Conversational Bots

A conversational bot is a program you talk with, usually in natural language. People use bots to find information, to have a concept explained, or to brainstorm ideas. Because one basic idea can be applied in very different settings, the most useful way to understand bots is to sort them by the job they do and by the kind of user they serve. This article follows that path. It starts with business uses, moves to a companion-style and research use, and then looks at bots that take on physical or game-world forms.

As you read, keep two questions in mind for every application: what is the bot trying to accomplish, and who is on the other side of the conversation? Those two questions are enough to tell most applications apart.

The source groups the applications of conversational bots into four areas: customer support, product copilots, social simulation, and 3D characters such as smart NPCs. The first two are business-facing tools that serve customers. Social simulation is a research setting in which groups of bots are studied. 3D characters are a presentation form, meaning a bot that appears as a visible character, most often in games.

The source gives two reasons companies adopt customer support bots. The first is cost: a support bot can help a company save costs. The second is customer experience: a support bot can improve it because it responds to users sooner than human agents do. Speed is the key idea. A customer who gets an answer immediately is often better served than one who waits in a queue for a person.

What distinguishes a copilot from a support bot is its purpose. A support bot answers questions when a customer asks them. A copilot walks alongside the customer while the customer works through a multi-step process, helping them understand what to do next. The copilot's attention stays on the task itself, not only on the single question in front of it.

Customer support bots and product copilots both serve customers, so they are easy to confuse. The table below separates them by the kind of job each does, using only what the source states.

Try the same reasoning on a different message: 'What is the return window for this product?' This is a single factual question with a single answer, so the bot is behaving as a customer support bot. The same platform can host both kinds of bot, so classify by the job being done in each conversation.

The wave of text-based bots that followed the success of ChatGPT is only one interface. The source points out that text is not the only way to interact with a conversational agent. Voice assistants such as Google Assistant, Siri, and Alexa have been around for years. A bot can also be presented as a 3D character, and the source says 3D conversational bots are already common in games and are gaining traction in retail and marketing.

A less obvious application comes from research rather than products. The source reports that researchers have used a group of conversational bots to simulate a society, which lets them conduct studies on social dynamics. Park et al. (2023) is cited as an example of this work.

The point for a learner is that bots are not only tools that serve people directly. They can also be the setting in which researchers study how groups of agents behave and influence one another. In this use, the bots are the subject of the study rather than the service being delivered.

Without AI, NPCs are typically scripted. They perform simple actions and speak from a limited range of dialogue. AI-powered NPCs, which the source calls smart NPCs, can be much smarter. The source states that intelligent bots can change the dynamics of existing games such as The Sims and Skyrim, and can also enable new games that were never possible before.

The flowchart below traces what happens after a player speaks to an NPC in each case. The top lane is a scripted NPC: the response is chosen from a limited set of prepared lines. The bottom lane is an AI-driven smart NPC: the bot produces a response in natural language. This is a teaching model built from the source's description of scripted versus AI-powered characters, not a description of any particular game's internal code.

Read the two lanes side by side. Both begin with the same player action, but the middle step differs. The scripted path depends on a bounded set of lines, which is why the source describes these characters as speaking from a limited range of dialogue. The AI path replaces that bounded set with a bot that can respond to what the player actually said. This single change is what the source means when it says AI can make NPCs much smarter.