intro
Hi, I'm Alex Leuchte, Head of Growth and Practice Lead AI at Etribes. I have now founded several companies that I am very proud of, including Keleya. With Keleya, I have developed a complete ecosystem of digital products related to pregnancy to support pregnant women on their way through this emotional and demanding phase of their lives. In my private life, I am also often testing and experimenting innovative concepts, including building a weather balloon that sent a camera and a small SIKU space shuttle into space, which is now on my desk, or the restoration of my beloved VW van.
I am now at Etribes with the clear goal of digitizing Germany. Personally, I see myself as a founder, not as an advisor. As a person who rolls up their sleeves and builds things themselves, with the mission to get companies and products from 0 to 1, from idea to implementation. Since I regard this unconditional will of “pragmatic execution” as a skill set lacking in the German economy, I am making it available to precisely these companies.
Would you like to learn more about building prototypes and AI? Then feel free to write to me directly via LinkedIn or via the contact form below!
This is how AI creates the next “iPhone moment”
“We have the next iPhone moment!” — But what does that mean and what does artificial intelligence have to do with it? In 2007, the iPhone was launched and fundamentally changed the way customers interact with digital technology. For the first time, you no longer had to sit in front of your PC to visit a website because everything was easy to use in your pocket. Companies that responded quickly were able to benefit enormously from this technological disruption, while established companies such as Nokia missed out on this trend.
AI is also one such disruption. The biggest challenge that CEOs, CTOs and everyone else out there must face today is how to close this technological gap in order to participate in the disruption and benefit from it sustainably. In this article, the term “artificial intelligence” refers to the sub-segment of generative AI or the large language model, which we often associate with products such as Chat GPT.
In this context, let's use the iPhone again as a reference: The fascination for the iPhone was omnipresent, but no one really knew how to actually tap into new business potential on the basis of this revolutionary device. Over time, however, completely new products were created that were not even remotely predictable at the time the iPhone was introduced.
In this blog post, you will learn what generative AI is, how it works and how it can be successfully implemented in companies. I'll also give you exclusive, practical insights into several prototypes that we at Etribes have implemented using generative AI.
What is generative artificial intelligence?
The German definition for generative AI (generative artificial intelligence) is generative artificial intelligence. Generative AI is a form of artificial intelligence that can generate new content based on existing information (training data) and user requirements (user prompts). This content can include text, images, videos, audio content, program codes, 3D models, molecular structures, and more. In some cases, the generated content is barely distinguishable from human-created content. Generative AI differs from discriminative AI, which is aimed at differentiating and classifying inputs, but does not generate any new content itself. (Source)
Generative AI can be divided into unimodal and multimodal AI. Unimodal AI only creates or processes a specific type of data, while multimodal artificial intelligence is qualified for various types of data. Generative AI uses methods and technologies such as trained neural networks, supervised and unsupervised machine learning (deep learning) and various AI algorithms. (Source)
Skeptics vs. Explorers: The Two Sides of Generative AI
On the one hand, there are the skeptics: As in any technological disruption, there are so-called “keepers” who stick to the status quo, according to the motto: “old fashioned intelligence works pretty well.”
On the other hand, there is the “Creative Explorer”. That is currently probably 99% of users on LinkedIn who use generative AI as a personal assistant. For example, by writing 30% more efficiently or reducing your research by 15%.
But in the end, it's not just about individuals writing emails faster, researching more efficiently, or improving their documentation, it's about everyone in your organization working more efficiently and conscientiously. After all, generative AI, as we understand it today, is based on publicly available world knowledge. Every organization, on the other hand, has its own data that they can use to make the results of AI usable for themselves.
That's why smart people are now asking themselves the questions: What can AI do for me? How does it actually work? Which work processes in my organization are frequent, manual and suitable for the use of artificial intelligence? How can I improve my customer interaction and customer loyalty through the use of AI? Is my organization even ready and able to work with AI? How can I enrich and improve generative AI with my company data?
How does Generative AI work?
