AI solutions: integrate AI into your company
We analyse your workflows, show where AI takes work off your team and implement it with data protection in mind: from choosing suitable AI systems to running language models on your infrastructure or in a data centre in Germany or the EU.
What are AI solutions for companies?
AI solutions for companies are applications of artificial intelligence that take on or support tasks in existing workflows, such as answering questions from documents, sorting requests or writing minutes. BIST Consulting covers the whole path, from an AI opportunity assessment and the choice of suitable AI systems to operation and support. Multi-agent systems are a focus of BIST: a central agent assigns tasks to specialised agents, each drawing on its own data sources.
Services
AI opportunity assessment
In interviews and a workshop, we work out where AI helps your company.
- Interviews with management and business teams
- A workshop on tasks, data and the systems in use
- The result: a list of use cases, prioritised by your requirements
Choosing AI systems
The right AI system for each task, tested on your own examples.
- Large and small language models, plus reasoning models that work through a task step by step
- Decision models such as Clef or Laya, which vendors call System One models
- Speech to text, text to speech and image recognition
- Comparing quality, data protection, operating effort and ongoing costs
- A reasoned recommendation, tied to no provider
Multi-agent systems
A central agent distributes a complex task among specialised agents.
- Each agent draws on its own data sources and tools
- For tasks that span several systems and steps
- An architecture tied to no particular model or provider
Local language models
Language models on your own hardware or in a data centre in Germany or the EU.
- Inputs and documents go to no external AI provider
- Open models such as Llama, Qwen or Mistral, run with Ollama or vLLM
- Where it runs is up to you: in house or in a data centre
- A model size that fits your tasks
- Independent of any single provider, models can be swapped later
Knowledge bases and AI assistants
An assistant that answers questions from your own documents.
- For manuals, policies or product information
- Documents prepared in a searchable knowledge base
- Answers that name their source
- Documents with a period of validity only while they apply
- A quick start with our AI chatbot SOPHAI, or a solution of your own
AI in software development
We train your development team and guide the rollout of AI agents, controlled and traceable.
- Spec-driven development explained and taught, for example with OpenSpec
- Guided rollout of coding agents such as OpenCode or Pi
- Skills and project rules for recurring tasks
- MCP servers connect tools, knowledge sources and documentation
- Local models on request, so the source code stays in house
Workshops and training
Management and staff see where AI helps and learn to use it safely.
- An in-house presentation of where AI can help, for management and staff
- Workshops: where AI helps in daily work and where it does not
- Training in using language models and assistants, including checking results critically
- Handling confidential data correctly
- Content built on your own use cases
Ongoing support
Your system stays up to date and reliable after launch.
- Installing new model versions
- Checking answer quality against typical questions
- Adding new documents to the knowledge base
- Adjustments when your requirements change
Typical starting points and our answer
You want to use AI in your company but do not know where to start.
An AI opportunity assessment shows which use cases pay off first. In interviews and a workshop, we assess each use case.
AI opportunity assessmentYour staff often spend a long time searching manuals, policies or shared drives for answers that already exist in the company.
An AI assistant answers your staff's questions from your own documents and names the source of every answer.
Knowledge bases and AI assistantsYou want to use AI, but you must not hand confidential data to an external provider.
The language model runs on your own hardware or in a data centre in Germany or the EU. Inputs and documents go to no external AI provider.
Local language modelsYour team already uses AI services without it being clear which services they are and which data leaves the company.
In a workshop we record which AI services are in use and for what, and recommend privacy-compliant alternatives that your whole team can work with.
Workshops and trainingA task needs data from many sources and several steps, such as a customer enquiry that draws on your CRM, your inventory system and your documents.
A multi-agent system distributes such a task among specialised agents, each drawing on its own data sources and tools.
Multi-agent systemsYour development team wants to use AI agents, but in a controlled and traceable way.
We explain spec-driven development, train your team and guide the rollout: specifications with OpenSpec, coding agents such as OpenCode or Pi, skills for recurring tasks and MCP servers that connect your tools and your documentation.
