Opens in a new tab
Case studies/Pensions/PGGM

PGGM gives risk and compliance teams one view of every live AI

PGGM runs its AI models from a single platform, where risk and compliance teams can see what is live, how each model decides and what evidence exists. Case handling is twice as fast.

About PGGM

PGGM is a Dutch pension administrator and asset manager, managing pensions for millions of participants on behalf of pension funds in the healthcare and welfare sector. Its decisions affect people's retirement income, so it must be able to explain and justify every one of them to participants, pension funds and supervisors.

IndustryPensions and insurance
FocusOne line of oversight
The challenge

Challenges that stood in the way

No central view of AI

Models for use cases such as risk assessment, portfolio optimisation and fraud detection were spread across departments. Nobody had a single view of performance, compliance status or risk.

Decisions that were hard to explain

In a sector where outcomes affect retirement income, stakeholders expect a clear reason behind every AI-driven decision.

Growing regulatory demands

Meeting the EU AI Act and supervisory expectations meant keeping documentation and oversight current for every model, which was hard to do by hand.

The solution

How PGGM and Deeploy work together

PGGM uses Deeploy as its central platform for deploying, serving, monitoring and governing AI models. Data teams bring models into Deeploy, where performance, explanations, feedback and documentation are kept together. Risk and compliance teams work from the same platform, so oversight is based on live information rather than periodic reports.

The platform

How the Deeploy platform is used

Scattered models across departments are replaced by a single, clear view of every use case, with all AI models managed in one central Deeploy platform.

Explore AI InventoryAI Inventory in the Deeploy platform

Consistent oversight of performance, compliance and governance applies to every model in one place, keeping operational risk low as PGGM’s AI use grows.

Explore Risk & Control FrameworksRisk and Control Frameworks in the Deeploy platform

Performance and drift are monitored continuously and every prediction is traced, with model cards, deployment details and audit trails stored centrally, so audits and internal reviews are fast, easy and fully documented.

Explore Monitoring & Audit ReadinessMonitoring and Audit Readiness in the Deeploy platform

An EU AI Act-based AI code of conduct applies across every PGGM model through Deeploy, with central oversight and audit trails making adherence easy to show.

Explore Policy & GuardrailsPolicy and Guardrails in the Deeploy platform
Outcomes

What changed for risk and compliance

One inventory of every AI model

All models now run from a single platform. Risk and AI governance officers can see which models are live, who owns them, how they perform and where they stand on compliance. Operational risk drops because no model runs outside the line of sight.

Decisions that can be explained and reproduced

Every prediction can be traced and reproduced. Explanations show which factors drove each outcome, so a decision can be justified to a participant, a pension fund board or a supervisor. With those explanations to hand, cases are handled twice as fast.

Oversight based on live data

Deeploy tracks traffic, performance and drift for each model and alerts teams when behaviour changes. Problems are caught while they are small, and retraining happens when the data says it is needed, not when the next review comes round.

Expert judgement built into the models

Feedback from domain experts is captured and used to refine model decisions. Accuracy improves over time and false positives fall, which reduces manual reviews.

Audit-ready by default

Model cards, deployment details and audit trails are stored centrally and kept up to date. When an audit, a pension fund client or the supervisor asks for evidence, PGGM can provide it straight from the platform. The same records support its preparation for the EU AI Act.

Technical details

How a model is governed in Deeploy

A data team onboards the model into Deeploy, which handles deployment and serving.

Deeploy registers the model in the central inventory with its owner, use case and version.

Each prediction is logged with its explanation, so it can be traced and reproduced later.

Monitoring tracks traffic, performance and drift, and sends alerts when thresholds are crossed.

Domain experts give feedback on model outcomes, which flows back to the data team for retraining.

Model cards, deployment details and the audit trail are updated with every change.

Who uses it
Data teamsDeploy, monitor and retrain models
Domain expertsReview outcomes and give feedback
Risk and complianceOversee the model inventory and pull audit evidence
Internal auditReview documentation and trace individual decisions
Case studies

More success stories

Cybersecurity

Hadrian scales AI and MLOps, deploying models 80% faster

See case study
Banking

Bunq governs AI-driven banking with 40–60% fewer false positives

See case study
Manufacturing

Vlisco keeps factory AI accurate, explainable and owned

See case study
Hospitals

Healthplus.ai earns clinicians' trust in infection prediction AI

See case study
Insurance

RISK keeps pricing and claims models accountable in real time

See case study
Mental healthcare

NiceDay keeps therapists in charge of every AI recommendation

See case study

Stay in control
after go-live

Monitor, trace and explain every AI system, and keep the evidence ready for review.