Across Southeast Asia, enterprises are moving more customer journeys into digital channels while adding AI-enabled features to services that were already expected to perform reliably across varied markets. Banking apps, government portals, retail platforms, and telecom self-care tools now have to work not only in controlled test environments, but also on older devices, congested mobile networks, and inconsistent connectivity conditions. As AI makes services more dynamic and personalized, the gap between internal performance metrics and actual user experience is becoming harder to ignore.

This issue is especially visible in Southeast Asia, where market conditions differ widely from one country to another. Singapore’s high-quality connectivity and device base can give enterprises a clean view of performance, but the same assumptions may not hold in Indonesia, the Philippines, Vietnam, or Thailand. In these markets, device fragmentation, routing inconsistency, 4G congestion, cross-border latency, and uneven infrastructure can shape whether a digital service feels reliable to the user.

For Gargi Dasgupta, Chief Technology Officer of Mozark, the challenge is that many organizations still measure digital services largely from the inside out. Cloud dashboards, application performance monitoring tools, and internal telemetry can show that systems are available, APIs are responding, and infrastructure is healthy. Yet those signals may miss whether a biometric login, document upload, recommendation engine, or AI-assisted journey actually completes on the user’s device under real-world conditions.

In this TNGlobal Q&A, Dasgupta discusses why AI is changing the assumptions behind software testing, where traditional observability falls short, and how enterprises in sectors such as banking, telecom, retail, and government can rethink digital experience validation. She also explains why real-world telemetry is becoming more relevant to infrastructure, compliance, and governance as regulators and users place greater emphasis on service quality, resilience, and accountability.

Across Southeast Asia, enterprises are rolling out AI-enabled digital services faster than before. From your perspective, what new kinds of performance blind spots does that create, especially in markets where device quality, network conditions, and user environments vary widely?

Speed is the new normal. AI has reduced shipping cycles from quarters to weeks, and most enterprises are adopting the change and are loving the velocity. Although validation hasn’t caught up to the speed of development.

The reason for lagging in validation is that AI-driven services don’t behave the same way twice. They personalize content, swap payloads, and increasingly act on the client side. The backend service dashboards might still show green even though users on a mid-range Android in Pahang stare at a blank screen. Or a recommendation engine that excels in Singapore times out on congested 4G in Cebu.

The gap between what we ship and what users actually get is widening fast as the media payload and JavaScripts get bigger, and agentic workloads need stable, high-bandwidth connectivity. The blind spots are growing and causing real issues for users.

Many organizations already rely on observability tools, cloud dashboards, and internal telemetry. Where do those tools still fall short when the goal is to understand what end users are experiencing in the real world?

Observability tools track what’s happening inside the perimeter, and they do a great at the jobs they are built for, like APM, log aggregators, and cloud dashboards. They can also indicate internal application health like CPU usage, Error rates, P99 latency, Container uptime, etc. It truly is great at tracking and observing the health of the application from the inside out.

All that telemetry stops at the network edge. It doesn’t tell you what render time looks like on a three-year-old Android. It can’t see a biometric flow failing under packet loss. When an API returns 200, but the UI never updates, no dashboard catches it. So you may have 99.95 percent availability on paper, but a customer in a heavily congested network in the middle of the day in Manila might be seeing a timeout. We are missing the outside-in view of the observability.

You can’t fix this with better APM. You need telemetry from the device itself, under representative network conditions, in the geographies you actually serve. That’s a different category of measurement, and most app providers have not realized it yet.

Southeast Asia is not a uniform operating environment. When you look across markets such as Singapore, Indonesia, the Philippines, Vietnam, and Thailand, what are the biggest differences in digital experience conditions that enterprises tend to underestimate?

One might consider Singapore as the representative of the SEA region. However, it differs vastly from the rest of the region. Singapore is a premium, miles ahead of the rest — premium devices, fiber everywhere, sub-20ms latency.

Indonesia and the Philippines sit at the other end. Mid- and low-tier Android phones dominate with older chipsets, constrained memory, and inconsistent OS updates. Services tested in clean conditions degrade noticeably the moment they hit real-world devices.

Vietnam and Thailand are somewhere in between. Advancing quickly on 4G and selective 5G, but local routing inconsistencies and patchy connectivity in tier-2 and tier-3 cities create real variance. Indonesia’s archipelagic geography makes it even harder

Enterprises need to test and cover use cases built around the device and network mix in each specific market. Otherwise, you’re shipping blind across most of your user base. The enterprises miss this hard truth; testing based on the demographics and in their geographic location is the key to customer satisfaction.

In sectors such as banking, telecom, government, and retail, what kinds of service failures are most likely to pass internal checks but still break down for users on everyday devices or weaker network conditions?

These failures share a signature. They don’t show up as failures in the pristine in-house test conditions.

Take an example of Banking, biometric authentication is sensitive to latency variance, so the session times out on the client while the auth service marks it successful. Step-up flows have compounding problems in high latency and lower end devices causing friction with UI freeze or app crashes.

