Engineering consultancy
AI that takes work off your teams and keeps your data in the company
We help companies hand real work to AI, such as checking invoices, answering service questions or preparing quotes. We look for the processes where that pays off, build the system on your servers or in your own cloud account, and test it on your past cases before your team relies on it. Your data stays in the company throughout.
The first call is free and takes 30 minutes. After it, you get a fixed price for each step in writing.
Our team has built software and production AI for
Where companies usually start
All use casesAccording to Bitkom's 2026 survey, 57% of German companies with 20 or more employees use AI, but 87% say they use its potential little or not at all. The processes below are good places to start, because they come up often, follow rules and can be measured.
By industry: AI in manufacturing; AI in chemicals, pharma and medtech; AI for energy and utilities; AI in logistics; AI for banks and insurers; AI for law, tax and audit firms
How we build it
The layers underneath every use case
Every use case rests on the same four layers: servers or a cloud account, your data, the agents themselves and the platform they run on. Many providers cover either the infrastructure or the agents. We do both, because an agent is only as reliable as what sits underneath it.
All services01Infrastructure
Servers or cloud accounts for AI
GPU servers in your own data centre, or Azure OpenAI and Amazon Bedrock in a cloud account you control. The LLM gateway and model serving sit on top.
02Data
Your company data, connected
We connect file shares, SharePoint, your ERP and the ticket system together with their permissions. Assistants then find the right source, and nobody sees documents they are not allowed to see.
03Agents
Assistants and agents that do the work
They answer from documents, prepare cases and automate steps in a process, and a person approves anything that matters. Coding agents for engineering teams are one example.
04Agent platform
A platform on your servers when you need one
Once several teams use agents, they need a shared gateway, model serving, tools and logs. We set this up as infrastructure as code, and either your team or we run it.
How it works
Three steps, each with its own fixed price
After the first call you get a written proposal. You can stop after any step and keep what has been built so far.
- 01
Scoping or acceptance test
We pick one or two use cases with a measurable goal and work out how much hardware or cloud capacity they need. If you already have an AI system, we run an acceptance test on it instead and write up the results.
- 02
Build
We build the assistant or agent and the platform underneath in milestones. The acceptance criteria are written into the contract.
- 03
Run
After handover your team runs the system, or we run it for you on a monthly retainer, with regression tests and a quality report.
AI for your role
All rolesA managing director has different questions from a CFO, an IT lead or a works council. Pick the page that fits your role.
Your data stays under your control
Your servers, your cloud or both
We work this out with you during scoping, before you buy anything. The technical setup is the same either way: Kubernetes, a LiteLLM gateway in front of the models and infrastructure as code. If you move later, you do not have to start again.
On-premise AIOn your servers
This makes sense when data must not leave your network, or when usage is high and steady enough for your own hardware to pay off.
In your cloud
An account you control on AWS, Azure or the provider you already use. This suits a quick start, or a load that changes a lot.
Both, behind one gateway
Some data needs stricter rules than the rest. Sensitive work then runs on your own models, and everything else on hosted models in an EU region.
What we build
Assistants and agents on your own documents
Each one starts with a use case you can measure, such as hours saved or cases prepared. We build it in your systems and hand it over with training.
Service assistant
Technicians and the hotline get answers from manuals, spare parts lists and past tickets.
Quotes and tenders
For each new quote, the assistant finds the requirements, earlier offers and the right product data.
Process agents
An agent prepares one step of a process, such as checking an incoming order, and hands it to a person.
A small model for one task
For a task that comes up very often, a small model trained on your data can be cheaper to run than a large one.
Testing
Tested before go-live
A demo shows that a system can answer. An acceptance test shows how often it answers correctly: 100 to 200 cases from your own documents, a required pass rate for each criterion agreed in advance, and a confidence interval around every result.
| Criterion | Required | Cases | Measured | Confidence interval | Result |
|---|---|---|---|---|---|
| Answer is correct and complete | 85% | 200 | 94.0% | 89.8% to 96.5% | Met |
| Cites the right document | 90% | 200 | 95.0% | 91.0% to 97.3% | Met |
| Hands over to a person when it should | 90% | 120 | 94.2% | 88.4% to 97.1% | Not decided yet |
The third criterion is still open, because the required rate lies inside the interval. If the pass rate holds, about 200 cases would settle it.
Clients
What clients say
“They were embedded in our team, which was really effective.”
“We are impressed with their engineers' high-quality work.”
“Their responsiveness, friendliness, and successful project delivery were top-notch!”
- 01
Your environment, your rules
Every data flow is agreed with you before work starts.
- 02
A data processing agreement in every contract
With the documentation your data protection team needs.
- 03
Teams and tasks, never individuals
Works council requirements are handled before anything goes live.
- 04
Everything we build stays yours
Code, test sets and runbooks, written for your team to run.
Agentic coding
From assisted coding to bounded autonomous workflows
Autonomy expands one task type at a time. Your team starts with assisted coding, moves to supervised agent tasks and hands narrow, well-tested work to background agents only when pilot results support it. Each step has approvals, fallbacks and stop conditions agreed in advance, and decisions rest on measured cost per accepted change. Your engineers keep review and release authority throughout.
