Many tasks are repetitive. When using AI for these tasks, you must specify the context, desired tone of voice, target audience, and other preferences each time to achieve a good result. This can be made easier. AI chatbots, also known as AI assistants, are a tool for repetitive, specific tasks. A chatbot is given fixed rules and instructions that you can reuse repeatedly. For example, get inspired by the culture chatbots (opens in new tab) from DEN. In this article, we discuss three functionalities of generative AI tools, provide examples of tasks for which you can use AI chatbots, and share tips to get started yourself.
Three Ways to Use Generative AI
To understand what an AI chatbot is, it helps to compare three ways of using generative AI tools.
Individual Chats
This is the method most people are familiar with. You ask a question to an AI tool and get an answer. You don’t build on previous conversations, and there are no fixed rules or instructions. This method works best for quick, one-time tasks. Think of generating ideas (brainstorming), looking up information, or having a text written once.
Project Environments
Some generative AI tools (ChatGPT and Claude) offer a project feature. In a project, you can upload multiple documents, work longer within the same context, and build on previous conversations. The project feature is useful when you want to perform multiple tasks where the same context (documents) is beneficial. For example, brainstorming about an event, developing a project plan for one of the ideas, finding speakers for the event, writing a website text about the event, etc. Or writing a grant application.
However, a project environment is not yet a custom AI chatbot.
Custom AI Chatbot
AI chatbots are a tool for repetitive and specific tasks. You set the instruction once: the role, the goal, the context, audience, tone of voice, desired output, etc. Additionally, you can add relevant documents. You can then repeat this task infinitely. Because the chatbot always receives the same instruction, the answers are more consistent. This is very useful if, for example, you always want to maintain the same writing style for your organization or if you always want to apply the same structure in a project plan.
Individual Chats | Project Environment | |
Usage | One-time questions or tasks | Multiple tasks within the same project; building on earlier chats |
Added Files | One or a few files useful for the specific task | One or a few files useful for every task |
Result | Variable | Flexible and consistent |
What Can You Do with a Custom AI Chatbot?
AI chatbots can be used in many situations. Below are some examples for inspiration:
- Communication and marketing – (Re)writing texts in your own house style, simplifying to B1 level, making texts SEO and GEO-proof, or checking readability.
- Education – Creating lesson material about the collection, reusing educational formats with new content, or adapting content for teenagers.
- Policy and projects – Structuring project plans according to a fixed format or rewriting annual reports and summarizing minutes.
- Data and evaluation – Summarizing survey results, processing visitor feedback, and analyzing evaluations.
- Testing target audiences – Testing communication or concepts by asking questions from a described target audience/persona. Not a replacement for a real test panel, but a good first check.
What Tools Can You Use?
Not every AI tool offers the ability to create chatbots, and you always need a paid account for this. Below is an overview of the main options.
Provider | Product | Chatbots | Explanation |
OpenAI | ChatGPT | CustomGPT’s | With a free account, you can use CustomGPT’s created by others. You can only create your own CustomGPT’s with a paid account. |
Microsoft | Copilot | Agents | Works well within Microsoft 365. Copilot Agents are not the same as AI agents. (opens in new tab) |
Gemini | Gems | Integration with Google Workspace possible. |
At the time of writing, other (more responsible) AI tools do not offer this functionality or it is much more complex to do so.
How to Build Your Own Custom AI Chatbot
Want to create your first AI chatbot? Follow these steps! There are also many instructional videos available on how to create an AI chatbot. For example, the following for CustomGPT’s (opens in new tab), Copilot Agents (opens in new tab), or Google Gems (opens in new tab).
Step 1 – Choose One Specific Work Process
Start with one clearly defined task. For example: rewriting texts for your audience, structuring a project plan, or one of the other mentioned examples. The more specific your task, the better and more consistent the result.
Step 2 – Write a Clear Basic Instruction
A good instruction describes the role, the context (what should the AI know about your organization, role, or task), the tasks (what can the AI do and what not?), the audience (for whom is it intended?), the style (what tone and language level?), and the desired output (article, summary, 10 ideas). Use our five tips for prompting (opens in new tab) and ask AI to improve your prompt. Additionally, ask AI to improve your prompt.
Example of a simple basic instruction: “This AI helps rewrite information for museum visitors. The language level is B1. The texts are informative, neutral, and understandable. Do not add new information to the article.”
Step 3 – Upload Relevant Documents
Determine what additional information is needed to perform the task well. Think of the writing or house style, policy plans, project descriptions, educational formats, or previous grant applications. Upload only what is necessary for the task.
Step 4 – Test with Real Examples
Use examples from practice and check: is the output correct? Does the agent stay within the rules? Is the tone consistent? Is the quality good (enough)? By working with real examples, you quickly see what works well and what needs adjustment.
Step 5 – Improve and Implement
Refine instructions based on your tests, add documents, or remove them. Then share the chatbots within the workspace of your AI tool with colleagues and clearly describe what the chatbot can and cannot be used for.
Delve into the Risks of AI
The use of AI is not without risks (opens in new tab). Ensure that you and your colleagues have the right knowledge and skills (opens in new tab) to use the chatbot. Make agreements on how to handle AI in an AI policy (opens in new tab) for your organization.
Creating an AI chatbot is not complicated. Check the chat history of your AI tool to see which tasks you often repeat, and start with one of these tasks. With clear instructions and the right documents, you can go a long way. This way, you create your own “assistant” that knows your workflow, helps you save time, and thinks along with you when needed.
AI Chatbots, Custom GPTs, and AI Agents: What’s the Difference?
An AI chatbot is a general term for a digital assistant you can chat with to perform tasks, such as writing texts or summarizing information. A custom chatbot (e.g., a custom GPT) is configured by you with fixed instructions, allowing you to repeatedly perform the same task with consistent results. In ChatGPT, these chatbots are called Custom GPTs, in Microsoft Copilot Agents, and in Google Gemini Gems. An AI agent goes a step further: it can perform multiple steps and sometimes independently take actions in other systems. This article mainly discusses custom chatbots for repetitive tasks.
More articles per topic
