with your company knowledge, your systems and the AI model of your choice.

  • Hosting in Germany or on-premise
  • Free choice of model
  • Pilot in 6 to 8 weeks

30 minutes on a video call with the AI Kit team. You see a process like yours live, and afterwards we set up a test environment for you.

The schnell.digital AI Kit is the AI platform for mid-sized businesses that turns any AI model into reliable work in your processes, run in Germany or on your own servers.

  • GDPR compliantData processing agreement included
  • Hosted in GermanyData centre in Germany
  • EU AI PactSigned by schnell.digital
  • Made in GermanyDeveloped in Augsburg

Why AI has changed little so far

Individuals save time. The process stays the same.

01

You have tried ChatGPT or Copilot. Some employees now write faster with it.

02

Requests, quotes and complaints still run as before: someone copies, checks and forwards by hand.

03

The model is rarely the problem. It lacks your knowledge, access to your systems and clear rules.

An AI model on its own does no work. It needs instructions, your knowledge, access to your systems and clear limits. Experts call this layer the AI harness: everything built around an AI model so it works reliably. The AI Kit is exactly this layer.

Think of the shop floor: the model is the engine. The AI Kit is the machine around it, with controls, safety guards and a counter.

How a process is created

From a sentence in your own words to a process anyone can read

  1. Describe the task

    You describe the task the way you would explain it to a colleague.

  2. Get a proposal

    The wizard recommends a Workflow or an Agent and explains why.

  3. Read, adjust, test

    Each step reads as a sentence with gaps. The gaps are the settings: model, knowledge, role and rules. A test run in the editor shows the result before the process goes live.

More about Workflows and Agents

Your description

If an enquiry contains a complaint, prioritise it and forward it to the back office.

Turned into a process you can read step by step

Workflow editor in the AI Kit: the step “Analyze reviews” reads as a sentence with gaps, for example “I’m a customer feedback analyst” and “my goal is to summarize customer reviews reliably and classify them consistently”.

Where you start

Three processes mid-sized companies start with

Agent with Word output

Creating quotes

Today
The back office gathers earlier quotes, price lists and terms and writes the quote by hand.
With the AI Kit
Sales fills in a form. The Sales Offer Agent finds matching line items in earlier quotes, calculates the budget and puts the draft in the Inbox as a Word file.

24 hrs of work go into a detailed tender response on average.[1]

Sales use case
Inbox in the AI Kit: offer draft for a new customer with a table of line items, quantities, unit prices and the earlier offer each price comes from.
Workflow

Routing service emails

Today
Someone reads every email in the service inbox, classifies it and forwards it by hand.
With the AI Kit
A workflow reads the mailbox, summarizes each email, detects complaints and routes them to the right team. Replies go out directly or wait as drafts for review, just as you decide.

€8 is the average cost of a customer contact by phone, chat or email.[2]

Customer service use case
Wizard for new automations in the AI Kit: a service process is described in your own words, with examples such as “When an email arrives, summarize it and forward to the right person”.
Excel in, Excel out

Analyzing customer reviews

Today
Someone reads the reviews from the shop export, summarizes them and maintains a list.
With the AI Kit
The workflow reads the Excel export, classifies each review according to your guidelines and delivers a finished Excel file with an overview.

6 details per review: summary, sentiment, category, product, priority and action.

See it in the editor
Workflow “Customer Review Analysis” in the editor: the trigger “Upload review export”, followed by the actions “Read reviews”, “Analyze reviews” and “Create Excel report”.

New in autumn 2026

Agents that use your workflows. Results as Excel and Word.

Inbox: the Sales Offer Agent asks the agent “Service Advisor”. The handover and the reply on discount and warranty are expanded.

Agents use workflows and other agents

The Sales Offer Agent asks the Service Advisor and creates the Word document through a workflow.

Inbox: the Sales Offer Agent has placed the offer draft as a Word file in the chat.

The AI delivers Excel and Word files

Real xlsx and docx files, also based on your company template.

AI configuration of the Sales Offer Agent with model, knowledge bases, tools and the Pseudonymization switch.

Control built in

Pseudonymization via a switch on the agent, and an approval step with a deadline.

