Co-Build: Your Process, Your Agent
4-6 weeks. Source code access. No SaaS.
We build AI agents for your specific process - not as a standard product, but as a co-build with your team. You receive the source code, prompts and rule sets. After the transfer, you operate the agent independently.
At a Glance
4-6 Weeks
From discovery to a production agent in your infrastructure
Full Source Code
Code, prompts, rule sets, documentation - everything belongs to you
No SaaS
No monthly fees. No vendor lock-in. Independent operation possible.
Decision Layer
Every agent with audit trail, Human-in-the-Loop and rule set versioning
Model-Agnostic
Claude, ChatGPT, Gemini, Llama - swappable without code changes
Exit Strategy
Training, documentation, handover. Fully independent after 12-18 months.
Three Phases to a Production Agent
Every Co-Build follows a structured process. The result is a production agent in your infrastructure - with source code, documentation and exit strategy.
Phase 1 - Week 1
Discover
- Process analysis with your domain team
- Identify and formalize rule sets
- Map decision points
- Decision Layer design
- Clarify system landscape and integrations
Result: Process blueprint with decision matrix
Phase 2 - Weeks 2-5
Build
- Agent development with your team
- Implement Decision Layer
- Rule sets as versioned configuration
- System integration (SAP, DATEV, etc.)
- Testing with real data
Result: Production agent in your infrastructure
Phase 3 - Weeks 5-6
Transfer
- Source code handover
- Team training for independent operation
- Documentation of code, prompts, rule sets
- Set up monitoring and alerting
- Document exit strategy
Result: Your team operates the agent independently
For Which Areas
Co-Build is suitable for any process that is based on rule sets and is currently handled manually by people. Typical areas:
Compliance & Legal
Regulatory reviews, contract review, clause extraction, deadline monitoring.
Operations & Supply Chain
Order processing, supplier management, quality inspection, complaint processing.
Shared Services
Ticket routing, request classification, internal knowledge provision, service desk automation.
Industry-Specific
Export controls, customs clearance, insurance claims, credit assessment - any rule-based process.
What You Get
- Source Code: Complete agent code in your repository
- Prompts: All prompt chains and system prompts, documented and versioned
- Rule Sets: Domain rules as versioned configuration, not hardcoded
- Decision Layer: Audit trail, Human-in-the-Loop, works council or employee oversight constraints
- Documentation: Architecture, decision logic, test cases, deployment guide
- Exit Strategy: Training, handover protocol, support period
Co-Build vs. SaaS vs. Inhouse
| Co-Build | SaaS Platform | Inhouse | |
|---|---|---|---|
| Time-to-Value | 4-6 weeks | 2-4 weeks | 6-12 months |
| Source Code Access | Full | No access | Full |
| Vendor Lock-in | None | High | None |
| Governance / Audit | Decision Layer included | Platform-dependent | Must be built |
| Ongoing Costs | Infrastructure + optional maintenance | Monthly license | Personnel costs |
| Customizability | Unlimited | Within the platform | Unlimited |
In Practice: Compliance Review in a Shared Service Center
A shared service center reviews 120 supplier contracts per month for compliance - manually, with checklists in Excel.
Before - manual
- 45 min per contract (reading, checklist, documentation)
- ~90 hours/month for the team
- 12% error rate in clause detection
- No versioning of review logic
After - Co-Build Agent
- 82% autonomous review (standard contracts)
- ~19 hours/month (exceptions + spot checks only)
- <2% error rate via rule-based review
- Decision Layer with full audit trail
Result: 71 hours saved per month. Agent productive in 5 weeks. Source code, rule sets and prompts owned by the client. Independent operation after 14 months without external support.
Deep Dive in the Agent Briefing
Our article series for decision-makers implementing AI agents in the enterprise.
Other Agent Types
The Specialist
Document Agents
Read, understand and classify documents - with true language comprehension.
The Coordinator
Workflow Agents
Orchestrate processes across systems - SAP, DATEV, SharePoint.
The Knowledge Keeper
Knowledge Agents
Answer questions from rule sets - with source references and rule versioning.
Frequently Asked Questions about Co-Build
What exactly do I get with Co-Build?
The agent's source code, all prompts and prompt chains, configured rule sets, Decision Layer with audit trail, documentation, test cases and an exit strategy for independent operation.
Do I need my own AI team?
Not to start. During the co-build, our engineers work with your domain team. In the transfer phase, we train your team for independent operation. Internally, you need at least one technical contact and one domain process owner.
How does Co-Build differ from a SaaS solution?
With SaaS, you rent a platform and depend on the vendor. With Co-Build, you receive the source code, prompts and all rule sets. You can develop the agent further yourself, run it on your own infrastructure and switch vendors at any time.
What happens after the 4-6 weeks?
The agent runs productively in your infrastructure. We offer optional maintenance and further development afterwards - but no obligation. Most clients take over full independent operation after 12-18 months.
Which process should your first agent handle?
30 minutes. Your use case. Concrete recommendation.