Why Skanska launched a new AI assistant and how it plans to keep data secure

Earlier this month, Swedish contractor Skanska launched a new artificial intelligence (AI) chatbot called Skanska Sidekick.

The new internal AI tool is aimed at helping Skanska’s workforce to be more productive by assisting them in harnessing knowledge from across the entire company. Skanska said it hoped Sidekick would help to inform better decision making and improve value for customers.

Construction Briefing spoke to Will Senner, leader of Skanska USA Building’s Data Solutions team to find out more about how it works and its benefits, how Skanska aims to keep its data and those of its customers secure, and how he expects the technology to develop in the future:

Why did you decide to launch Skanska Sidekick?

With the unveiling of ChatGPT and other AI chatbots, companies across the world were exposed to the power of natural language processing and the associated productivity gains these tools can deliver. 

However, our cyber security team flagged numerous risks with these publicly available tools - company-specific, proprietary information entered into publicly available tools can leak and be used in future iterations of the model.

Built on our private Microsoft Azure environment, Skanska Sidekick harnesses the advanced language capabilities of the GPT-4 model while preventing data leaks and ensuring the security of our customers’ data. This means our teams can safely and securely enter company-specific, proprietary information while improving our ability to develop content, summarize information and conduct research.

How does it work and what sort of benefits do you expect it to bring in terms of efficiency, communication, and customer satisfaction?

Skanska Sidekick provides a chatbot interface that users can interact with to answer questions.

Unlike publicly available AI chatbots, Skanska Sidekick allows users to upload multiple documents at once. This dynamic and unique feature allows users to query against information provided in the uploaded documents and provides insights, summaries, and answers based on the user’s requests.

Our teams have been using the tool to generate content; summarize information like meeting minutes, create data analysis expressions (DAX) codes and Excel formulae; and learn about a variety of complex subjects or how best to speak to them with partners and customers.  

An example of a prompt and Skanska Sidekick’s response (Image supplied by Skanska USA)

What are Skanska Sidekick’s limitations?

Skanska Sidekick offers the same functionality as the public-facing GPT4 model, which is limited by the chatbot’s 2022 knowledge cut-off date. For queries that are tied to more specific construction-related insights – project-specific information, construction methods, ADA requirements – users can use Sidekick’s document upload feature.

Moving forward, we have a roadmap to integrate additional information which is more specific to our projects and our shared knowledge and expertise across Skanska.

What safeguards are in place to ensure the security and privacy of customer data?

Information entered into the chat is not used to inform future model improvements, even for other users within Skanska.

This ensures that confidential client information stays with the individual users.

As we integrate additional Skanska specific data and knowledge, the team is focused on protecting existing access rights and privacy so that any information which is only intended to be shared with a certain group of people is not inadvertently exposed to a wider audience via the AI.

Will its introduction require training for Skanska employees and will you offer a guide to customers on how to use it effectively?

Our team has created dynamic training materials to help employees understand the strengths and limitations of Skanska Sidekick and generative AI tools.

Our training materials include tips for effective prompting, details on use cases for specific roles within the company, and internal channels for our teams to share opportunities they discover. With this being a new and rapidly evolving technology, we are focused on adapting these training materials over time to incorporate new use cases and best support our employees as they make these tools more and more a part of their day-to-day work.

How do you hope this technology could develop in the future. What more would you like it to be able to do that it can’t do currently?

Skanska began investing in our data strategy over five years ago, creating an effective foundation to allow us to quickly respond to and adopt disruptive innovations like generative AI. 

With any AI tool, having accessible and clean data is a critical starting point, so those early efforts have proved incredibly important.

In the future, we will layer in more Skanska-specific information in the tool. This means connecting data from projects and information from our internal knowledge-sharing repositories to Skanska Sidekick, allowing employees better access to project-specific insights and enterprise knowledge.

We believe this has incredible potential to improve knowledge sharing across the business and enhance how we manage risk and develop base value solutions for customers in a way that truly leverages the unique power and insights of Skanska across the globe.

Brought To You By