Tencent Maps' LBSwork Reuses WorkBuddy's Same-Source Harness Framework—Can It Replicate the Success?

WorkBuddy handles office work, CodeBuddy helps write code, DataBuddy handles data analysis, and MusicBuddy has run off to help musicians with creative work.

A few days ago, Tencent Maps released a desktop AI product called LBSwork, aimed mainly at industries such as retail, travel, and culture and tourism, offering capabilities including map development, data analysis, and business decision-making.

Although it isn't called MapBuddy, it uses the same Harness framework as WorkBuddy.

I went through the official LBSwork website on purpose. Looking at the familiar chat box, task list, knowledge base, experts, skills, and connectors in the product preview images, there is more than a hint of WorkBuddy about it. Don't believe me? Then take a look at the two product screenshots below.

LBSwork product interface👆🏻

WorkBuddy product interface👆🏻

It seems this WorkBuddy playbook has already started being adopted by other business lines at Tencent.

The question is, can WorkBuddy's successful experience really be replicated at scale?

01 Tencent has started using the WorkBuddy playbook everywhere

In March this year, WorkBuddy officially launched.

After several months of development, it has become one of the more representative Tencent AI products, especially in the desktop AI office Agent direction, where WorkBuddy's market performance is quite good.

According to data previously published by Analysys, in September 2026, WorkBuddy's desktop monthly visits reached 85.34 million, ranking first among the 20 AI-native office agents it tracks. This counts visits, not unique users.

With results like these, other business teams inside Tencent would probably be tempted too.

In the past, when developing an AI product, technical teams often had to handle model calls, task management, tool execution, permission control, and other issues themselves. Now that WorkBuddy has built up a set of Agent technical frameworks, other business teams have the opportunity to reuse them.

LBSwork is a new example.

Tencent Maps did not build a completely separate Agent framework from scratch, but instead adopted the same Harness as WorkBuddy.

With this kind of foundation, business teams can focus more energy on their own professional capabilities.

Tencent Maps plugs in map data, route planning, spatial analysis, and other capabilities; the music business plugs in music creation tools; and the data analysis business integrates databases and data processing capabilities.

The same technical approach can keep being used just by switching industries.

For a company with as many businesses as Tencent, this is indeed attractive.

With so many businesses under Tencent, if every team developed AI Agents from scratch, the duplicated investment would be easy to imagine. If technical frameworks can be reused, product R&D efficiency naturally has a chance to improve.

Readers often ask: since they are all based on similar technical frameworks, why does Tencent keep launching new standalone workbenches? Will every business department have its own Work in the future?

02 Can Tencent Maps also help bosses open stores?

First, let's look at what LBSwork can actually do.

The official website designed seven map experts for it, including map visualization, daily travel, marketing insights, retail site selection, ride-hailing, crowd heat, and logistics fulfillment.

Among them, crowd heat and logistics fulfillment are still marked "coming soon," while the other five are the scenarios highlighted at this stage.

Image from an official website screenshot

Among these features, I am more interested in retail site selection.

Suppose you are preparing to open a milk tea shop. The most headache-inducing thing may be choosing a location.

How many people are in the business district? Are there many young consumers? How many Mixue Bingcheng and Chagee stores are already nearby? Does this street actually have consumption potential?

In the past, this information often relied on offline scouting, commercial data analysis, and the experience of expansion staff.

LBSwork now wants to bring AI into this.

If a company has enough historical operating data, it can also try to build a sales forecast model to estimate the future sales of a new store.

A few minutes later, getting a report with maps, charts, and analytical conclusions is indeed quite tempting.

Of course, a beautifully written report and whether a store can make money are completely different matters.

So my biggest interest in this feature lies in whether it can accurately handle real commercial data.

Other scenarios are also somewhat interesting.

For example, LBSwork can generate map dashboards and visualization screens from a single natural language sentence, and even turn the places Li Bai traveled throughout his life into a footprint map.

This part is a bit like Vibe Coding, except it is specifically aimed at map application development.

There is also a relatively niche scenario: ride-hailing liability determination.

When a driver complains that navigation took a detour, or a passenger believes the route had a problem, the platform needs to check trip records, analyze road conditions, and then determine the cause. LBSwork demonstrates the ability to have AI analyze anomalies using trip logs and assist in handling work orders.

Compared with store site selection, this kind of business scenario is actually easier to measure in terms of efficiency improvement.

However, I have not yet obtained internal testing access to LBSwork, so the actual effect of these capabilities cannot be verified. In particular, analysis involving location big data and industry data still requires corresponding account authorization.

The official website introduction is very rich, but how much real-world capability the product has will still need to be tested by users.

03 Amap and Baidu: Tencent, you're late

But Amap and Baidu have long been doing similar things.

Let's start with Amap.

On July 16 this year, Amap officially released the business intelligence agent "Amap Ask Store."

Image from a DoNews report

This product is mainly aimed at merchants. It can do site selection analysis, business district assessment, and store operation diagnosis, and can also generate analysis reports, supporting collaboration across PC and mobile.

For example, if you want to open a coffee shop, you can have it analyze candidate locations and compare conditions in different business districts.

So, does it sound a bit like LBSwork?

Moreover, Amap also has its own spatiotemporal intelligence agent platform, providing development capabilities such as MCP and Skill, and its layout already covers multiple industry scenarios.

Now let's look at Baidu.

Baidu Maps' Huiyan business has long provided services such as commercial site selection, customer group analysis, regional research, and performance forecasting.

For map development, Baidu has also launched Maiya MAPYA, which can use natural language to assist in querying map APIs, generating code, and completing development tasks.

