The roundtable opened a clear challenge. AI initiatives frequently remain isolated, underfunded, and disconnected from strategy. Closing that gap requires more than technical skill. It requires cross-functional sponsorship, governance that scales with risk, and a direct line to business outcomes: P&L, customer experience, operational efficiency.

The practitioners also pushed back on the assumption that bold, visible AI deployments are the most valuable. Back-office models and embedded ML often deliver better returns, faster, with lower risk than generative AI pilots designed to impress. The lesson: start with the problem, not the technology.

Cases that held up under scrutiny

Sofie Perslow's ML-driven campaign optimisation across 1,300 grocery stores produced millions in incremental profitable sales within six months — while reducing cannibalization and building genuine confidence in data-driven decisions across the organisation.

Anthony Belpaire's contact centre transformation began as an FAQ bot and evolved into a context-aware conversational assistant integrated with CRM. Stepwise deployment, clear guardrails, and explainability at every stage reduced call volumes and raised customer satisfaction – critical in a regulated industry where trust is non-negotiable.

Łukasz Samborski mapped a fragmented AI project landscape into a coordinated portfolio, identified gaps in technology, data, and skills, and introduced an automation-first policy alongside democratised low-code tools. His core lesson: scaling AI without reskilling the organisation is not scaling at all.

Silvio Giorgio's computer vision deployment across hundreds of stores – thousands of cameras, processing in real time – tackled loss prevention and customer experience simultaneously. The decisive design choice was human-centric: soft prompts rather than public alerts preserved customer dignity and drove compliance. Finance-linked measurement made the EBIT contribution visible and defensible.

What leaders should do differently

The roundtable converged on five practical principles.

• Make trust operational. Benchmarking AI against human or legacy baselines, logging out-of-tolerance events, and routing sensitive decisions to humans are not optional extras. They are the foundation of long-term adoption.
• Measure what matters to the business. AI investments that cannot be connected to revenue, cost, or experience metrics will not survive the next budget cycle.
• Invest in people, not just platforms. Mandatory training, adoption incentives, and broad upskilling prevent the organisation from splitting into AI-fluent and AI-left-behind camps.
• Govern by risk, not by instinct. Three-lines-of-defence structures, EU AI Act alignment, and differentiated controls for high-risk and low-risk use cases create conditions for responsible scale.
• Use traditional and generative AI deliberately. Classical ML remains the workhorse for structured, high-volume decisions. Generative AI is powerful for unstructured tasks and interfaces — but only with the guardrails to match.

The bottom line

AI ambition is widespread. The capacity to convert it into business value is not. Organisations that treat trust, governance, and workforce enablement as foundational — not as afterthoughts — are the ones that will scale.

The technology is ready. The leadership question is whether organizations are.

---------

Learn more about Amrop's Global Digital Practice.