Artificial intelligence can make individual tasks faster without necessarily making an organization more productive. In ASEAN, Workday research found that 86 percent of employees say AI has improved their day-to-day work experience, yet around one in five lose more than seven hours a week moving information and reconciling data across disconnected systems.
TNGlobal reported on the findings in August. The research covered professionals in Indonesia, Malaysia, Singapore and Thailand and was part of a wider global study of active AI users.
In this TNGlobal Q&A, Jess O’Reilly, General Manager, ASEAN at Workday, discusses why AI often exposes structural friction rather than creating it, how organizations should measure net productivity, and what changes when AI agents can take actions across enterprise systems.
Workday’s research presents an apparent contradiction: 86 percent of ASEAN employees say AI has improved their day-to-day work experience, yet one in five reportedly lose more than seven hours a week moving information and reconciling data. How should leaders distinguish productivity problems caused by AI itself from problems that AI is simply exposing in already fragmented processes and technology stacks?
These findings are not contradictory. AI can improve the experience of an individual task while also exposing the friction in how work gets done across an organization. The question for leaders is whether AI is creating that friction or revealing friction that was already there.
In most cases, AI is revealing the friction. For example, when HR, finance and IT operate on separate systems, data may be duplicated, outdated or inconsistent. Employees then spend time moving information between systems, reconciling different versions of the same data and making sure everything is accurate. That work existed before AI. However, what AI changes is the speed and scale at which organizations are trying to operate, which makes the inefficiency much harder to ignore.
AI can also create a productivity tax of its own, but that is a narrower problem and usually a deployment one. Where tools are rolled out without oversight or training, time saved generating an output is spent checking and correcting it.
The report says 66 percent of employees spend at least half their time translating and coordinating between systems and teams. What does that work typically look like in practice, and which types of handoffs or information gaps create the greatest drag on productivity?
A lot of this work comes down to employees acting as the connection between systems that were never designed to work together.
For example, HR may own people data, finance owns the numbers and IT manages technology and systems. When those functions operate on separate platforms, employees can spend significant time pulling information from different sources, reconciling discrepancies, updating records and making sure everyone is working from the same version of the data.
Workday’s research shows that more than eight in 10 employees spend significant time moving information between systems, while 84 percent spend significant time coordinating work across teams or business systems. Another 82 percent regularly spend significant time reconciling inconsistent information. The accumulation of each point of friction across the workflow means that every time an employee has to find information, translate it, reconcile it or confirm it with another team, time is taken away from higher-value work
You have described the loss of context between systems as a structural problem. What kinds of context are most often lost when work moves between AI tools and enterprise systems, such as permissions, business-process state, data provenance, definitions, or prior decisions, and why does preserving that context matter for the quality of AI outputs?
When AI tools and enterprise systems are disconnected, information can be separated from the processes and workflows that give it meaning. This matters because AI is probabilistic. Unlike traditional software, which is predictable and does the same thing every time, AI can be confidently wrong. In areas such as HR, finance and IT, where the margin for error is effectively zero, that distinction is particularly important.
An AI tool may be able to generate an answer or complete an individual task, but it may not have visibility into the broader business context, the latest information or what another part of the organization has already done.
That is one reason we see employees spending so much time reconciling information. Workday’s research found that 82 percent regularly spend significant time reconciling inconsistent information across multiple systems. When AI is embedded into the flow of work, rather than layered onto fragmented systems, employees can spend less time validating and moving information and more time acting on it
Only 30 percent of organizations in the research have embedded AI into the core of their business. What separates genuinely integrated AI from simply adding more AI tools to existing workflows, and what should enterprises prioritize before connecting AI more deeply to HR, finance, IT, or other sensitive operational systems?
The difference is whether AI is embedded within the systems and workflows where work actually happens or simply added as another tool alongside them. A standalone AI application may sit outside the system of record, meaning employees still need to move information between the AI tool and their core business systems, reconcile data and ensure that both systems are working from the same information. That can add another layer of complexity rather than removing it.
Before connecting AI more deeply into sensitive functions, organizations need to establish the right foundation: a clear understanding of how work happens, trusted business processes, defined handoffs and guardrails, as well as security, accuracy and audit trails.
The opportunity is to move away from standalone AI applications and toward AI that is embedded directly into core business platforms. That is how organizations can reduce the manual work around AI rather than simply adding another tool for employees to manage.
AI can make an individual task faster while creating additional work through checking outputs, reconciling conflicting information, or correcting errors downstream. How should organizations measure the net productivity impact of AI rather than relying on task-completion speed or adoption rates? Which metrics should be established before deployment?
Ultimately, organizations should look beyond task-completion speed and ask whether AI is enabling better-quality work and better outcomes, rather than simply measuring how much work gets completed.
This means looking at factors such as the amount of rework required, time spent reconciling information, manual handoffs and whether employees are actually spending more time on higher-value work. The quality of the output matters too. If AI produces something quickly but an employee then spends significant time checking and correcting it, the organization has not necessarily gained productivity.
The opportunity is to redesign work around what AI and people each do best. AI can take on more routine, repetitive and administrative work, while people focus on areas that require judgment, creativity, critical thinking, collaboration and decision-making.
As enterprises move from copilots toward autonomous or agentic systems, is there a risk that organizations recreate the same fragmentation problem with multiple agents acting across different applications? What interoperability, identity, access-control, observability, and governance capabilities become necessary when AI can take actions rather than simply generate recommendations?
If organizations simply add agents on top of fragmented systems, they risk recreating the same problem in a more complex form.
The shift from AI as an assistant to AI as an active participant in work makes context and governance even more important. AI agents can operate within the flow of work, but they need the right business context, trusted information and guardrails to do so effectively.
That means organizations need to think about how agents work together across the enterprise rather than evaluating each agent in isolation. Agents need to operate within defined permissions and enterprise controls.
Technology integration alone does not necessarily remove organizational friction. What changes in process ownership, employee roles, training, and decision rights are needed so that automation actually reduces administrative burden instead of shifting it elsewhere in the organization?
Technology integration removes friction only when leadership sets the direction and gives teams the authority to redesign the process beneath it. That means naming an owner accountable for the outcome rather than the system, and being explicit about which decisions are automated, which are flagged for review, and who answers when an output is wrong. Leaders also need to make adoption safe. Employees will only use these tools well if they are trained to question them and are not penalized for the learning curve. Reskilling is what makes oversight real rather than procedural, and that investment has to be visibly backed from the top.
The ASEAN findings draw on respondents in Indonesia, Malaysia, Singapore, and Thailand, with different sample sizes in each market. What conclusions can reasonably be made at the ASEAN level from the research, and did Workday observe any meaningful differences among markets, functions, or employee groups that help explain where AI-related friction is most acute?
At an ASEAN level, the research points to a broader shift in how work is being done. Employees are embracing AI and seeing benefits from it, but many organizations are still dealing with fragmented systems, disconnected data and gaps in the skills and frameworks needed to use AI well.
That said, ASEAN is not a single, uniform market. Digital maturity, workforce structures and operating environments differ considerably across the region. What the findings help us understand is AI adoption does not automatically translate into productivity. The larger issue is that AI is being introduced into organizations where information, workflows and systems remain fragmented. The next phase of adoption needs to move beyond adding more task-level tools and focus on integrating AI into the flow of work.
Editor’s note: This Q&A has been lightly edited for clarity and TNGlobal house style. The responses remain those of the interviewee.
Share your perspective: TNGlobal welcomes contributed insights and expert commentary from across Asia’s technology and innovation ecosystem. Submit a contribution for editorial consideration, or explore more conversations in our interview archive.