AI in software developmentYou have to choose between several AI systems and need an independent assessment.
We test the systems on your own examples and give a reasoned recommendation.
Choosing AI systems
From our projects
ars mundi: an art adviser chatbot in the online shop
For ars mundi, we developed the “ars mundi Kunstberater”, an art adviser. The chatbot asks visitors to the online shop what they are looking for and shows them matching works from the range, with the artist and the price.
Geomapping: a chatbot for customer support
For Geomapping, a software provider for CAD, GIS and driving schools, we developed a chatbot that helps in customer support. It answers recurring questions from customers, so the support team can focus on more demanding requests.
Technologies we work with
Language models
- Llama
- Gemma
- Phi
- Qwen
- Mistral
- DeepSeek
- GLM
- Kimi K2
- gpt-oss
Running them
- Ollama
- vLLM
Vector databases
- Qdrant
- Pinecone
Developing with AI
- OpenSpec
- OpenCode
- Pi
- MCP
Why BIST
Research meets practice
We know new AI concepts from research before they reach finished products.
Data protection from the start
As TÜV-certified data protection officers, we plan every solution so that your data stays protected.
Consulting and implementation from a single source
We support you from the first conversation through to operation.
Vendor-independent
We are not tied to any vendor and recommend what suits your task.
Two dedicated contacts
Both founders know your project. You always speak to someone who builds it.
Talks
What AI can do today, where its limits lie and how to introduce it responsibly.
- AI in mid-sized companies: from trying it out to putting it to work
- Language models in house
- What AI can and cannot do
Frequently asked questions
What are LLMs?
LLMs (large language models) are AI models trained on very large amounts of text, which enables them to process language and generate text. A language model answers questions, summarises documents and drafts texts; well-known models include OpenAI's GPT, Meta's Llama and Mistral AI's Mistral Large. BIST Consulting runs language models locally, on your own hardware or in a data centre in Germany or the EU. Your data then does not go to an external AI provider.
What are decision models?
Decision models are small AI models that answer a question with fixed options and a probability instead of generating text. Vendors also call this category, which emerged in 2026, System One models. Typical tasks are sorting requests and routing them to the right place. Examples with open weights are Clef from Cloudflare and Laya, which is particularly small at 421 million parameters. Both are released under the Apache 2.0 licence and can run on your own hardware. When choosing AI systems, BIST Consulting checks whether a decision model is enough for a task.
What does an AI project cost?
It depends mainly on your data, how you want to run the system and how it connects to your existing systems. If the model runs on your own hardware, you also buy that hardware; alternatively, we run it on a server in a data centre in Germany or the EU. We work as a project with an agreed scope and price or with a monthly budget of hours. The intro call is free of charge.
What hardware does a local language model need?
What matters most is the memory of the graphics card, because it determines which model size runs fast enough. Smaller models already run on a well-equipped workstation; larger ones need a server with a powerful graphics card. We first check whether your existing hardware is sufficient. If it is not, we run the model on a server in a data centre in Germany or the EU.
Are local language models as good as ChatGPT?
For many business tasks, yes: answers from your own documents, summaries or drafts, for example. For very broad general knowledge or complex reasoning, large cloud models are often still stronger. That is why we test models on your own examples before every recommendation and show you where a local model reaches its limits.
How do you protect confidential data?
We design every system with data protection in mind: we clarify which data is processed, where it is stored and who can access it. If the model runs on your premises or in a data centre in Germany or the EU, inputs and documents do not go to an external AI provider. Both founders are data protection officers certified by TÜV NORD.
How does AI help in software development?
AI agents take on routine work such as tests, refactoring or documentation and speed up new features. To keep the result reliable, we work with spec-driven development: first comes a specification, for example with OpenSpec, then AI agents implement it, and your team reviews the result. Skills and MCP servers give the agents knowledge of your systems. With local language models, your source code never leaves the building.
Let’s talk about your project
Before we write a proposal, we take the time to understand your company and your requirements. The intro call is free of charge.
Book an intro call