In simple terms, artificial intelligence consists essentially of receiving input and then delivering output. Meanwhile, something is happening with this data. It takes over the transfer from input to output and carries out certain process steps. However, these processes are the core where AI can provide maximum support.
To explain it in more detail: This process is driven by three types of prompts, the system prompt, assistant prompt, and user prompt.
- System prompt:
The system prompt defines the system in which the AI should move. If, among other things, you are a staffing agency, you must tell the AI that they should act in this area. If you're a mechanical engineering company that sells cranes, the AI should also be aware of this. It's important that AI knows the size of your company and what products it offers.
- Assistant prompt:
With the help of the so-called assistant prompt, you can build a data pipeline that uses various data sources that may need to be updated continuously. This data is fed into the AI in real time and can significantly improve the results generated.
- User prompt:
With User Prompt, you interact directly with the AI. You enter information or ask questions to define what you actually expect from artificial intelligence.
How are projects with generative AI implemented at Etribes?
Step 1
First, we identify use cases that are relevant to you and your company. The examples below, which I'll go over in this post, may be completely irrelevant to you, as you and your company work in a completely different way. But that's exactly what the first step is about: finding out what works for your organization applies equally to B2B and B2C scenarios.
We look at connecting points in various areas of the organization, whether in marketing, customer support, customer engagement, finance, research, or even HR. However, that is not the complete list. There are many uses that can look completely different for your organization. This is why this step is so crucial, as it is not a plug and play solution. It requires individual consideration and adjustment to the specific needs and circumstances of your organization.
Consider classic key account planning, among other things: In your organization, it may take place once a month or every few days. That may not be the first port of call we would choose to use AI, particularly on a systemic level. However, if your organization is a staffing agency that regularly calls people, for example for cold calls, then this is a process that may take place in your organization hundreds or thousands of times per day. In this case, the process has a significant impact on sales due to sales activity. At the same time, it causes high costs due to the large number of employees who have to carry out this process.
Step 2
This is followed by our philosophy of “built, learn, and iterate” — i.e. building prototypes. Instead of spending a lot of time on PowerPoint slides, business case calculations, or other things, it's about building a working product and working with it. We are firmly convinced that measurable success is possible within a few weeks. One example of this is when we have two teams compete against each other, one with support from AI and one without. Through a clear test, we not only enable users to interact with the technology, but also enable users to reduce their fear of AI on a completely normal, everyday level while understanding the magic and benefits that artificial intelligence can offer. Ultimately, we aim to achieve concrete results and create added value. Our approach is designed to not only use the technology itself, but also to guide the people who use it on their daily journey and help them understand and utilize the benefits of AI.
I have brought the following examples with me: B2B Customer Research, B2B Customer Sales Documentation. In doing so, we considered two criteria. On the one hand, it is about impact, i.e. whether this process step can either solve an efficiency problem or have the potential to increase efficiency. For example, it can significantly increase the conversion rate and thus lead to increased sales. We also look at the measurability, for example how often this process step occurs in your organization, so that you can not only measure the idea of the impact, but also measure the success of the prototype in practice and adapt it accordingly, depending on how this use case is implemented in your organization.
Prototype 1:
How Generative AI supports B2B companies with cold calling as part of the go-to market
#1 - STATUS QUO
First, let's introduce a persona: Marvin is an employee in a staffing company whose primary responsibility is not to make mistakes. He makes around 30 phone calls a day and the conversion rate, i.e. the proportion of successfully sold services, in this context is around 2 to 5%, depending on various factors. To improve this rate, the company has set that every employee should have 15 minutes to research and research about the potential customers they call. You may often get a cup of coffee or rely on experience and no longer complete the task as well as at the beginning. As a result, the quality of phone calls varies greatly. When we look at the overall process, it is on the one hand about prioritizing the customer list, then researching the people who should be called and finally documenting the phone calls. In the end, we hope for a successful conclusion. The problems that arise are manifold: - The prioritization of the customer list is determined subjectively by the supervisor. - The research work requires time and is not always carried out consistently. - The length of phone calls varies greatly, from 20 seconds to ten minutes.
