Productive Conflict Improves AI Decisions, Expert at MSP Show

At the MSP Show 2026 in London last week, service management consultant Alex Cosma told attendees that productive conflict and structured debate improve technology and AI decisions. She urged leaders to invite dissent rather than avoid it, warning that avoiding confrontation can hide flaws and lead to larger problems later.

Cosma described a pattern she calls “artificial harmony,” where teams mistake quiet agreement for success. In meetings where no one pushes back, problems can remain undiscovered until they cause operational issues. She pointed to examples such as processes that are routinely ignored, user portals that attract complaints without direct feedback, and projects that remain marked green despite growing operational problems.

On AI, Cosma warned that relying on machine outputs too early can short-circuit human debate. She told the audience: “AI shines in convergence. It makes us faster and more consistent in routine decisions. But humans shine through divergence.” She recommended starting with divergent thinking to explore different viewpoints and question assumptions, then using AI to narrow and refine options.

A key point in her session was to keep disagreement focused on ideas rather than people. “Never challenge the individual – challenge the idea,” she added. Cosma described a case where architects, process managers and service leaders spent a full day arguing a difficult service-management issue. The debate grew tense, she noted, but participants reached a result that improved on any single proposal.

Cosma highlighted the risk of groupthink and the tendency for teams to prefer social harmony. “If everyone agrees with me in a meeting, I get suspicious,” she observed. She urged leaders to make dissent an expected part of decision-making and to create psychologically safe environments where staff can raise objections.

Her practical recommendations included training leaders to invite challenge early and structuring disagreement so it reveals hidden issues and tests assumptions. Cosma said those steps can improve process design and help ensure AI tools operate on examined assumptions.