From idea to investable AI

Why de-risk AI investments now

AI ambition is rising faster than financial certainty and business confidence, forcing leaders to decide where to place early bets while minimising the risk of sunk costs. In Grundfos’ IS Service Desk, ticket volumes are rising while headcount must remain flat. The priority, therefore, is to enhance responsiveness and quality through technology without expanding the team. 

Rather than building first, Grundfos chose to learn first. We designed an assumption-led innovation sprint to stress test and thereby de-risk the most promising use case – a generative AI chatbot in Microsoft Teams that structures and routes ServiceNow tickets – and to establish a credible roadmap from assisted intake to self-help and automation.

Inside the sprint

Assumption validation can accelerate learning. So, we began with a Use Case Day to prioritise opportunities and align stakeholders. In preparation for the sprint, we mapped over 50 assumptions across three lenses: desirability, feasibility, and viability, then prioritised a handful for in-depth validation.

During the sprint, we:

• Built prototypes of a Teams-based chatbot for structured ServiceNow ticket creation.
  
• Conducted interviews with end users, service agents, onsite technicians, and a Grundfos Machine Learning developer.
  
• Tested chatbot flows, gathered reactions, and refined scope and success criteria.
  
• Investigated platform constraints and integration paths in Azure and ServiceNow.

From insights to roadmap – sequencing value from Gen 1 to Gen 3

The validation work across desirability, feasibility, and viability surfaced the root causes of user frustration in the current workflow and clarified both solution preferences and adoption risks. Consequently, generation 1 of the solution evolved further towards enabling employees to describe IT issues in natural language, allowing the chatbot to autonomously generate structured tickets in ServiceNow. Over time, in generations 2 and 3, the vision is to expand the chatbot’s capabilities to offer direct resolution of IT issues, reducing reliance on human support.

This led to the formulation of generation 1 on the roadmap. The entire roadmap outlines a modular development approach from generation 1 to 3.

This generational framing de-risks the journey by sequencing capability, clarifying dependencies (data, models, routing logic, knowledge articles), and making investment decisions more tangible to stakeholders.

What made the difference?

The process was guided by fundamental principles that significantly de-risk AI investments:

• Validate assumptions early with real users and system constraints

• Prioritise speed-to-learning over speed-to-building to avoid costly surprises

• Treat stakeholder alignment and ownership as explicit deliverables

• Sequence value with a Gen 1–3 roadmap; link dependencies, KPIs, and decision gates

• Build a minimal but durable operating model (content, models, governance) from day one

From concept to investable AI

Grundfos’ innovation sprint demonstrates how to successfully translate AI ambition into a credible, low-risk plan: stress-test assumptions, validate with real users, and structure the path from assisted intake to automation. The result is an investable roadmap that sequences value and builds internal readiness – a repeatable way to de-risk, decide, and deliver.