TeleAgent AI Agent Hits 1.5M Users: Workplace AI Shift
China Telecom's TeleAgent passed 1.5 million users in four months, with 99% of its 50,000+ marketplace skills built by end users. This article breaks down what that milestone means for your automation strategy.
The Short Answer
China Telecom's TeleAgent AI agent has crossed 1.5 million registered users, and it did it in about four months. Announced on October 9, 2026 at PT Expo China 2026, the workplace automation platform now counts more than 200,000 daily active users, a steep climb from just 27,000 registered users in mid-June. For business leaders watching…
China Telecom's TeleAgent AI agent has crossed 1.5 million registered users, and it did it in about four months. Announced on October 9, 2026 at PT Expo China 2026, the workplace automation platform now counts more than 200,000 daily active users, a steep climb from just 27,000 registered users in mid-June. For business leaders watching the AI space, this is not just another impressive growth chart. It is real evidence that AI agents are moving from clever demos into everyday production tools that complete work end to end.
The story matters for a simple reason. Most AI headlines celebrate what models can understand or generate. TeleAgent is being celebrated for what it can finish. That distinction, from understanding to doing, is reshaping how companies think about business process automation, and it carries practical lessons for any organisation planning its next automation investment.
The TeleAgent AI agent story: from 27,000 to 1.5 million users in four months
TeleAgent is built by China Telecom AI Technology Co., known as TeleAI, as a native AI application for the workplace. It was first unveiled at the World Artificial Intelligence Conference 2026 and serves a broad audience: individual professionals, small and medium-sized enterprises, and large state-owned conglomerates. The growth curve is what caught the industry's attention. Going from 27,000 registered users in mid-June to 1.5 million by early October represents roughly a fifty-five-fold increase in under four months, with over 200,000 people now using it every single day.
At the PT Expo China 2026 opening, Liu Guiqing, Executive Director, President and Chief Operating Officer of China Telecommunications Corporation, summed up the significance with a line that is already being quoted across the industry: TeleAgent marks the shift of AI from understanding what is said to getting things done. The company sees it as proof that autonomous AI agents capable of end-to-end task execution are moving from demos into large-scale enterprise adoption. (Source: China Telecom announcement)
The product itself is designed to remove friction. TeleAgent runs natively on Windows, macOS and leading Chinese enterprise operating systems including Kylin OS and UnionTech OS, with no complex server setup for individual users. A drag-and-drop workspace lets people drop files in and let the agent handle execution from start to finish. For larger customers, there are two distinct editions:
- Enterprise SaaS edition: supports multi-device collaboration and scenario customisation, so teams can tailor the agent to their own workflows.
- Private-deployment edition: released in mid-August, this keeps models and data entirely within the customer's own environment, addressing the data sovereignty concerns that stop many regulated industries from adopting cloud AI.
Behind TeleAgent sits TeleAI's Xingchen foundation model series, which the company says already powers more than 500 enterprise workflows across 20 industries. In September 2026, TeleAI also released Xing4.0-29B, an agentic large model designed to run on a single GPU, signalling that serious agent capability no longer requires data-centre scale hardware.
What 50,000 user-built skills reveal about the TeleAgent AI agent
The most interesting number in the announcement is not 1.5 million users. It is 99 percent. TeleAgent's skill marketplace has grown from roughly 40,000 to more than 50,000 skills, and 99 percent of those skills were built directly by end users solving their own productivity problems. Pause on that for a moment, because it inverts the traditional automation story.
In classic robotic process automation, the pattern was always top-down. The IT department or an outside consultant studied a process, wrote a bot, deployed it, and then maintained it. The bottleneck was developer hours, and the classic failure was the bot IT built that nobody actually used. TeleAgent flips this. The people who feel the pain of a repetitive task are the ones building the skill that eliminates it. The domain expert becomes the automation author.
Here is the original insight most coverage of this milestone is missing: a user-built skill marketplace is not just a feature, it is a measurement of workflow documentation. Every skill in that marketplace represents a process that someone understood well enough to teach an agent. For years, the hardest part of automation projects was not the technology. It was getting employees to describe, step by step, what they actually do all day. TeleAgent's drag-and-drop workspace and natural interaction model lowered the bar for that description far enough that 50,000 real workflows got captured. The platform did not just automate work. It harvested the undocumented knowledge that automation projects have always struggled to extract.
For businesses planning automation, the lesson is direct. Your organisation already contains thousands of small, repetitive workflows that no IT roadmap will ever get around to. A marketing coordinator's weekly reporting ritual, a procurement officer's supplier follow-up sequence, an HR assistant's onboarding checklist. Individually, none of them justifies a development project. Together, they are the long tail of productivity, and user-driven agent platforms are the first technology genuinely shaped to capture that long tail. If you are mapping out where automation could help, start by asking your teams a different question than consultants usually ask. Not "what is your biggest process problem" but "what is the annoying thing you do every Tuesday." That is where your 50,000 skills live.
From chatbots to doers: why agentic AI changes automation strategy
The TeleAgent milestone lands inside a much larger shift. Gartner forecasts that 40 percent of enterprise applications will embed task-specific AI agents by the end of 2026, up from under 5 percent in 2025. McKinsey reports that 88 percent of organisations now use AI in at least one business function, though only 23 percent are scaling an agentic system. The experimentation phase is nearly universal. The scaling phase is where the winners are being decided.
The difference between a chatbot and an agent is the difference between advice and action. A chatbot drafts your email. An agent drafts it, attaches the right files, checks the recipient list against your CRM, sends it, and logs the follow-up. That end-to-end completion is what Liu Guiqing meant by getting things done, and it is why agents demand a different evaluation mindset. You do not judge an agent by the quality of its answers. You judge it by the reliability of its outcomes.
