Are you stuck in AI's 'frozen middle'?

Southeast Asian companies are ramping up their artificial intelligence (AI) investments, but more than half remain stuck between experimenting with the technology and scaling it across their businesses, according to Adobe.

The "Adobe 2026 State of AI Readiness Report" found that 57% of organisations in Southeast Asia are caught in what it calls the “frozen middle”, where companies are piloting or deploying AI without the coordination needed to scale it. Only 6.3% qualify as AI Leaders, having scaled AI across multiple workflows and established the operating models needed to sustain adoption.

The findings are based on a survey of 461 senior executives across Southeast Asia with authority over AI investment, as part of global research conducted by Incisiv on behalf of Adobe. The wider study surveyed 5,633 senior executives across 12 industries and 11 markets, with respondents representing organisations generating at least US$100 million in annual revenue.

Don't miss: B2A, AEO, GEO and beyond: your guide to marketing’s new AI language 

Despite 73% of companies in Southeast Asia increasing their AI investment year on year, 48% are still working on their first AI use case or have stopped after an unsuccessful attempt. Almost 60% also believe they are underinvesting in AI.

The gap between experimentation and deployment can be lengthy. Six in 10 organisations take at least six months to move from AI experimentation to initial deployment, while almost one-third take 12 months or longer.

Where AI initiatives stall

A key barrier is the disconnect between business teams and IT. According to Adobe, 78% of surveyed organisations still approach AI through a ticketing model, in which business teams submit requests and IT delivers solutions.

This can make it harder to align technical capabilities with business needs before projects begin. Adobe found that approximately 36% of AI proposals do not make it past IT evaluation, while a further 18% are rejected at security review. The rejection rate at security review nearly doubles in highly regulated industries such as financial services, healthcare and life sciences.

Unclear ownership is another hurdle. Only 12% of companies in Southeast Asia have a designated AI owner with budget authority and accountability for outcomes, compared with 89% of AI Leaders.

Measurement is also lagging behind adoption. Just 8% of organisations comprehensively measure AI return on investment (ROI), with most tracking usage rather than output quality, rework or the cost of reviewing AI-generated work.

Meanwhile, 81% of businesses said their AI skills shortfall had remained the same or worsened over the past year. More than half are expanding existing roles to include AI, while 21% are introducing new skill requirements.

Furthermore, Adobe's research suggests that organisations making greater headway with AI are distinguished by how they structure their operations, rather than by how much they spend on technology.

The findings echo a wider gap between AI adoption and business returns across Southeast Asia. A McKinsey & Company report reported by MARKETING-INTERACTIVE in April found that nearly half of organisations in the region had moved beyond AI pilots, but around 60% reported less than a 5% impact on earnings before interest and taxes (EBIT), while nearly one in five reported no financial impact at all.

The report also found that only around one in five companies were scaling agentic AI in sales, marketing and customer engagement, despite nearly nine in 10 organisations planning to experiment with AI agents in the following year. Adobe's findings point to a similar challenge, with unclear ownership, fragmented collaboration between business and IT, and limited ROI measurement holding companies back from turning AI adoption into sustained business value.

Related articles:  
AI momentum builds across Southeast Asia, but gaps persist  
Only 9% of AI brand experiences deliver true consumer value  
90% of IT leaders report gaps in readiness for AI-driven cyber threats