Our services for AI agents and workflows
Automation pays off where people move data between systems – and that is exactly where undocumented deviations often arise in GxP. We automate those paths with agents and workflows without losing the regulatory logic.
Start
Process capture & use case
We capture volumes, handling times, error sources and GxP relevance. That yields a defensible automation use case with clear inputs and outputs – before technology is chosen.
Deliverable: process pack with feasibility, risk and ROI estimate
- Boundary: automation vs. custom software
- Data flow between CRM, ERP, LIMS and email
- Prioritisation by audit risk and effort
Agent
Configure AI agents
Agents for bounded tasks with a clear success criterion: written task description, defined data sources and a limit at which control passes to a person.
Deliverable: working agent with documentation of task, sources and escalation
- Pre-qualification of enquiries and routing
- Knowledge search with a source for every answer
- Draft reports for human review
- OpenAI, Claude – model choice justified per use case
Automation
Workflows with Activepieces & n8n
We connect agents and business systems via workflows you can read and audit. Both platforms are open source and self-hostable – data stays in your infrastructure.
Deliverable: production workflow with log per run and error handling
- Integration with HubSpot, Salesforce, SAP and lab IT
- Email, approvals, notifications
- Retry and alert to responsible owners
GxP
Validation & CSV
For regulated use we validate workflows to CSV standards. Audit trail, change control and data integrity are planned before implementation.
Deliverable: URS, risk assessment, test cases and validation report
- Version control of workflow configuration
- Logging: input, result, timestamp
- Integration into change control
Compliance
EU AI Act & transparency
Automated decisions sit under GMP law and the EU AI Act at once. We document risk class, human oversight and transparency duties – including Art. 50 labelling where required.
Deliverable: role and transparency documentation under Regulation (EU) 2024/1689
- Risk classification and prohibited practices check
- Human oversight at GxP decision points
- Link to EU AI Act check
Operations
Run, monitoring & handover
We operate the platform during the project or hand over runbooks, monitoring and training. MCP integration enables standardised connection of further AI systems.
Deliverable: operated automation or documented handover to your IT
- Self-hosted in your environment
- Extension to further workflows as a programme
- Case story: Access to web app
What sets us apart: An AI agent does not replace release. We build human oversight into the workflow – instead of documenting it afterwards.
Process with manual transfers?
Book an initial callAI pathfinder
Which starting point describes your project?
Choose your scenario – you get the right entry point with references and contact route.
Automate manual transfers
Data from system A to system B, email approvals, notifications: we map the process and build the first workflow in Activepieces or n8n.
AI agent for internal knowledge search
Make SOPs, validation documents and supplier files searchable – with a source for every answer and handover to QA when uncertain.
Validate workflow for GxP operation
URS, risk assessment, test cases and validation report – the workflow sits under change control like any other computerised system.
Clarify EU AI Act duties
Risk class, transparency, human oversight: we document deployer duties and link them to your GxP process.
Programme with multiple workflows
Shared templates, version control and change control across parallel automations – PMO logic for your agent portfolio.
Automation from one source means process, agent, workflow and validation evidence from one team – without a handoff gap between business, IT and QA.
Reference: Access-to-web-app migration in under 270 hours. Read case story
Three roles in an audit-ready workflow
GxP decisions stay with people. The agent supports, the workflow logs – the system remains under change control.
Release & oversight
Qualified staff review GxP-relevant results and release – Annex 11 and Art. 14 EU AI Act.
Task & escalation
Bounded sub-steps with source citation; handover when uncertain or at GxP threshold.
Log & version
Audit trail per run, configuration version, change control on changes.
An AI agent does not replace release. EU GMP Annex 11 requires additional review for critical data, and Art. 14 of Regulation (EU) 2024/1689 demands effective human oversight for high-risk systems. We build both into the workflow.
EU AI Act check
For provider and deployer duties, use the public EU AI Act check: eight questions on risk classes, prohibited practices, transparency, human oversight, documentation and GPAI under Regulation (EU) 2024/1689.
Start EU AI Act checkTechnology stack
What you receive
Who are AI agents and workflows for?
IT leadership, business units and QA in pharma, biotech and labs – when manual data transfers, undocumented email approvals or hard-to-search SOP libraries create audit-relevant risk.
Regulatory framework: EU GMP, ICH and FDA references
Automated decisions sit under two rule sets at once: GMP law and the EU AI Act.
- EU GMP Annex 11: risk management, audit trail, access control and additional review of critical data.
- Regulation (EU) 2024/1689: risk classes, documentation, logging and human oversight for AI systems.
- ISPE GAMP 5, 2nd Edition: risk-based test depth, including for configurable and learning components.
AI agent or classic workflow?
Not every automation needs a language model. Rule-based workflows are often the faster entry – agents where interpretation and knowledge search are required.
AI agent
- Interpretation, knowledge search, drafts
- Source citation and escalation to people
- Higher documentation and AI Act effort
Workflow (rule-based)
- Fixed rules: if A, then B
- Data transfer, approvals, notifications
- CSV often simpler, less AI Act complexity