Works with
LLM cost optimisation
Spend by team and task, then budgets, defaults and routing.
Harness engineering
Context, checks and review rules that make agents reliable.
Coding agent adoption
One shared setup and training on real tickets.
Evaluations on your repositories
Replays of merged changes before you switch tools or models.
Team
The people behind the work
Mickey GrafFounder and CEOFrom the founder
Models get cheaper every quarter. Running them well doesn't
New models arrive every few months. Whether they pay off depends on the work around them: context, checks, routing and measurement. That is the work we do.
Mickey GrafFounder and CEO
Ask AI for a summary of Cloudsail
Pick a topic and open it in your assistant.
Summarise what Cloudsail (cloudsail.com) does, which services it offers and how a first project with them runs.
FAQ
Common questions
Does AI pay off for a company our size?
Usually yes, if the process runs often, follows rules and takes up hours of people's time. We check this on your own cases first: the hours and errors per case today, the quality the system reaches in a test and what it would cost to run. If the numbers do not work, we tell you before anything is built.
Where should we start with AI?
With one process you can count, such as quotes per week, invoices checked or questions to the service desk, and a clear idea of what success looks like. We build that first system, measure it and plan the next step from the results.
What does AI cost to run?
The model is usually not the biggest item. In Bitkom's 2026 survey, companies using AI named infrastructure, data preparation and integration into existing systems as their biggest cost items. Before you decide, we show you what each handled case would cost, and each step gets a fixed price agreed in writing after the first call.
Does our data leave the company?
Only if you decide it may. We run the models on your own servers or in your own cloud account in an EU region, connect your documents with their permissions and keep the logs in your own storage. Your data protection officer sees the data flow before anything is built.
What does Cloudsail do?
We set up the infrastructure for AI on your servers or in your own cloud account, connect your company data and build the assistants and agents that work with it. Every system is tested before go-live and again after each change. We also roll out coding agents for engineering teams.
How do we deploy an LLM on-premise so data never leaves the company?
Serve an open-weight model with an engine such as vLLM on your own GPU servers, put a gateway in front for keys and logs, connect your documents with their permissions, and test the answers before go-live.
Should our AI run on premise or in the cloud?
It depends on your data, your usage and how quickly you want to start. Your own servers suit data that may not leave your network and high, steady usage. A cloud account you control suits a quick start and a load that changes. We compare both during scoping, and the setup is the same either way.
Our data is scattered. Can agents still work with it?
Yes, if the few sources a use case needs are connected first, together with their permissions. You do not need a company-wide data project before you start.
Which processes should we give to AI agents first?
Work that comes up often, has a clearly correct result and can be checked quickly by a person. Rare cases, unclear outcomes and anything that cannot be undone come later, or stay with people.
How do we know the AI system works before we accept it?
Agree acceptance criteria in advance, test 100 to 200 cases from your own work and accept the system on its measured pass rates, each with a confidence interval.
Can we use AI without sending data to a US cloud?
Yes: run open-weight models on your own servers or with an EU provider. Choosing the EU region of a US provider decides where your data is processed. It does not change which law the provider is subject to.
Does the works council have to agree to an AI system?
In Germany, often yes. Co-determination applies when a system is objectively suitable for monitoring performance or behaviour, and logs usually make it so. Agree early on what is logged and who can see it. This is not legal advice.
What is the free AI usage review?
You send the usage and billing exports you choose to share, and we send back one page on where your AI and coding agent spend goes, what to change first and what to measure next. It is free and needs no access to your systems.
Do we have to buy hardware?
Not at the start. A first use case can run in a cloud account with GPUs in an EU region, and move in house once usage and results justify the hardware.
Who runs the system after go-live?
Either your team, with the runbooks and training we hand over, or we do, on a monthly retainer with a named engineer.
What does a first project look like?
It covers one process, team or model, and we agree how to measure it before we start. We take a baseline, change one thing at a time, test each change against the baseline and hand over the setup with training.
Are you consultants or engineers?
Engineers. The people who scope your project build it with your team and hand it over.
Will this work with our existing codebase and systems?
Yes. We work in the repositories, tools and processes you already have, without a rewrite or a new platform.
We already pay for Copilot, Cursor or Claude Code. Why has nothing changed?
Licences help individuals. Results come from the setup around them: the context agents see, the checks before merge, budgets and routing, and training on real tickets. That is the part we set up.
Do your agents decide on their own?
Only within limits you set. Each agent starts by preparing work for a person to decide, and takes on more only after you have reviewed its results with us.
Which models and tools do you use?
Whichever suit the task: commercial models such as Claude, GPT or Gemini, open-weight models on your own hardware, and the coding agents your teams already have.
What does it cost?
It depends on the scope and the size of your teams. You get a written proposal after the first call, and the first conversation is free.
What if a change doesn't pay off?
Then we leave it out. You keep the test set and the measurements either way.
Start with one use case
Bring a process you would like to hand to AI, or an AI system you need to sign off. You talk to an engineer for 30 minutes, and we do not need access to your systems.
Look for the draft in your email program
Send it from there, and an engineer replies in person with times for a 30-minute call. If no draft has opened, write to us at miki@cloudsail.com.