If you put the products of Tencent, Amap, and Baidu side by side, you'll find that everyone is doing similar things.

In the past, it was map APIs, location big data, and industry solutions. Now they are all adding large models, Agents, and Skills, along with natural language interaction interfaces.

Tencent's LBSwork has its own characteristics in the way it integrates products, but it will be difficult for it to establish an obvious advantage based on a desktop AI workbench alone.

Moreover, commercial site selection is inherently a professional market, and customers will not immediately replace their existing systems just because a company releases a new Agent.

For enterprises, what matters most is still data quality, business capabilities, and analysis accuracy.

For example.

You can't just have all three open a shop and see who's right.

In the end, you still have to compare each company's data basis, evaluation methods, and actual forecasting performance.

Tencent Maps' accumulated location service capabilities in the past provide a certain foundation in this regard. Amap and Baidu have also been operating for many years; none of them is a newcomer to the industry.

The AI competition in the map industry probably will not be decided just because everyone has built a similar WorkBuddy interface.

04 Copying WorkBuddy probably won't be that easy

How much of WorkBuddy's experience can actually be copied to LBSwork?

Image from the Tencent Maps official account

I think it needs to be looked at separately.

The technical framework can be reused, the interface interaction can be borrowed, and even product modules such as experts, skills, and connectors can adopt similar designs.

But user needs differ greatly.

WorkBuddy faces workplace office workers. Organizing documents, making PPTs, processing Excel, and searching for materials are tasks many people encounter every day.

Users also find it relatively easy to judge whether it is useful.

If a PPT is poorly made, you can just revise it; if Excel processing is wrong, running it again is no big deal.

Many of the customers LBSwork serves are professional clients.

What these customers care about is often data security, system stability, business process fit, and analysis accuracy.

Especially commercial site selection.

If WorkBuddy makes a bad PPT, it wastes at most ten minutes of your time. If LBSwork gives a severely distorted site selection report, the boss may lose hundreds of thousands of yuan.

Users' trust requirements for the two products are obviously different.

There is also the business model.

WorkBuddy can attract a large number of individual users through general office needs and then gradually explore paid services.

But many of the professional scenarios LBSwork targets require location big data, industry data, and enterprises' own operating materials.

The threshold for use is higher, and the number of potential customers will also be limited.

The premise is that the product must deliver reliable results.

Therefore, Tencent can bring over WorkBuddy's technology and product experience, but it will be difficult to directly copy its user growth method.

Since LBSwork and WorkBuddy use the same technical framework, why not directly connect map capabilities into WorkBuddy?

For example, WorkBuddy already has a skill system. If users want to do business district analysis, map visualization, or route planning, can they directly install the corresponding Skill?

For ordinary users, this method may be more convenient.

Otherwise, in the future, when making a market analysis report, you would first open LBSwork to generate a map, then switch to DataBuddy to analyze the data, and finally return to WorkBuddy to organize the PPT.

Of course, standalone products also make sense.

LBSwork involves a large number of professional map capabilities, requires complex visualization interfaces, and also needs to connect to enterprise accounts and industry data. Professional customers may prefer to use standalone tools.

And as a general office Agent, WorkBuddy does not need to stuff all industries' complex functions into one client.

Judging from the currently public product layout, Tencent already has multiple directions including general office, code development, data analysis, music creation, and map analysis.

Some product functions may inherently overlap.

If enterprise users want to analyze data, should they go to DataBuddy or LBSwork? If they want to create a map dashboard, should they use CodeBuddy to write code or directly let LBSwork generate it?

If every team designs standalone Agents from its own business perspective, there may be more and more products, while users find it increasingly difficult to choose.

This reminds me of how major internet companies used to make apps.

When a demand became popular, different internal teams might all try to build it. After the products spread out, they began to consider entrance integration, account interconnection, and functional coordination.

If this continues in the AI era, it may ultimately really produce a "Tencent Buddy family bucket."

As AI Agents proliferate, coordination between products will indeed become a problem.

At present, Tencent is both building standalone Agent products and opening capabilities through Skills, MCP, and connectors. Whether a unified entrance will ultimately form is still hard to judge.

06 Can Tencent create a second WorkBuddy?

From general office to programming, data, music, and then maps, more and more business teams are beginning to try Agent products.

The emergence of LBSwork shows that the technical capabilities accumulated by WorkBuddy are being used by other businesses.

This is a good thing for Tencent. After all, mature technology can be reused, which is always better than every team reinventing the wheel.

But I do not think every business needs to build a separate WorkBuddy.

Some are suitable as standalone apps, some are enough as a Skill, and for others, the capabilities can be directly embedded into existing business systems, which users may actually prefer.

Tencent has launched so many products in the past, but in the end only some of them can truly retain users over the long term.

LBSwork has only just begun internal testing, and competitors Amap and Baidu have already accumulated experience in related directions. Whether Tencent can gain an advantage with its own map data, enterprise customer resources, and WorkBuddy's technical foundation still needs to be observed.

I am quite interested to see whether Tencent will connect these Buddies in the future to form a complete AI product ecosystem.

For example, WorkBuddy receives tasks, DataBuddy processes data, LBSwork does map analysis, and finally other tools generate reports.

If this level of collaboration can really be achieved, the product and business resources Tencent has accumulated over the years may be able to play a greater role.

Of course, the premise is that they can really do the work well.

Tencent can launch WorkBuddy-style products in batches, but users will not download and install them one by one just because they are all named Buddy.

This article comes from the WeChat public account "Everyone Is a Product Manager" (ID: woshipm), author: Guai Ge, published by 36Kr with authorization.