Many enterprises are now exploring Agent‑based intelligent agents to revamp internal workflows. Today we will break down the case of Jiangsu Tower’s “Taqing Digital‑Worker Agent” and analyze how it was implemented.
The Legacy Challenge: Site‑lease Reimbursement, a Long‑standing Headache for Front‑line Teams
Site‑lease reimbursement may sound straightforward: receive supplier contracts, perform cross‑checks, enter data into systems, complete approval workflows, and release payments. Yet for Jiangsu Branch of China Tower, this process has long consumed massive front‑line manpower.
Why is it so cumbersome? Behind each reimbursement lie multiple interconnected tasks: document sorting, multi‑system data entry, cross‑department workflow coordination, and constant compliance with risk‑control rules — where every step must be error‑free.
What complicates matters further is that know‑how and expertise are “people‑dependent”.
“In the past, this work relied entirely on senior staff mentoring new hires. Staff turnover would immediately drag down operational efficiency,” recalled a regional manager at Jiangsu Tower. Seasoned employees master workflows, are familiar with common pitfalls, and know how to resolve exceptional cases — but such expertise exists only in their heads. New team members have to start from scratch, resulting in volatile efficiency and inconsistent error rates.
The Nanjing subsidiary exemplifies this pain point. A seven‑person team was bogged down in repetitive manual data entry, verification and workflow follow‑ups, leaving little bandwidth for higher‑value work.
What the Agent Solved: Transforming Human Expertise into System‑embedded Capabilities
Jiangsu Tower adopted a different approach: instead of forcing employees to adapt to rigid workflows, they enabled systems to replicate human‑like judgement.
Taking the Nanjing subsidiary as a pilot, the enterprise built the AI digital employee dubbed “Taqing Digital‑Worker”. Its implementation consists of three core pillars:
First, Jiangsu Tower carried out granular digital decomposition of the full site‑lease reimbursement workflow. It translated senior employees’ judgement criteria, operational steps and hands‑on experience into a library of role‑specific rules interpretable and executable by the digital worker. Expertise is therefore no longer tied to individual staff, but solidified within the system.
Second, conventional RPA robots only execute hard‑coded scripts and fail whenever page formats change. By contrast, “Taqing Digital‑Worker” is equipped with an AI brain. It can comprehend tasks, parse contracts and invoices, extract key information, and make autonomous judgements following predefined rules.
Crucially, it is granted permissions and tool access to directly operate business systems — with no need for interface revamps on legacy systems, greatly lowering roll‑out barriers.
Third, tangible efficiency gains are reflected in hard metrics: over 10,000 working‑hours saved annually, with a stable business processing accuracy rate above 99%.
The original seven‑person team at Nanjing Tower has been downsized to two supporting staff members. Instead of drowning in manual data entry, these employees handle offline edge cases, exception resolution and review work that remain beyond the digital worker’s capacity. The operating model has shifted from “human‑led, machine‑assisted” to “machine‑led, human‑supported with exception fallback”.
Three Most‑asked Business Questions
Q1: Is it secure for digital employees to operate business systems directly? Are data risks a concern?
This tops every enterprise’s list of concerns. Jiangsu Tower embedded security as a core constraint from the project’s inception, building three‑layer safeguards:
- Technical layer: Private intranet deployment, with zero data exfiltration outside the enterprise security perimeter. All operations run on internal networks without exposure to public internet.
- Governance layer: Granular permission control and tool whitelisting. Strict authorization governs which data and business systems the digital worker may access. Permissions are well‑scoped and all operations follow predefined rules.
- Oversight layer: Full audit trails for every action, with mandatory human review for high‑risk links. Humans retain final oversight for high‑risk items, establishing a hybrid model of autonomous execution backed by human supervision and fallback.
Q2: Do we need system overhauls or custom interface development? Will implementation be labour‑intensive?
No system‑wide revamp is required. This is a key design advantage of “Taqing Digital‑Worker”. Rather than developing dedicated interfaces, it operates existing systems at the operating‑system level — simulating human behaviour by reading screens, simulating mouse clicks and keyboard inputs.
Enterprises avoid heavy‑cost system reconstruction and lengthy IT backlog waits for interface development, substantially lowering deployment barriers.
Q3: Can Agents deliver similar value for smaller companies outside the tower‑infrastructure sector?
“Taqing Digital‑Worker” is not limited to China Tower’s internal use cases. Jiangsu Tower has packaged this practice into a replicable digital‑worker framework, with planned expansions to other high‑volume standardized scenarios including contract management and electricity‑bill reimbursement.
If you are considering launching similar initiatives, here are three practical recommendations to get started:
- Jiangsu Tower avoided large‑scale enterprise‑wide AI roll‑outs from day one. Instead, it selected site‑lease reimbursement as a focused pilot. Likewise, identify one suitable scenario within your organisation — such as expense‑report auditing, contract‑clause comparison or report data entry — and validate end‑to‑end performance before scaling.
- The “Taqing Digital‑Worker” case demonstrates that large‑scale system overhauls, custom coding and lengthy IT projects are not prerequisites. Modern AI platforms support low‑code configuration, enabling non‑technical stakeholders to deploy functional digital workers within a short timeframe. Launch small‑scale pilots first, then iterate gradually.
- Security governance including data protection, permission management and activity auditing was built into initial project design, rather than patched on after incidents occur. If you kick off an Agent project, treat security governance as a mandatory requirement at project initiation, not a post‑launch add‑on.
The transformation from a seven‑person team to two human operators augmented by a digital worker illustrates that enterprise‑grade Agents are no longer science‑fiction concepts. They are already reshaping job roles within back‑office finance departments, and adoption barriers may be far lower than many organisations anticipate.