Artificial Intelligence: Ready for the AI Revolution?

Artificial Intelligence Header — Source: iStock/mihailomilovanovic

Artificial intelligence (AI) cannot think for you. However, these technologies can help free up your mental capacity to explore new thoughts and put existing ones to the test. They allow you to use your potential to master new challenges by completing less demanding routine tasks more quickly. As a result, teams not only work 40 per cent more efficiently, but they also consider their work to be more fulfilling.

High-quality project results, a fast time to market and relaxed, happy employees. Is that not how working life should be? And it can be – if you actively foster AI expertise within your company and know how to use which tools in what situation.

Learn how dotSource is driving this development in companies and gain valuable tips for your business in this article.

AI Expertise Through Transparency

No change process can succeed without transparent communication within the company – and the fact that the rapid development of AI technologies requires companies to change their processes or even existing business models is nothing new.

But, if almost every company department will be affected by these new technologies sooner or later, where is a good place to start?

Raise Critical Awareness

The allure of new technologies is strong – especially when it comes to AI solutions that promise to take over mundane tasks. But before you get started, you should have an idea of the legal framework within which you can use these technologies for your daily business.

Firstly, you need to be aware of the fact that all content shared with your AI solution will be saved and processed. So, before you begin using a tool, inform yourself about the company behind it and how trustworthy it is. Ideally, have your legal team check it and publicise the results to the whole company.

Furthermore, you should never share customer-related data or your company’s internal content with these tools. Thanks to NLP (natural language processing), tools like ChatGPT are particularly user-friendly. This often tempts users to share sensitive data with these tools. Take time beforehand to think about how you can anonymise requests – if this is not possible, you should not be using an AI tool for this particular case.

If you have generated an image with the help of AI, it needs to be marked as such. Some tools, such as DALL-E 2, already do this for you and watermark the images they create – in the case of DALL-E 2, with coloured squares on the right side of the image. If there are people in the picture, you technically need their consent in writing – however, since it is usually impossible to know who is on an AI-generated image, it is best to abstain from the use of these images altogether.

Test New Options

Every day, new products and solutions emerge that make use of current AI developments. These bring seemingly limitless opportunities for use within companies, teams or even for individuals. So, naturally, questions arise such as:

Which tasks can be completed more quickly with the help of AI?
What effect does this increase in efficiency have on business processes?
How can my team use this to contribute to the superordinate business goals?
Which changes need to occur to make perfect use of this potential?

To answer these questions, employees need to develop new skills in dealing with AI tools. They need to ask themselves:

Which tools can I integrate into my daily work?
How can I optimally communicate with AI solutions?
What limitations do these tools have?

Do not forget, practice makes perfect – or, in this case, makes experts. To achieve satisfactory results, people and machines will have to perfect the way they work together.

However, oftentimes daily business does not leave a lot of time to test new technologies, let alone to develop completely new strategies or intelligent business models.

dotSource used its annual Hackathon to take an in-depth look at the opportunities, challenges and innovative ideas surrounding AI. Over the course of two days, employees worked together on questions related to the use of AI:

In digital marketing when it comes to automatically creating metadata
In software development to establish concrete guidelines for its use
In product data processing to automatically enrich data
In internal processes to easily summarise effort bookings

Share Acquired Knowledge

Through cross-team collaboration, experiences with new developments can be shared and driven forward collectively. However, in order to promote real change, regular personal interaction and communication is needed on a company-wide level. It is helpful to designate people who are responsible for sharing AI knowledge and driving developments forward.

In this regard, dotSource founded a community of practice (CoP). This community consists of employees from different departments who exchange information on new insights and challenges to learn from each other.

The members of this CoP then go on to share their knowledge in presentations which all employees of the company can attend, allowing them to ask questions and raise concerns.

AI Expertise with the Right Prompts

Knowledge is also necessary when it comes to prompts, the text input which you want your tool to respond to. These are especially important when an AI solution processes your request via NLP. This is based on a large language model (LLM), which was previously trained with large amounts of information to generate the best possible response to your request. To gain a satisfactory answer, you need to be clear about the output you expect from the tool.

There are various methods you can use to nudge AI tools into the desired direction.

Prompt Engineering: An Overview of Methods

Of course, you can submit your requests to AI tools in German. However, the English library of the LLM that several AI tools are based on is much more comprehensive and will deliver more precise responses.