In retail, image-heavy carousels and AI-driven recommendations just don’t render on lower-end devices. Users see blank states and leave. On the analytics side, it looks like a normal drop-off.

Government portals are notorious for eKYC and document upload flows that die on unstable cellular. The user gives up before the request reaches the server, so it never even hits the error logs. Telecom self-care apps have similar issues with OTP delivery and service activation.

These are silent killers. They erode trust, inflate support costs, slows digital adoption.

As AI becomes more embedded in customer-facing services, what changes in the testing and validation process itself? Do enterprises need to think differently about digital assurance when the service layer is becoming more dynamic, personalized, or agent-driven?

QA used to assume determinism. Same input, same output, same code path. AI breaks that assumption. A few things have to change.

First, the way we write test cases. Deterministic assertions don’t work for personalized, agent-driven flows. We need to validate intent and behavior. Did the agent take reasonable action? Did the journey complete? Was the inference quality acceptable?

Second, coverage. Dynamic services behave differently across device-network combinations, so testing one or two reference configurations isn’t enough. We need thousands.

The bigger shift is when we test. Pre-release validation isn’t sufficient when model drift and content drift are continuous. Assurance has to run in production continuously.

In my opinion, the hardest shift is organizational. Quality stops being a gate owned by QA. It becomes a closed-loop discipline shared across engineering, SRE, and product.

Regulators in different markets are paying closer attention to digital service quality and accountability. How do you see that changing enterprise priorities, particularly for organizations operating across multiple jurisdictions?

Regulators are pushing past data privacy. Financial authorities are now looking at service availability, transaction completion rates, and resilience. Telecom regulators want better coverage and quality reporting. Accessibility mandates are getting tighter in several markets.

What’s common across all of them is the move from policy adherence to evidence-based oversight. They want proof of consistent user experience at a local level, not regional averages that hide poor service in specific markets. For organizations operating across multiple jurisdictions, that is complicated. They now need to adhere to different benchmarks, different audit cadences, and different reporting formats per country.

The practical effect is that digital experience validation is shifting from being an engineering concern to governance. Enterprises that build auditable, localized evidence of real-user performance into their operating model will move faster than those that are still performing quarterly compliance exercises.

There is often a tendency to treat digital experience as an application issue, while infrastructure teams focus on uptime and performance at the system level. Do you think digital experience validation is starting to become an infrastructure concern in its own right? Why?

It already is, and it’s overdue. The old separation between application teams and infrastructure teams doesn’t hold up anymore.

AI services are sensitive in ways traditional applications aren’t. When any of the areas, like network conditions, edge proximity, device capability, and inference latency, degrade, the user blames the brand. Not the network provider. Not the CDN. So infrastructure teams can no longer measure themselves only by server uptime and CPU utilization. They own the substrate that determines whether the service actually delivers, and they need to own the outcome measurement that follows.

What this looks like in practice is real-world experience telemetry becoming a peer signal to system telemetry in the SRE workflow. Capacity decisions, edge placement, and CDN routing are all evaluated against measured user experience.

Looking ahead, what would a more mature approach to digital experience assurance look like for Southeast Asian enterprises over the next two to three years? What are the capabilities or habits that will separate the organizations that adapt well from those that continue to work reactively?

Reactive organizations chase incidents. Mature ones see them coming.

The capabilities that separate the two are not complex. First, continuous synthetic and real-user testing across the actual device-network footprint, integrated directly into CI/CD. Not as a release gate only, but as an always-on signal.

Second, an explicit process to test on the edge, on low-end devices, in a congested cellular network, with cross-border latency, having intermittent connectivity. These aren’t corner cases in Southeast Asia, but are everyday conditions for most of the user base.

Third, closing the loop. Real-world telemetry has to feed back into roadmap and architecture decisions of feature shipment, market prioritization, and edge infrastructure choice.

The teams that internalize these habits will protect their AI investments and scale across the region.

Mozark recently raised new funding as demand for digital experience testing grows. What does that tell you about how enterprises and investors now view this category, and where do you think the market is headed from here?

Our recent funding signals a shift in market awareness. Enterprises and investors are starting to understand that real-world digital experience validation isn’t a peripheral testing function. It’s core to any AI-driven business.

Code generation, deployment, and AI integration are not bottlenecks anymore. These are made easier now with AI native SDLC. The hard part is delivering services reliably to real users across fragmented devices, networks, and geographies. That’s where the next layer of competitive advantage will be built.

What I see ahead is the convergence of development, infrastructure, assurance, and compliance teams all working from a shared view of real-world performance. The category will consolidate around platforms that operate at scale across geographies, with the depth to support regulated industries. Southeast Asia, given its diversity and pace of AI adoption, will be one of the leading proving grounds.

Amogy CEO Seonghoon Woo on the path from clean energy pilots to deployment [Q&A]