noun_Email_707352 noun_917542_cc Map point Play Untitled Retweet Group 3 Fill 1

ChatGPT and beyond: What is the impact of generative AI on information work

For the first time in history, AI is available for all. To benefit from ChatGPT and other ground-breaking tools, it's good to get hands-on and play. There is a new skill to learn - Prompt Engineering.

Data Insiders / June 07, 2023

From customer service to coding and even traveling, Lukas Lundin from Microsoft shares his insights on how generative AI can transform the way we work and live.

AI (Artificial Intelligence) as a term was coined already in the 1950s and started to gain new interest during the 1980s and 90s with neural networks and computer vision. Natural Language Processing (NLP) took focus in 2000s with models that were able to understand human language. In recent years, Large Language Models (LLMs) – transformers – have revolutionized the game by being able to consume large amounts of data without supervised learning.

OpenAI’s GPT (Generative Pre-trained Transformer) model gained significant attention as it was able to perform various natural language processing tasks, such as text generation, summarization, and question-answering. ChatGPT, using GPT model as its foundation, made the ground-breaking technology available to everyone, attracting users in record time.

“Generative AI was a user interface problem that was solved. ChatGPT gave a good user interface to an already existing service. The advantages of these tools are accessibility and ease-of-use” says Lukas Lundin, Data and AI Go-to-market Manager at Microsoft, who has a front-row seat to all the latest developments.

Listen to Data Insiders podcast episode with Lukas Lundin on Spotify.

Learn to ask good questions – Prompt Engineering is the new skill for knowledge workers

Other recent tools from OpenAI are Codex, which can convert human language into code and transform one piece of code into another programming language and DALL-E that generates images from human language.

Lukas Lundin advises that organizations should encourage everyone to experiment with these technologies and practice prompt engineering, which is becoming an essential skill to every information worker.

You don't need to be a data scientist or know anything about machine learning. What you need to know about is prompt engineering, which is basically a fancy term for asking good questions.


Lundin understands that organizations have concerns about AI, especially regarding security and privacy. Microsoft has recently brought GPT models to everyday productivity tools such as Word, Outlook and PowerPoint and provides access to OpenAI’s services in Azure environment. This enables companies to stay in control of their data. One solution to safeguard that the models are not abused, to generate malicious code or spam for example, is Azure Open AI Service’s content moderation filtering that runs both the input prompt (questions) and generated content through an ensemble of classification models that detect the misuse.

How do organizations use generative AI

The explosion of pre-trained models has made companies to experiment and find ways to incorporate these technologies into their tasks and operations. Lukas Lundin sees four categories of use cases getting most traction in an enterprise space:

  1. Customer support. Pre-trained models can assess free text or speech from different channels and classify and assign them according to urgency or other value. They can also generate answers and respond to customers.

  2. Documentation and reporting. AI services can extract information from various sources of structured data and generate reports and documentation based on them.

  3. Coding-related tasks. Models such as Codex can be used to read through code changes, and even create documentation and review if the code changes are following the organization’s chosen development patterns.

  4. Enterprise search. Large language models can improve semantic search within organizations through vectorization. The model is fed with information and documentation, and it translates words into numerical vectors so that computers can understand them. When a question from a user is translated into another vector, you can compare the vectors and find the closest match between a question and an answer.

New use cases are emerging all the time as organizations and individuals are exploring the possibilities of AI.

Read also: ChatGPT will change work faster than expected

This is the first time in history we can offload intelligent tasks to a computer. We had the industrial revolution where machines became autonomous and replaced a lot of manual labour. Now we will have the same happening with information work.


Generative AI solutions as a gateway

Lukas Lundin is an AI enthusiast both at work and outside the office. He is using AI tools to create marketing content, write summaries and emails, but also in his free time to find new recipes and interesting sights to visit.
“On a vacation in Tokyo, I visited different districts and Bing Chat worked as a travel guide during the trip,” he explains.

Looking ahead, Lundin expects that transformer models will work as gateways to different AI services.

“I think, in the future, we will have a GPT model that takes in questions and accesses various other models and resources to complete the task. It will be able to orchestrate whole pipelines,” Lundin predicts.


Stay ahead of the data revolution with Data Insiders, the #1 Data Podcast!

Insiders brings together experts to discuss the trends and phenomena around this hot topic in a way that's easy to understand and entertaining. Learn the truth behind the hype and buzzwords and gain the knowledge and support you need to make your business data-driven. Presented by Tietoevry, with Thomas Kirk as your host. Tune in and join the data revolution with Data Insiders!

Join now to keep on receiving the newest data insights.

Data Insiders
The #1 Nordics Data Podcast

The #1 Nordics Data Podcast, invites top professionals and visionaries of data-driven business to share their learnings and insights. Find all episodes on Spotify!

Share on Facebook Tweet Share on LinkedIn