Prompt Method	Explanation	Prompt Example / Entry
Instructions	Tell the AI tool what to do or ask a question.	Which keywords match the keyword »field service management«? Sort the keyword by relevance and proximity to the main keyword in a table.
Role Prompting	Give the AI tool a role and provide it with additional context. Ask yourself: Which person could best give me information on a certain topic?	Let´s pretend you are […]
Few Shot Prompting	Use examples to show the AI tool what kind of output you expect.	Develop higher-level attributes and list possible variants from the following products: […] Example: Higher-level attributes and possible variants: […]
Chain of Thought Prompting	Use examples to explain how to arrive at a solution. The AI tool repeats this procedure and is significantly less prone to errors.	The odd numbers in this group add up to an even number: 4, 8, 9, 15, 12, 2, 1. A: Adding all the odd numbers (9, 15, 1) gives 25. The answer is false. The odd numbers in this group add up to an even number: 15, 32, 5, 13, 82, 7, 1. A: [LLM will directly answer when a response pattern is specified]
Zero Shot Chain Prompting	Let the AI tool explain its thought process in as much detail as possible.	Let‘s think step by step. Give me a step-by-step guide on how you went about creating higher-level product attributes and possible variants for each of the products.

These prompt methods can also be combined to receive answers to more complex questions or build on previous requests. Furthermore, AI responses can be adjusted through specific prompts. For example, a text can be generated in a certain style, format or tone.

One example for this could be: »Write a technical blog article about ChatGPT. Write in a motivating and professional tone.«

The first NLP prompt can determine the entire conversation. That is why it can be worth it to develop »priming prompts«. These are comprehensive prompts that combine several of the previous methods to outline one specific scenario. This does not just control the output, but also reduces the complexity of subsequent user requests.

You can find a collection of helpful prompts that can be used for ChatGPT models here.

Furthermore, there is a Chrome extension which helps you formulate prompts. At dotSource, this extension is primarily used to create prompts for code generation: PromptStorm fills in blank CHatGPT prompts

AI Expertise with the Right Tools

In the last few weeks, we have discussed several AI tools on Handelskraft, including ChatGPT, DALL-E and Microsoft Cloud. Now, we want to share how these AI tools can specifically help development teams generate high-quality code quickly. However, you should never blindly rely on the output of these tools, but rather see it as inspiration or food for thought. Ask yourself if the response is correct and use your expertise for an optimal result.

ChatGPT

ChatGPT has recorded 100 million active users within two months and thereby brought AI to the forefront of the societal mind. First used to draw inspiration for e-mails, presentations and blog articles, it has increasingly found use in software development. By entering short prompts, developers can generate code, script or snippets with one click.

The tool can help you create automated workflows and therefore complete repetitive tasks a lot faster.

ChatGPT can also lighten the workload when it comes to documentation. The right prompt can create a well-structured document containing all the relevant information on the code as well as the entire project.

GitHub Copilot

When it comes to AI, Microsoft has clearly positioned itself as a pioneer. While Microsoft Copilot facilitates administrative tasks in multiple areas, GitHub Copilot supports your developers in generating code. While they are still busy with the input of their current code, the tool is already making suggestions for code snippets or entire functions.

Furthermore, bugs in existing code sequences can be analysed before comprehensive testing is undertaken by a quality assurance tester. This way, developers can carry out programming tasks twice as quickly.

Bing AI

Microsoft offers another useful AI tool in the form of ChatGPT integration into the Bing search engine. It can be integrated directly via Microsoft Edge. Bing Chat can be used to make requests or interact directly with websites.
For example, we used this tool

for suggestions on how to handle new development tasks, e.g. for creating new product attributes
in dealing with new tools, e.g. for a tutorial on how to link two systems (GitLab and Jenkins)
to support bug fixing, e.g. to suggest necessary configurations for existing code
to create bash scripts for automatically creating tasks on a Unix- or Linux-based operating system

Find further alternatives to ChatGPT here.

Master the Digital Revolution – Download the White Paper for Free Now!

The digital world is constantly changing. Efficient tools, new commerce strategies and modern methods of customer communication offer your company countless opportunities to get ahead of the competition. To ensure your success, it is important not to lose sight of the big picture. The »Digital Transformation« white paper provides you with helpful tips on how to master the permanent change.

Fill out the form now and receive a free copy in your inbox!