Agentic AI meets ROI – our recap in 4 days, 3 talks and many tangible learnings
Four days of FUSION, three presentations – and one common thread: Agentic AI only succeeds with orchestration, governance, and measurable benefits. In this recap, you'll find the most important product news, our highlights for quality engineering, and the short story of how we took AVALON from hackathon winnings to production.
Our 3 talks in brief
1. Main Stage – Agentic Testing (with Ingo Philipp):
Ingo, Benjamin Tan, Johannes Reitermayer and I have shown how you can safely move from PoCs in regulated environments to production – with clear guardrails, KPIs and roles in delivery.
2. Extended Community Session – UiPath Playground:
Deep dive with the community into architecture, governance and “controlled agency” – what do you allow the agent to do, and what not?
3. Speakers' Corner (Main Hall):
A pragmatic blueprint for setting up a test data agent, integrating it into Jira/CI/CD, and tracking the ROI.
What UiPath has delivered (the most important product news)
Guiding principle: Agentic AI meets ROI. Less hype, more end-to-end orchestration of humans, bots and AI agents – with verifiable added value.
Maestro everywhere: Orchestration + Case Management + Process Apps connect people, automation, and agents to create real business flows (e.g., claims, lending, disputes). This is the connecting fabric for agentic scenarios.
AI Agent Builder (Studio): Unified agent building – from low-code to fully coded agents – including templates, debugging, and performance optimization. Quality is considered right from the start of the development process.
Natural Language Automation: Screenplay translates intentions into desktop actions; API workflows allow you end-to-end processes without UI fragility.
Quality Engineering, significantly expanded: Test Cloud for scale-out, integrated performance testing, self-healing for UI tests and autopilot in the test manager – the step towards “agentic testing” across the entire SDLC.
Trust & Governance: Stronger guardrails, PII protection and improved audit capabilities to ensure that autonomy remains secure from both a corporate and audit perspective.
Ecosystem moves: Closer connections to OpenAI, Microsoft Azure AI Foundry, NVIDIA NIM and Snowflake Cortex AI – moving from insight to action faster, without losing governance.
Why this matters for QA & Engineering
- Faster results: Vertical Solutions + Maestro reduce last-mile effort in real-world processes.
- Scalable with confidence: Agent guardrails and audit trails create trust in regulation and auditing.
- More quality at lower costs: Performance tests from the cloud, self-healing and autopilot increase coverage & speed, reduce maintenance.
Our use case: AVALON – synthetic test data for Avaloq (with LGT)
The hackathon prize resulted in a practical agent: AVALON generates compliant synthetic test data "on demand" for Avaloq and writes the artifacts centrally to our Test Data Orchestrator (Jira). No production data is used in the test – yet high variance and realistic scenarios are achieved.
Here's how it works (in short):
You describe the need in natural language (e.g., "a minor customer from Zurich who is rejected during the credit check").
The agent generates consistent, Avaloq-compatible data sets (including PL/SQL for import).
TDO centrally documents the test data and feeds it into your workflows (Jira/CI/CD).
The effect:
- Test data via prompt, preparation time reduced
- PII risk eliminated
- “Shift Left”: realistic data early on for Dev & QA.
AVALON's victory in the "Agentic Testing" track at AgentHack 2025 underlines the maturity of the approach – and how seamlessly UiPath components (Agent Builder, Test Suite, Orchestrator, etc.) work together.
Agentic Testing – what you can do today
- Define governance first: Set policies, approvals, observability, and KPIs for agents before you send them out.
- Choose use cases with high leverage: bottlenecks in test data, repetitive regression, performance risks in critical journeys.
- Think modularly about tooling: Maestro for orchestration/case handling, Agent Builder for skills, Test Cloud for scale, TDO for data flow & traceability.
- Measure ROI: Lead Time, Defect Escape Rate, Test Coverage, Stability (Self-Healing Events), Time to “Data-Ready”.
Who is this relevant for?
1. Banks & insurers: High regulatory requirements, sensitive data, many end-to-end processes.
2. Industry & MedTech: Compliance-critical processes, quality requiring documentation.
3. Platform teams (SAP, Avaloq, Core Banking): Test data bottlenecks, release pressure, need for automatic scaling.
What we do for you
- Assessment & Roadmap: Where can agents deliver immediate added value? What guardrails do you need?
- AVALON pilot (Avaloq): From "Prompt" to near-production testing – safe, auditable, scalable.
- QE enablement: Performance testing in the cloud, self-healing, autopilot in the test manager – as a package or modular building blocks.
Conclusion
FUSION 2025 marks the transition from AI features to operationalized agentic AI – with orchestration, data and process governance, and measurable outcomes. If you want to deploy agents productively, securely, and with real ROI, we start with your process – not the model.
Fancy a 60-minute sparring call?
Send us your target image (use case + stack), and we will outline the architecture, guardrails, and KPIs for a provable pilot.
