Chatbot Product Manager Guide (Part 2): Classification
-
If you've read the previous chapter, you've likely encountered at least 10 chatbot products. Completing the assignments may expose you to over 20. As a newcomer to the chatbot industry, you might feel overwhelmed. Therefore, learning to classify is essential knowledge for chatbot product managers, as it greatly benefits your systematic understanding of product frameworks.
This is part of a series. Recently, I've been training new colleagues in chatbot-related topics—this is the second lesson. The plan is to cover 8 topics in total, including:
Classification essentially means categorizing things based on certain criteria. There's no standard classification method—it's a highly open-ended question. You can classify products based on dimensions such as: whether they have a visual interface, single-turn/multi-turn conversations, closed-domain/open-domain, etc.
Here, I'll introduce two classification methods I believe are essential for product managers: classification by purpose and classification by platform.
From the perspective of purpose, chatbots can be divided into two types: task-oriented and conversational.
Task-oriented bots include: personal assistants, specialized bots, customer service bots, etc. Conversational bots include: chat companions, companionship bots, etc.
Why classify chatbots by purpose? The reason is that these two types of bots have different product focuses.
If your product is a task-oriented bot, you should focus on whether it helps users solve business problems.
For example, a personal assistant chatbot should be able to extract schedules (e.g., meetings, train tickets) from SMS or email content and remind users at the right time.
Similarly, a customer service bot should be capable of understanding non-standard phrasing, especially in industries like finance where users may not know the technical terms and use unconventional expressions.
In short, the focus of task-oriented chatbots is on the "task" itself, not the chatbot. Often, chatbots aren't necessary—people use them just for the sake of having a chatbot. The product's value only emerges when the business scenario fits chatbot usage.
People often get caught up in the superficial appeal of chatbots and try to force business applications onto them, unaware that task-oriented and conversational bots are fundamentally different—they just share the chatbot label.
If you neglect business value, your chatbot will fail. Because what task-oriented chatbots really compete with are traditional button-based business functions.
If your product is a conversational bot, your focus should be on the fun and relevance of the content.
For example, if a user asks, "What was Bolt's best race?" the bot might reply: "Hey, I know Bolt—he's the fastest man on Earth, nicknamed the Jamaican Flash." While it didn't fully grasp the user's intent, it recognized "Bolt" as an entity and gave at least a 60-point answer.
This case shows that, compared to task-oriented bots, conversational bots care less about logic—they just need to stay "relevant" (all answers relate to Bolt).
Additionally, content fun is crucial. Only fun keeps the conversation going, since the goal is to kill time.
With a vast corpus (and human moderation for quality), a tagging algorithm, and a retrieval tool, you can create a decent conversational bot—ensuring answers are somewhat relevant and look good.
Unfortunately, many task-oriented chatbots follow this path. Their ceiling is low—they only handle simple Q&A, barely needing chatbot tech; a search engine would suffice.
That said, not all conversational bots offer simple relevance-based answers. Some great products (e.g., Replika) deliver solid multi-turn conversations with internal logic.
From the platform perspective, chatbots come in three types: smart hardware, standalone apps, and embedded apps.
Smart hardware currently includes smart speakers and built-in phone assistants, with wearables likely joining soon.
Standalone apps are mobile apps like Replika or Migu Lingxi. Embedded apps are chatbots within other apps, like AliMe in Taobao.
If your chatbot runs on smart hardware, your core task is uncovering use cases.
Take smart speakers: users mostly play music, check weather, or set reminders.
On reflection, phones handle these tasks—so why use smart speakers? Simply put, smart speakers outperform or out-convenience phones in specific scenarios.
Smart hardware isn't just speakers—it could be smartwatches or phone assistants. You must identify competitors (often phone apps) and find where hardware does better or easier.
I believe smart hardware chatbots have the brightest future: it's a blue ocean with huge potential, while phone usage is saturated, making app-based chatbots challenging.
If your chatbot is a standalone mobile app, your priority is finding a niche and boosting retention.
Mobile internet's second half brings two key changes:
Currently, chatbots lack killer apps and lag in screen-time battles, making daily retention critical.
My advice: avoid pure conversational bots. Their goal is fun chats to kill time, but they compete with games, short videos, and livestreams—far stronger rivals.
Take Microsoft's Xiaoice—when did you last chat with it? It periodically "pings" users to stay relevant.
In our fragmented, entertainment-driven, low-effort era, chatting isn't a great pastime—especially since current experiences disappoint.
For embedded-app chatbots, learn to distinguish functions suited for bots from those that aren't. These products mix chatbots and traditional button-based features.
This raises a question: what are chatbots' strengths? It's broad—I'll cover it systematically in Part 4. For now, here are two examples.
Also, note that embedded apps often serve as customer service bots—chatbots' most mature commercial use, cutting labor costs for many firms.
Though titled "Chatbot Product Manager Guide," I don't expect readers to become experts after reading. If this helps you start in chatbots, I'm pleased.
To truly become a chatbot PM, reading isn't enough—independent thinking and hands-on practice are key.
Thus, each article ends with exercises to spark your own ideas, beyond the content.
Next up: Chatbot Product Manager Guide (Part 3): Technical Implementation.