There is also a caution in the data. Gartner projects that over 40 percent of agentic AI projects will be cancelled by the end of 2027, with unclear business value and inadequate data quality as the leading causes. The interest is real, but the execution gap is the story now. TeleAgent's success hints at why some deployments survive that gap: it started with individual users solving their own problems, then grew into enterprise editions. Adoption pulled the technology in, rather than a strategy deck pushing it out. Companies working with experienced AI and machine learning partners can apply the same logic: seed agents where the pain is felt, prove daily value, then scale.
Five practical takeaways for businesses automating workflows
What should a UK, US or Pakistani business actually do with this news? Here are five grounded takeaways drawn from the TeleAgent playbook and the broader 2026 agent landscape.
- Start from your own long tail. Before buying any platform, run a two-week exercise where every team lists the tasks they repeat weekly. You will be surprised how many small automations surface. A guide like total workflow optimisation can help you structure that audit before you spend anything.
- Insist on end-to-end execution. When vendors demo AI tools, ask them to complete a real multi-step task live, not just generate text about it. The agent that can finish a task reliably at 2am without supervision is worth more than the one with the cleverest answers. This is one of the clearest signs your business is ready to automate with AI: you know which outcomes you want, not just which tasks annoy you.
- Plan for both cloud and private deployment. TeleAgent's two-edition strategy is a template. Regulated data, client confidentiality, or national data rules may require models and data to stay on your premises. Choose platforms and partners that can work both ways, so a compliance requirement never forces you to abandon your automation.
- Measure daily active use, not licences bought. Track how many of your people actually use the agent every day. If that number is flat, the deployment is decoration, no matter how many seats were bought.
- Treat user-built automations as an asset to govern. When employees start building their own skills and workflows, as TeleAgent's 99 percent figure shows they will, you need light governance: naming standards, a shared library, and review of anything that touches customer data or money. Low-code and no-code agent workflows make this easy to start, and no-code and low-code AI automation workflows are a practical entry point for growing teams that lack a large engineering bench.
How to evaluate a workplace AI agent before rolling it out
With dozens of agent platforms now competing for enterprise budgets, a structured evaluation saves expensive mistakes. The table below distils the criteria that the TeleAgent story suggests actually matter, alongside what the old chatbot era taught us to check.
| Evaluation criterion | What to ask | Why it matters |
|---|---|---|
| Task completion reliability | What percentage of multi-step tasks finish without human rescue? | An agent is judged on outcomes, not answers. |
| Deployment flexibility | Is private or on-premises deployment available? | Data residency rules can block cloud-only tools. |
| Platform coverage | Does it run natively where your team works? | Agents that need server setup die in pilot. |
| User-built extensibility | Can non-developers create and share skills? | This is where the long tail of automation lives. |
| Governance and audit | Can you see what the agent did and roll it back? | Autonomy without audit trails is a liability. |
| Integration depth | Does it connect to your CRM, ERP and file systems? | Agents are only as useful as the systems they can touch. |
Run a pilot with one team and one painful workflow, not a company-wide rollout. Give it thirty days and measure three things: daily active usage, time saved per completed task, and the number of user-built skills or workflows that appear without IT involvement. If all three trend upward, you have found something worth scaling. If only licences were consumed, you have learned something cheap.
Frequently asked questions
What is China Telecom's TeleAgent?
TeleAgent is a workplace AI agent developed by China Telecom AI Technology Co. (TeleAI). It is a native AI application that executes office tasks end to end, from handling files through a drag-and-drop workspace to running multi-step workflows. It serves individual professionals, SMEs and large enterprises, with both a cloud SaaS edition and a private-deployment edition.
How fast did TeleAgent grow to 1.5 million users?
TeleAgent grew from about 27,000 registered users in mid-June 2026 to more than 1.5 million registered users by October 9, 2026, roughly a fifty-five-fold increase in under four months. It also passed 200,000 daily active users, which indicates strong habitual use rather than one-time signups.
What is the TeleAgent skill marketplace?
It is a library of more than 50,000 reusable automation skills for the TeleAgent platform, up from around 40,000 earlier. Remarkably, 99 percent of these skills were built by end users solving their own productivity problems, making it one of the largest user-generated automation ecosystems in the industry.
What is the difference between the enterprise SaaS and private-deployment editions?
The enterprise SaaS edition supports multi-device collaboration and scenario customisation for teams. The private-deployment edition, released in mid-August 2026, keeps the AI models and all data entirely within the customer's own environment, which suits regulated industries and organisations with strict data sovereignty requirements.
Can small businesses benefit from AI agents like TeleAgent?
Yes. The core lesson of TeleAgent is that the biggest automation gains often come from the long tail of small, repetitive tasks that individual employees know best. Small businesses can start with low-code or no-code agent workflows targeting one painful weekly task, measure daily usage and time saved, and expand from there.
What should companies watch out for when adopting AI agents?
The main risks are unclear business value, poor data quality, and lack of governance over what agents can do. Gartner projects that over 40 percent of agentic AI projects could be cancelled by the end of 2027. Piloting with one team, measuring daily active use, and requiring audit trails for agent actions are practical ways to stay on the safe side.
The shift from AI that understands to AI that gets things done is no longer a forecast. With 1.5 million users, 200,000 daily actives and 50,000 mostly user-built skills, TeleAgent is the clearest large-scale proof yet that workplace AI agents have crossed into production reality. The businesses that win this wave will not be the ones with the biggest models. They will be the ones that capture their teams' everyday workflows, govern them lightly, and let the people who feel the pain build the automation. If you are ready to explore what agentic automation could do for your organisation, talk to our team and we will help you map your first high-impact workflows.
Related Articles
Keep reading — more guides you may find useful.
Related Services
Services connected to this guide.
