Study: 74% of companies face surprise AI costs as governance gaps widen

A new study from Sensedia, Squadra, and MIT Sloan Management Review Brazil says 74% of organizations are seeing unexpected AI costs, compliance risks, or redundancy problems as they try to move beyond experimentation. The findings suggest companies now need governance, data access, and systems integration more than more AI pilots.

Why it matters: - The study shows AI has moved from a pilot-stage conversation to an operations problem for most companies. - Unexpected costs and weak controls can slow AI rollout, raise compliance exposure, and limit business value. - The findings suggest competitive advantage will come from governance and integration, not just model adoption.

What happened: - Sensedia and Squadra, in partnership with MIT Sloan Management Review Brazil, released a global study on AI adoption and enterprise governance. - The research found 66.2% of executives rank AI as a top business priority, but only 7.4% of organizations have broadly integrated AI into core processes. - The study says 74% of organizations are dealing with unexpected costs, technical redundancies, or compliance risks tied to decentralized AI use.

The details: - The study draws on survey data from APIX 2026, interviews with enterprise leaders, and research from MIT Sloan, Bain & Company, Gartner, McKinsey, and Deloitte. - Researchers said productivity gains from AI do not automatically translate into growth. - Sustainable differentiation depends on embedding AI into core enterprise architecture. - 54.4% of technology professionals said corporate data access for AI applications is unstructured. - 41.4% said they operate without formal risk management processes. - Only 4.8% of executives said they have complete governance models backed by dedicated platforms. - 80% lack structured agentic architectures. - 68% have not deployed Model Context Protocol in production. - Kleber Bacili, co-founder and CEO of Sensedia, said AI adoption is no longer the main challenge and that intelligent agents now need a single governance layer connecting data, systems and models. - Lisa Arthur, Sensedia’s SVP North America & APAC and Global Chief Marketing Officer, said adoption without architecture creates a costly governance problem and that companies need integrated orchestration to scale safely.

Between the lines: - The report frames AI as an infrastructure issue, not just a software choice. - Companies that spread AI tools across teams without shared controls may create duplicate spending and harder-to-manage risk. - API management is emerging as a central layer for observability, security, traceability and cost control in agent-driven systems. - Marcilio Oliveira, co-founder of Sensedia, said APIs connect intelligence, processes and decisions in an agent-driven environment.

What's next: - Companies are expected to shift from isolated AI experiments toward governed architectures that connect models, APIs and enterprise data. - The report points to growing interest in agentic systems, but says most organizations still lack the architecture to support them at scale. - The study highlights examples from Bradesco, Inter, MAPFRE, Hcor and Klabin as early evidence of how governed AI deployments can be applied in finance, healthcare and industrial settings. - Bradesco has deployed more than 600 production AI applications and estimates $48 million in annual value. - Inter uses more than 500 models for credit, fraud and personalization decisions. - MAPFRE uses intelligent agents in customer service and claims journeys. - Hcor is advancing data interoperability and AI-based clinical support. - Klabin has adopted a multi-agent platform with centralized governance and unified control.

The bottom line: - The report’s core message is blunt: AI scale depends on governance, not enthusiasm. Companies that want measurable returns will need tighter control over data, systems and agent workflows.