Tencent Cloud used AI agents for XLSMART migration

Tencent Cloud claims zero downtime moving 1,200 microservices for XLSMART In sum – what we know: - A fast, clean cutover – Tencent Cloud moved over 60 core apps, 1,200 microservices, and 15TB of data for XLSMART in 4.5 months with zero downtime. - AI did the heavy lifting – Proprietary tools like CodeBuddy, WorkBuddy, TokenHub, and DatabaseClaw handled discovery, architecture mapping, and live monitoring throughout the migration. - A reusable playbook – Tencent has packaged the 20-plus custom Skills into a productized migration platform tied to its Agent Development Platform 3.0 for other enterprise customers. Tencent Cloud has completed a large-scale public cloud migration for Indonesian telecom operator XLSMART in 4.5 months, with zero downtime, and it apparently did so by leaning heavily on AI agents at every stage of the process. Over 60 core applications, 1,200 microservices, 1,100 APIs, 900 business interfaces, and more than 15TB of core data assets all moved onto Tencent Cloud while XLSMART kept serving tens of millions of customers. Tencent used a portfolio of more than 20 agentic AI-powered cloud migration “skills” deployed across the entire lifecycle, from discovery through cutover and monitoring. That’s a meaningful departure from the traditional, script-based approach, where each phase relies on a mix of manual labor and one-off automation. Tencent frames the result as compressing tasks that would normally take months of manual work into weeks or days, through unified agentic pipelines rather than siloed tools. It’s a bold claim, and one that positions the company squarely against the US hyperscalers in Southeast Asia. Whether it holds up as a repeatable model or reads mostly as a flagship marketing win is the more interesting question. The XLSMART transition XLSMART is a new entity, formed from the 2025 merger of XL Axiata and Smartfren, and it now serves over 69 million users. Mergers of that size rarely produce clean IT estates, and this one was no exception. The combined operator inherited highly fragmented, heterogeneous platforms. The cloud transition initiative launched in December 2025 as part of a longer-term modernization strategy, with Tencent Cloud selected as the strategic technology partner. From there, the cutover ran 4.5 months. Throughout, XLSMART continued serving its customers without interruption, which for a telco of this size is arguably the whole point. Yessie D. Yosetya, XLSMART’s Director and Chief Information & Technology Officer, described the project as “more than a cloud migration,” but the creation of a stronger digital foundation that allows the company to innovate faster, improve operational resilience, and continuously enhance customer experience. What did the AI agents actually do? The migration ran on a set of proprietary Tencent tools. CodeBuddy handled coding tasks, WorkBuddy served as the AI agent workspace, TokenHub automated the migration pipeline, and DatabaseClaw covered agent-based database operations, migration, and optimization. These aren’t generic helpers bolted onto a manual process; Tencent describes them as running a single AI-driven pipeline spanning discovery, mapping, design, and deployment. The most concrete function is resource discovery. One skill automatically scanned cross-account environments and mapped hundreds of legacy resource specifications directly to Tencent Cloud configurations, generating deployable design blueprints in the process. This is the part of a migration that has traditionally meant engineers manually cataloging resources across multiple platforms, logged in spreadsheets — slow, error-prone, and easy to get wrong at scale. Replacing that with automated scanning is a genuine efficiency gain, assuming the mapping is accurate. A second skill inferred application architectures from the discovered assets, then produced Low-Level Design specification sheets and visual architecture diagrams. Anyone who has sat through the design-documentation phase of a migration knows how much engineer time it usually consumes, and undocumented legacy apps are the worst offenders. Having AI reconstruct that architecture and generate deployable diagrams on the fly is, on paper, one of the more compelling uses of agentic AI here. During the live cutover, AI agents supported system monitoring, identifying and addressing technical issues as they arose. Around the agents, Tencent deployed more than 20 core cloud products, covering security services, container services, databases, and intelligent analytics. It’s a lot of moving parts, and the zero-downtime result suggests they held together. To be clear, though, the public materials describe the monitoring and issue-handling only at a high level. How much of that “addressing issues autonomously” involved human engineers in the loop isn’t spelled out. Highly-skilled The 20-plus custom agentic Skills built for XLSMART haven’t been left as a one-off. Tencent has packaged them into a reusable AI migration platform aimed at other enterprise customers, effectively productizing the XLSMART playbook of AI-driven discovery, mapping, and migration. Poshu Yeung, SVP and Head of Tencent Cloud International, framed the project as a benchmark for large-scale AI-driven cloud migration, particularly for telecom and other complex industries, and confirmed the Skills have now been folded into that reusable platform. The migration platform ties into Tencent’s Agent Development Platform (ADP) 3.0 and its broader Intelligent Agent Strategy, which centers on scenario-based AI and autonomous agents embedded in enterprise workflows. ADP 3.0 is pitched at generating and integrating agents for cases like customer service, marketing, inventory management, and research, with XLSMART serving as an early, high-profile infrastructure example within that strategy. Strategically, the value of a case like this is that it demonstrates agentic AI operating in a mission-critical, high-scale telecom environment without a reported incident. That’s the pitch, at least. Set against AWS, Azure, and Google Cloud, all of which have long relied on more traditional, script-based migration approaches, Tencent gets a visible flagship reference to point to in emerging Southeast Asian markets. It’s a differentiator that plays well on cost and speed, which are exactly the levers Chinese providers tend to pull in the region. Indonesian expansion XLSMART isn’t Tencent Cloud’s only move in Indonesia. At MWC 2025, Tencent Cloud and Telkomsel signed an MoU to develop AI and cloud solutions together, covering AI-generated content and AI translation for B2B and B2C segments, Palm Verification eKYC for B2B digital identity, and cloud cost optimization. It’s an early-stage agreement rather than a delivered project, but it signals the same telco-focused push. Beyond the operators, Tencent Cloud has partnered with Mega Berjaya Teknologi to distribute migration services and AI features, including eKYC and Super-App capabilities, across domestic Indonesian businesses. That’s the distribution layer — a route to smaller enterprises rather than headline telco deals. Tthese moves accelerate the penetration of Chinese AI infrastructure and cloud expertise into Indonesia’s domestic digital economy. The XLSMART migration is the flagship, but it sits inside a wider effort to embed Tencent’s platforms across the country’s telecom and digital sectors — and, notably, in competition with US hyperscalers on their newer turf. Table of Contents