Collaboration
From use case to a running AI agent.
A fixed price for the setup, about 30 days to go-live, then operations with continuous tests. For work outside the kit I also take on projects or join your team.
Your AI agent at a fixed price, live in about 30 days.
- Today
1 Pick a use case
From more than 70 examples or described in the AI Project Planner. The plan with a budget recommendation arrives by email in 10 minutes, free of charge and with no obligation.
- In about 30 days
2 Live in your systems
Set up at a fixed price and tested with examples from your daily work. The agent only goes live once it passes the agreed test cases.
- Then ongoing
3 Operations with tests
I monitor the agent, roll out updates only after the tests pass and send you a report every month.
How the 30 days run
- Day 0Project startAfter signing we agree together on the infrastructure we build on. You collect real examples from your daily work: around 50 for a frequent task, all from the last months for a rare one. As soon as I receive them, the 30 days start.
- Week 1SetupI set up the agent in your systems and read your documents and examples.
- Week 2Quality round 1The agent works through your real cases, your specialist checks every result. I work their feedback into the agent.
- Week 3Quality round 2Your specialist reviews the revised version. Their judgement sets the test cases for acceptance.
- Week 4Acceptance and startOnce the agent passes the agreed test cases, it goes live with approval. In the training your team learns how to work with the agent, approve outputs and give feedback.
The fixed price depends on the size
- Size S: No system connection or just the mailbox, one language, with approval.
- Size M: Reads from one or two systems, such as CRM, ERP, file storage or Teams.
- Size L: Writes to your systems, connects three systems, works in several languages or creates images.
The plan shows the amount for your case as a budget recommendation, in 10 minutes and without a call.
Included in the fixed price
- Project start with an implementation plan and a checklist: access, documents and examples from your daily work
- Setup in your systems, approval by your team
- Two quality rounds with your specialist and acceptance against agreed test cases
- Training: how your team works with the agent, approves outputs and gives feedback
- 30 days of bug fixing after acceptance
- The third quality round after four weeks in operation
After that: operations for a monthly fee
Every change first runs against all test cases for quality assurance.
Basic
Runs safely, keeps the accepted quality level
- Monitoring and bug fixing
- Updates and model changes, tested against all test cases first
- Monthly report
Optimisation
Keeps getting better
- Everything in Basic
- Samples from live operation, rated against fixed criteria
- Your feedback analysed, one quality optimisation round per quarter
- One hour per month for adjustments
The agent runs on your side or on a server in Switzerland. If you want to run it yourselves, you get the runbook, tests and documentation.
Is this a fit for you?
A good fit if
- a clearly defined process comes up often, such as enquiries, quotes or invoices
- a rare process takes a lot of time and is worth a lot each time, such as a briefing and everything produced from it
- your systems have standard interfaces, such as the mailbox, Microsoft 365 or a common CRM
- there are examples the agent can learn from, such as past quotes, emails or briefings
Less of a fit if
- a process comes up rarely and has little value: then operations do not pay off
- the agent is supposed to make sensitive decisions without any approval
- many connections to home-grown systems are needed: then an individual project fits better
Three disciplines, one chain: data as the foundation, CRM as the engine, AI as the accelerator. I am an expert on all three layers and deliver from concept to finished systems, vendor-neutral and free of reseller ties.
Agentic AI for marketing and sales
Agentic systems that do real work: I design, implement and operate AI agents and multi-step workflows for marketing and sales, plus AI brains as central context layers over your company knowledge. On top, review and advisory for planned agentic setups before anything gets built.
CRM and marketing automation
From evaluation through implementation to operations: customer lifecycle marketing that measurably converts, enablement as part of every implementation, so that best practices and state of the art reach your team and make it more efficient, and MarTech operating models for real-time, trigger-based marketing.
Data foundations
The foundation underneath: connecting data sources, building data warehouses, cleaning, migrating and transforming data, and shipping ML models as data products right where marketing uses them.
Hands-on with CRM and marketing automation, Agent frameworks, Machine learning, Cloud and infrastructure
Salesforce Marketing Cloud, MoEngage, HubSpot, Adobe Journey Optimizer, Microsoft Dynamics, Mautic, Brevo, ActiveCampaign, Customer.io, BSI Customer Suite. Plus contributing to custom-built systems and developing complete custom CRM and marketing automation solutions.
Google ADK, OpenAI Agents SDK, Claude Agent SDK, LangChain, LangGraph, CrewAI, Microsoft AutoGen, Semantic Kernel, LlamaIndex, n8n
scikit-learn, XGBoost, LightGBM, TensorFlow, Keras, PyTorch, Hugging Face, MLflow, Vertex AI, Azure Machine Learning
AWS, Microsoft Azure, Google Cloud, Infomaniak (VPS, AI Services with open-source models, Swiss Backup), your own servers (any Linux server), Snowflake, BigQuery, Azure Synapse, Kafka, Terraform, Docker, Cloud Run, Azure Functions, Power BI
The handover curve
- BuildI build, your team works along from the start.
- Together, hands-onHandover: my share falls, your team's share rises.
- Your team, operations with meYour team works with the agent, approves outputs and gives feedback. Monitoring, tests, updates and adjustments stay with me.
What your team can do afterwards
Your AI agent keeps running, even without me.
The agent runs on your side or on a server in Switzerland. Repository, runbook, test cases and infrastructure as code are with you from the start.
What happens if …?
… I drop out?
A partner I have worked with for years steps in. Repository, runbook and tests are with you.
… you want to end the collaboration?
Your team is prepared from day one and takes over operations with the runbook and documentation.
… a new model comes out?
The update is tested against all test cases before it goes live, with a report.
… the agent makes mistakes?
Approvals run in several stages where needed. Every mistake becomes a test case, and outliers trigger an alert.
- A person approves before anything goes out
- Tests against all cases before every release
- Releases with a report
- Monitoring and alerts on outliers
- Costs in view
Afterwards: operations for a monthly fee, or your team takes over with runbook and tests
For work outside the kit: CRM and marketing automation, data foundations or larger agentic systems, in SMEs and large groups alike.
Starting point: the AI Project Planner, 10 minutes, free →
Individual project
- Fixed scope
- Fixed timeframe
- Fixed price
A clearly scoped initiative outside the kit, delivered in a fixed timeframe at a fixed price. With handover to your team or your IT provider, runbook included.
Part-time staffing
- 20 to 100 %
- Time-limited
- Subject to availability
I work as part of your team for a limited period: to push a discipline forward, drive a topic hard internally and train the team, or to cover for a key person.
As a fractional Head of MarTech or Head of AI
I develop your MarTech and AI stack, implement AI use cases and train your team. Fractional means you get this role part-time, without filling a full position, the way many companies already work with a fractional CISO.
In your team
I slot into your team and take on whatever role is needed.
Stand-alone
I deliver fully on my own and only need access, feedback and approvals from you.
Or just one part:ConceptImplementationOptimizationOperations
Also in tenders: as a subcontractor, named key expert or in a bidding consortium →
CONTACT
Let's talk about your project.
Tell me about your idea: on a video call from anywhere in German-speaking Switzerland, or over lunch in the Zurich area. I'll tell you straight away how I would build it.
I reply within one working day with a proposed time.
Prefer email? felix.zeeb@geistzeit.ch