Stumbling blocks and tips – using chatbots and ChatGPT in the right way

1. Insufficient understanding of the technology

Generative AI and chatbots are complex technologies that have to be understood to be used effectively. It’s important to familiarise yourself with the fundamental mechanisms, opportunities and limits of these systems. A common misconception is that AI can think and act independently. In actual fact, it’s based on algorithms and data provided by people.

Want our advice? Familiarise yourself with the basics of generative AI and chatbots.

2. Misunderstandings in communication

The way questions and instructions are phrased greatly impacts the quality of the answers chatbots give you. Unclear or ambiguous phrasing often leads to misunderstandings and irrelevant answers. This is why it’s essential to phrase enquiries clearly and precisely to be able to obtain the information you need. This requires a certain amount of practice and understanding of how AI actually works.

Want our advice? Learn from typical errors to improve your prompts – we’ll tell you more about this later. Make sure you organise appropriate training for your employees.

3. Overestimating the capabilities of a chatbot

Another possible stumbling block is overestimating what AI and chatbots can actually do. They are powerful tools but they do have their limits, particularly when it comes to understanding and emotional intelligence. Users should have realistic expectations of the answers and solutions these technologies offer. Not all human capabilities can be replaced.

Want our advice? Don’t get ChatGPT to create an entire promotional campaign on the basis of just one prompt. Instead, use ChatGPT to generate ideas and drafts which can then be developed further by teams of creative people.

4. Disregarding data protection and security

Data protection and security are crucial aspects when it comes to using AI technologies. It’s important to protect sensitive information and ensure that any interaction with chatbots and generative AI complies with data protection policies. Not paying this aspect sufficient attention can lead to data leaks and other security risks, so make sure you deal with personal and confidential data conscientiously and responsibly.

Want our advice? Make sure that your company’s chatbot is compatible with current data protection standards. This helps you to build your customers’ trust in you.

5. Lack of customisation to specific requirements

Every company and every user has specific requirements and expectations of using AI. Standard solutions are often not sufficient to fulfil these needs. That’s why configuring chatbots and AI systems – and adapting them to the individual circumstances and goals – is essential for their successful deployment. This requires in-depth understanding both of your own requirements and of the opportunities these technologies offer.

Want our advice? Take a good look. For example: A company starts using ChatGPT to answer simple customer enquiries and forwards more complex cases to human employees. Customer satisfaction increases due to the efficient initial answer level.

The five biggest errors you can make when writing ChatGPT prompts

Prompts are the entries you make in ChatGPT – in other words, what you write in the chat window of ChatGPT. Incidentally, “prompts” is a good word when it comes to looking for further tips online. We have collected the key stumbling blocks and top tips for you here.

1. Prompts that are too vague or are ambiguous

One frequent error when it comes to dealing with ChatGPT is that you write vague or unclear prompts. This often leads to answers that don’t meet expectations. A prompt should always be phrased specifically, purposefully and as clearly as possible to ensure the required answer is obtained. This requires an in-depth understanding of how ChatGPT processes information.

Want our advice? Use precise, clearly phrased prompts. Ask ChatGPT whether it needs you to provide it with further information or details.

2. Incorrect interpretation of answers

Another problem is the incorrect interpretation of the answers provided by ChatGPT. Users should understand that the answers are based on previous training and don’t always reflect the context or nuances a human would be able to provide. It’s important to evaluate the answers in the context of the questions asked. They are based on algorithms which might not be able to answer all aspects of a question.

Want our advice? Check answers from ChatGPT carefully and don’t just be satisfied with the first answer. If necessary, ask for clarification.

3. Inappropriate expectations of the response time

It’s often the case that users expect really fast answers from ChatGPT. Usually, ChatGPT does react immediately, but complex queries or technical limitations can impact the response times. Users should have realistic expectations of the speed of the system.

Want our advice? When dealing with generative AI, just remember that good things come to those who wait.

4. Ignoring context and previous interactions

ChatGPT is capable of keeping the context in mind over several messages. This aspect is often ignored, resulting in every query incorrectly being looked at separately. To communicate effectively, users should take the context and previous interactions into consideration in their prompts. This makes more coherent and relevant answers possible – especially in longer dialogues and complex discussions.

Want our advice? Describe the context of your query to ChatGPT. You can assign a role to ChatGPT (“You are an experienced financial market analyst and…”).

5. Lacking follow-up and checking of the answers

Another mistake we often see is that answers from ChatGPT are not checked. ChatGPT might be a powerful tool but its answers are certainly not infallible. Users should always consider the answers very carefully, particularly when specialist knowledge or details are involved. This is the only way you can make sure that the information is correct and suited to the particular context.

Want our advice? The best idea is to take a second look, and don’t just take everything ChatGPT tells you at face value.

Become a pro by getting support from the pros

To summarise, you can see that the effective use of generative AI and chatbots such as ChatGPT requires both a fundamental understanding of the technology and deliberate and precise interaction. If you avoid the stumbling blocks listed above and apply our tips, this will certainly increase your satisfaction with AI applications and the quality of the results they give you.

As experts in artificial intelligence and language, we would be happy to help you when it comes to putting these technologies to optimal use to help drive forward your company or your projects.