Jeremy Willmott, Senior Director, Digital Strategy Group APAC, Adobe
Most companies have AI initiatives running separately across different teams and functions. The adoption journey is similar, with pilots launching, tools going live and roadmaps filling up. But, often momentum slows after that, and people start asking why progress has stalled.
We set out to find an answer. Working with Incisiv, we surveyed 461 senior executives from a cross-section of industries in Southeast Asia, all with authority over AI investment. We also went deeper on adoption opportunities and challenges through in-depth interviews.
We expected the barriers to broader AI adoption to centre on technology, and depend on which models are being used, how much spend is allocated, or how early companies begin their AI adoption journey. Instead, the strongest predictor of whether AI scaled is organisational readiness.
After all, two companies can run the same models, spend the same money, and hire similar teams, yet our research suggests that one can compound its return on investment while the other stalls. The difference comes down to the operating model, including who owns AI, how it's measured, how business and IT work together, and how the work itself gets redesigned.
Where AI deployment halts
Among the surveyed organisations in Southeast Asia, approximately 57% are stuck in the Frozen Middle, meaning that they're piloting or deploying AI, but the coordination needed to scale is limited. For agentic AI specifically, that number is 40%.
One hurdle is the time and commitment it takes to reach scaled deployment. Even moving from experimentation to initial deployment takes at least six months for six in ten organisations. For almost one third, that time horizon is 12 months or more.
These longer timeframes can hinder momentum, which also helps explain why 48% of organisations are still working on their first AI use case or have stopped altogether after an unsuccessful attempt.
However, for most, tools are in production and early results exist, but there's no enterprise operating model to support orderly and impactful adoption. It means functions cross the finish line one at a time instead of together and new pilots launch while old ones plateau. One chief strategy officer put it plainly, saying that when he finally counted the number of AI launches that had actually changed how the business ran, "the number was bad."
Only 6.3% of Southeast Asian companies, representing 27 organisations in our dataset, have both scaled AI to multiple workflows and built the operating model to sustain it. We call them Leaders, and they tend to convert pilots to production at nearly four times the rate of other companies. Conversely, approximately 79% are classified as nascent, early in both AI deployment and operating model maturity, the highest in the Asia Pacific region.
With 73% of companies lifting AI investment year-on-year, and almost 60% believing they are still underinvesting, the organisational settings that help turn investment into adoption and measurable value come into sharper focus. Here, the traits of Leaders who are already achieving both can be instructive.
Structural shifts are the real work
The breakdown of operating model maturity across key dimensions is similar in every industry. The research finds that organisations tend to rank more highly for the strategy and governance dimensions, while structure and workflows lag.
That's not an accident, given a decision-maker can approve a policy or sign off on a strategy in a meeting, whereas changing roles, moving decision rights, and reworking how teams operate takes much longer. That’s one reason why companies keep putting it off.
Meanwhile, Southeast Asian Leaders have higher than average scores for the structure dimensions, defined as dedicated ownership and accountability, and workflows, which indicates how deeply AI is embedded in processes. Leaders’ maturity in these areas outstripped the regional average by 67% and 59%, respectively.
When we asked executives what they'd do differently, the answers all came down to decisions they wished they’d made upfront. This includes setting clearer metrics, securing budget sooner, naming an owner earlier, bringing IT in earlier, and investing more in change management. In these cases, the foundational calls that should have come first were made after deployment, by which point the opportunity to establish a baseline had already passed.
Building pathways from experimentation to production
Of every 100 AI ideas proposed by Southeast Asian organisations, about 15% make it to production. Approximately 36% of proposals don’t make it past IT evaluation because business teams often lack the architectural and data capabilities needed to scope implementation effectively. Another 18% of proposals are rejected during security review and the rate nearly doubles for highly regulated industries such as financial services, healthcare, and life sciences.
Idea generation isn’t the problem. What we found was that companies often lack pre-approved pathways that help move initiatives into production, including approved integration patterns, standardised security reviews, and established budget categories. Without these mechanisms, proposals must navigate a new or different process, significantly reducing their chances of success.
The same applies to employees, roles and skills, with the latter remaining the third most common barrier to scaling AI. This remains a widespread and stubborn challenge, with 81% of businesses saying the skills shortfall remained unchanged or worsened over the past 12 months. Compounding the issue, more than half of organisations are expanding roles to include AI and 21% are adding new skill requirements.
Five things Leaders do differently
- Name a real owner. Compared to the global average of 20%, only 12% of Southeast Asian companies have a designated AI owner with budget authority and accountability for outcomes. This compares to 89% among Leaders.
- Measure from day one. Only 8% of companies measure AI return on investment comprehensively. Most just track usage rather than quality, rework, or the real cost of reviewing AI-generated output.
- Put business and IT in the same room, early. Of the surveyed companies, 78% still run AI like a ticket system where the business team submits requests and IT delivers solutions. However, Leaders build the model together and align on what success looks like even before the project brief is written.
- Treat failure as data, not blame. Over a third of companies repeat the same AI mistakes because no one identifies and flags the patterns. Leaders keep their destructive-failure rate under 1% by creating an environment where failures can be discussed and used as learning opportunities.
- Identify your champions and empower them. Among Leaders, 41% consider this as their single most effective move, and it's also the least expensive. Find the people already making AI work in real-world workflows. Give them recognition, a modest budget to extend what they have built, and the autonomy to scale their work to other teams.
The gap compounds.
Global Leaders pull 1.4 to 1.6 times more value out of the same technology as everyone else. The models are the same, but the difference lies in how organisations are built around them.
Ownership drives measurement. Measurement gives business and IT teams shared evidence. This shared evidence drives a collaborative partnership, which makes failure safe to examine. Examined failures surface champions. Champions strengthen the case for more ownership. If these measures are not incorporated, organisations in the Frozen Middle will fall further behind with every cycle.
Before you launch another AI idea, do these five things
- Name one owner
- Set a baseline before launch
- Put business and IT in the room before the brief is written
- Treat failure as evidence, not a career risk
- Find and fund the people already making it work
Read Adobe’s 2026 State of AI Readiness report to see how organisations compare globally and what the Leaders do differently.
Recommended for you
https://business.adobe.com/fragments/resources/cards/thank-you-collections/generative-ai