#2 - PROBLEM STATEMENT
Depending on how the phone call goes, a different amount of documentation is entered into the CRM system. This results in a wide range of quality sales documentation. This ultimately makes it difficult to identify best practices and success factors. The quality of documentation varies, there is a lack of standardization, or important data is missing.
#3 - RESULT
The solution here was to completely automate the research part and significantly simplify documentation by introducing automatic transcriptions. As a result, we were able to reduce the time required for research from 15 minutes to an efficient three minutes, which significantly increased the efficiency of telephone calls. At the same time, we were able to significantly improve our knowledge of Marvin's customers by collecting information about what customers need and when. This in turn has significantly increased the conversion rate of appointments. After the appointments, we took a proven approach, which is the use of voice messages. Nowadays, Marvin can do the documentation by recording a type of voice message. This recording is then processed using transcription technology. Based on certain rules, a management summary is created and follow-ups are then identified and carried out. This resulted in a significant improvement in data availability and quality. In addition, the automated execution of follow-ups was able to increase the follow-up rate by 42%, which is a significant improvement given that there was often no documentation before. In the medium term, this can have a significant impact on increasing sales.
Prototype 2:
How Generative AI creates a seamless customer experience for online fashion retailers
#1 - STATUS QUO
Now we come to another example that, to be fair, comes from the B2C sector, but it shows how easy it is for companies to overcome their own barriers in their heads. It's about retailers who are usually dependent on being the last step in the customer journey. In addition, the average shopping cart is heavily influenced by which products the customer selects or which discounts lead him to buy these products.
#2 - PROBLEM STATEMENT
The classic customer journey in this case is as follows: First, there is inspiration, such as the question of what you could wear to a wedding next week. Then you think about what you already have in your wardrobe and what might go with your partner's outfit. Planning then begins, where you think about where you need to go and what kind of clothing would be suitable for this. At some point, you start buying the clothes you want, either in a physical store or online. During this customer journey, as a customer, you are heavily dependent on the various categories and filters available on the websites of online retailers in order to find a coherent outfit. Of course, this requires effort and there are various steps during the buying process that can make the entire customer journey unnecessarily complicated. This example shows how companies can use innovative approaches and technologies to simplify the customer process by overcoming barriers and complexity in the customer journey.
#3 - RESULT
First, we developed a scraper that uses the retailer's web shop and downloads all relevant data such as images, product data and sizes. We then created a web application where we ask the user about their specific occasion, such as an outfit for a wedding, instead of just looking for a t-shirt. For this purpose, we have developed a Prompt system that describes the environment of the outfit in which it should be located. A forum is then created in which all product data is integrated and users are asked to put together an appropriate outfit. The outfit is generated and throughout the trip, the final name and environment of the outfit are integrated into the web application. An image is created that represents the name of the outfit and shows the products included. The web application also displays a story that matches the outfit. The available sizes for the individual products are displayed together with a direct link to the web shop. In addition, any discounts are pointed out to show how much money you could save if you bought the entire outfit. We are also currently working on finding a solution for actually presenting the outfit to people and how they can try it on.
Final Conclusion
AI has created a new “iPhone moment” by fundamentally changing the way companies work and create value. This disruptive force presents companies with challenges, but also offers enormous opportunities.
The development of AI has already shown that it promotes efficiency in various areas and generates added value. Now is the time to actively take the step and develop and test the first products. Because at Etribes, we believe that it is best to implement it quickly in order to have something tangible and thus plan and tackle the next steps in a more concrete way.
Together with our top programmers, we have developed AI prototypes to optimize sales processes. For example, these prototypes automate call preparation and enable voice-based documentation of sales calls in order to make all associated processes more efficient and faster.
Let's think together about where AI can help you, for example in a first short introductory phone call. Just get in touch with us using the form below or directly via LinkedIn!
.png)