Also new: an editor with a step sidebar and cards, and sign-in with your company account.

Everything new in the autumn release

Use case with example calculation

How the Sales Offer Agent works in sales

Lumisette Inc is an office furniture manufacturer from our demo environment. Its sales team creates offer drafts with the Sales Offer Agent. Click through the four steps.

Form of the Sales Offer Agent with customer, new customer switch, special terms and requirements.
1 / 4 · Fill in the form

What it costs when quotes take too long

The calculation comes from our sales use case. Replace the assumptions with your own numbers.

Go to the sales use case

What slow quotes can cost you

Example calculation with assumptions
quotes
€
%
%

Lost revenue per year

€1,125,000

Quotes × order value × win rate × share of lost orders

For comparison: around three in four B2B buyers buy from the vendor they spoke to first.[3]

All prefilled values are assumptions. Enter your own numbers. The prefilled values describe a mid-sized project sales team with 500 quotes a year. You are best placed to estimate the share of orders you lose because quotes take too long.

30 minutes on a video call with the AI Kit team. You see a process like yours live, and afterwards we set up a test environment for you.

Control and sovereignty

Your data stays in Germany or on your own servers

The AI Kit defines what happens to your data and who may do what, and you can trace it.

Hosting in Germany or on-premise
The cloud edition runs in a data centre in Germany. On-premise means the AI Kit runs as a single Docker container on your own servers, with local models even without internet access.
DPA included, no training use
You get a data processing agreement (DPA). The AI Kit accesses your company knowledge, and the models are never trained on it.
Roles and audit log
Five roles, from Admin to AgentUser for agent-only access, define who may do what. The audit log records who changed what and when.
A person decides
An approval step pauses the process until someone decides. Emails can be created as drafts that a person releases. Names and addresses can be pseudonymized before the model sees them.
EU AI Act
Traceable steps, logs and human approvals help you meet the transparency and oversight obligations of the EU AI Act.
Audit log in the AI Kit: entries for sign-ins, started runs and changed automations, with CSV export. Only metadata is recorded, never content.
  • GDPR compliantData processing agreement included
  • Hosted in GermanyData centre in Germany
  • EU AI PactSigned by schnell.digital
  • Made in GermanyDeveloped in Augsburg

Free choice of model and costs

You choose the model and pay the provider directly

The AI Kit works with OpenAI, Anthropic, Mistral and local models via Ollama. You choose the model for each step and switch it without rebuilding the process.

With your own API key, the access key to your account with the model provider, the provider bills you directly. schnell.digital adds no markup to model costs.

  • Model selectable per step, for example a small one for sorting and a large one for replies
  • Your own API key, billed directly by the provider
  • Local models for operation without internet, with no cost per token
  • Effort setting with five levels from Low to Ultra: defines how thoroughly an agent works and what it may consume
Model configuration in the AI Kit with the provider's own access token, model selection and the reasoning effort setting.

Model costs for one process per month

Example calculation with assumptions
runs
tokens
€

Model costs per month

€100

You pay these directly to the model provider, with no markup from schnell.digital.

All prefilled values are assumptions. Enter your own numbers. 20,000 tokens per run is a reference value for a process with knowledge retrieval and several steps. €5 per 1 million tokens roughly matches the blended price of OpenAI’s GPT-5.4 at 80% input and 20% output tokens (OpenAI price list, October 2026). Smaller models cost far less, local models nothing per token.

Rollout

From first meeting to running pilot in 6 to 8 weeks

One process, one fixed plan. Then you decide on the expansion with real numbers from the pilot.

  1. Week 1

    First meeting and demo. You pick a process that pays off.

    Who: Managing director or head of department, with us

  2. Week 2

    Process review and workshop: define the process, knowledge, systems and rules.

    Who: Department and IT, with us

  3. Weeks 3 to 5

    Set up and test: connect knowledge, build the workflow or agent, run tests with real examples.

    Who: Us, or your team after the training

  4. Weeks 6 to 8

    Pilot in daily work with real cases, approvals and metrics. Training for the people who maintain the processes.

    Who: Department, with us

Two ways to get there

Roll it out yourselves

You do it yourselves

You book the licence, and we train you. After that, your team describes and maintains the processes on its own.

