How to succeed with generative AI

  – By adopting models for generative AI, you can improve everything from product presentations, financial reporting and internal processes to how you talk to your customers via customer services, says principal SAP consultant and AI expert at PearlGroup, Bjørn Hellem.

  – Undoubtedly, there is enormous potential and great competitive power in the technology. But it's about how you do it, he adds.
 
Generative AI is entering most industries at full speed as a ground-breaking technology, and companies compete to be the quickest in adopting it. Today's most talked about form of generative AI is ChatGPT from OpenAI. Although generative AI is far more than ChatGPT, it can serve as a comparison to understand the scope of the technology.
 
– The chatbot is trained on vast amounts of unsorted data that is openly available on the web, and without being fine-tuned to any specific area, subject or industry, it answers questions and completes tasks by grinding together all available data and generating a general answer. The quality is varying - along with several ethical issues, such as copyright, says Harald Vedvik, PearlGroup's other AI expert, who also holds the title of principal SAP consultant.

Tailor your company's AI with your data

ChatGPT is a so-called basic model based on a massive number of open data sources and thus it serves as a kind of advanced Google available to everyone. The two AI experts in PearlGroup say there are other criteria for companies that want to succeed with generative AI. Here, smaller models based solely on the company's data is applicable.
 
  – Imagine a specific type of ChatGPT tailored to your company with basic models based on smaller amounts but own data - such as product information, financial data or reports. Open Source technology opens up the opportunity to direct AI towards innovation, improvements in efficiency or anything else you want to focus on in your own company, says Bjørn Hellem.
 
Many companies spend considerable resources on writing product texts that contain a lot of relevant product data from suppliers and preferably a sales language that targets a specific target group. You can now entrust this job to generative AI along with several other tasks.
 
  – The generative AI can build up text and product descriptions that meet the criteria you ask it to consider, and it can also help you categorize new products. The scope for possibilities is enormous, says Hellem.

How does generative AI work?

How does ML work?

3 methods for using foundation models

Helpful assistant for a variety of purposes

In other words, you can feed your company's AI model with the data you want it to respond to and tell how it should react and behave.
 
  – This means you get relevant answers to what you ask about and retain control over ethical guidelines as you have set these guidelines yourself. You get the "answers" wrapped in the desired language, adapted to selected target groups - and with good grammar, sentence structure and language that means you only need to proofread, says Vedvik.
 
Generative AI can increase efficiency in the company while adding value to your services to customers in several ways.
 
 – Another possibility is to feed the AI with financial figures and use it as a chatbot against its own data. For instance, ask it about which products were the most sold or the turnover in a particular department in Q1 2022 – or use it as an assistant when you need to write reports, says Vedvik, and adds:
 
  – If the company has an online store, you can use an AI model to develop a virtual AI assistant and train it to talk to customers just like a real person who makes personalized and detailed suggestions, he says.

How to get started with your AI?

The first step to getting started with generative AI for your company is to find a suitable basic model for your company's needs. Vedvik and Hellem are ready to assist.
 
   –  We at Pearl can help you get started. Firstly, we identify your company's desire or vision and then find the right model match. Many of the AI models are pre-trained on a number of different needs or scenarios, so you only need to input your data. The effort required to get started is small. Hellem believes it is a low-hanging fruit that can have significant gains in the long term.
 
For forward-leaning companies that want to invest in innovation, there are also opportunities to finance the start-up with generative AI models.
 
  – We have been involved in several innovation projects with customers that have aroused enough interest in partners we work closely with, such as Amazon, to have triggered funding for the project. If you want to come to us with an innovation case do have ambitions on the opportunities, the two AI experts advise.

Three popular tools based on generative AI:

Sagemaker

Allows you to build an AI model entirely from scratch and train it according to your company’s needs. It gives developers the ability to quickly and easily build and train machine learning models, without having to manage infrastructure.

 

Amazon Forecast

A prediction model that is pre-trained to provide forecasts for future scenarios based on historical data. For example, the service can predict customer demand after a campaign, adjusted for variations such as season, holidays, and weather reports.

 

QuickSight Q

A generative AI model aimed at business reports, which allows you to “talk to the report” to retrieve insights about business data. The service is a self-service analytics tool that enables you to make informed business decisions.