  • Processes in your own words, no programming
  • Training and documentation included
  • Support retainer available if needed

Licence from €349 per month

View licences and pricing

Have us roll it out

We handle the rollout

We choose the process with you, implement it, put it into operation and stay with you afterwards. Clear packages and scopes.

  • Use case selection and implementation by us
  • Sessions across the entire term
  • Scope described in writing

Packages from €950

See the consulting packages

Connecting your own systems, running on your own servers, training for larger groups: we handle anything beyond the packages individually. Request a custom offer →

Pricing

The licence starts at €349 per month

You pay a fixed licence per package. Your model provider bills you directly for model costs.

Cloud

from €349 per month

Operated in a data centre in Germany. Updates, monitoring and support included.

Inhouse

from €990 per month

The AI Kit on your own servers, with no storage limit and access to local resources.

All prices excl. VAT. Contract term 12 months. All licences and prices →

30 minutes on a video call with the AI Kit team. You see a process like yours live, and afterwards we set up a test environment for you.

Comparison

Why not just ChatGPT, Copilot or a custom build?

All three have their place. What matters is what you want to use AI for.

ChatGPTMicrosoft 365 CopilotCustom developmentschnell.digital AI Kit
Built forAssistance for individuals in a chatAssistance in Word, Outlook, Excel and TeamsExactly your process, as you program itFixed workflows and agents for entire processes
Your knowledge and your systemsThe provider's uploads and connectorsData from Microsoft 365, other systems via Copilot StudioYou develop every connection yourselfKnowledge bases, databases, email, Excel, Word, webhook, HTTP and MCP
Model choiceModels from OpenAIModels provided by MicrosoftFreeFree per step, including local models
Operations and dataThe provider's cloudMicrosoft cloudYour choice, and you run and maintain itData centre in Germany or on-premise
Control in the processEveryone checks results themselvesConfigurable with Copilot StudioAs far as you program itApproval step, roles, audit log, emails as drafts

As of October 2026. Features and licensing models of the products mentioned change constantly.

Questions about the AI Kit

Frequently asked questions

What is the schnell.digital AI Kit?

The schnell.digital AI Kit is the AI platform for mid-sized businesses that turns any AI model into reliable work in your processes. It provides everything around the model: instructions, knowledge, connections, model choice, control, measurement and operations. Experts call this layer the AI harness.

Which AI models can we use, and what do they cost?

OpenAI, Anthropic, Mistral and local models via Ollama. You choose the model per step. With your own API key, the provider bills you directly, and schnell.digital adds no markup. Local models have no cost per token.

Where is our data stored, and are models trained on it?

In the cloud edition, in a data centre in Germany. In the Inhouse edition, on your own servers. A data processing agreement is included. The AI Kit accesses your company knowledge, and the models are never trained on it.

Do we need programming skills?

No. You describe processes in your own words, and each step reads as a sentence with gaps in the editor. Your team keeps working in its familiar systems. Whoever maintains processes gets training.

How long does the rollout take?

From the first meeting to a running pilot takes 6 to 8 weeks. You roll out the AI Kit yourselves or book a consulting package from €950.

What does the licence cost, and how long does it run?

Cloud licences start at €349 per month, Inhouse licences at €990 per month, each plus VAT. The contract term is 12 months. You pay model costs directly to the provider.

Patrick Schnell

Let's look at your first process

In the demo, we show you the AI Kit on a process similar to yours. You leave with an assessment of whether and how it pays off for you.

We built the AI Kit because in more than 75 projects since 2017 we kept seeing the same thing: the model is rarely the problem. What's missing is the environment in which it works reliably.

30 minutes on a video call with the AI Kit team. You see a process like yours live, and afterwards we set up a test environment for you.

Prefer to check for yourself first? Start the process check (5 minutes)

Sources

  1. [1] Loopio, 2020: RFP Response Trends & Benchmarks
  2. [2] Gartner, 2019: Gartner Says Only 9% of Customers Report Solving Their Issues Completely via Self-Service
  3. [3] 6sense, 2025: B2B Buyer Experience Report

Example calculations are based on assumptions stated right next to the calculator. All images are screenshots of a sample tenant.