SoftServe Launches AI Assistant SOFI – Already Used by Thousands

Iryna Tkach, Director of Talent Management at SoftServe, has written a column for Scroll.media about the company’s internal development — an AI assistant called SOFI, designed to help teams quickly access corporate information. In her article, Iryna outlines the entire process, from assembling the development team to the successful company-wide launch.

We all spend a lot of time searching for information every day. In a large organization with thousands of employees, the number of information-related inquiries exchanged between teams and departments is massive. Considering feedback and needs from our colleagues, we decided to create an internal AI assistant to simplify navigation through the sea of corporate data.

That’s how SOFI (SoftServe Intelligence AI Assistant) was created— an AI tool that helps employees easily find the information they need and perform daily tasks more efficiently. It supports both specialists and managers, enabling them to focus on what really matters.

Within the first two weeks, SOFI had 2,000 unique users who submitted 20,000 queries.

Building a Cross-Functional Team

Before launching new projects or validating ideas, we often conduct internal research or surveys. The same approach was taken for our AI tool. In 2024, we conducted an in-depth internal study involving managers at various levels, which revealed several key areas needing improvement. Managers mentioned that it was hard to access information and highlighted the need for help with onboarding new employees and finding internal resources for professional growth.

From the beginning, we knew developing an AI assistant would require combining different areas of expertise. We assembled a cross-functional team that included tech specialists, product managers, and HR professionals:

• Tech specialists handled the uninterrupted development, integration, and optimization of the AI solution;
• Product managers focused on the user experience, ensuring SOFI was intuitive and user-friendly;
• HR specialists were responsible for change management, supported the implementation of innovations, and looked into the ethical issues related to the AI tool.

By combining these diverse perspectives, we developed an assistant that is technically robust, easy to use, and tailored to the company’s needs.

A Thoughtful Approach to the Pilot Launch

We initially introduced SOFI to a pilot group — and enthusiasm was high. Over 100 employees volunteered to take part in testing. We carefully selected participants to ensure diversity across technical and non-technical departments, which helped us gather a broad range of feedback.

We provided clear usage guidelines and encouraged structured feedback on user-friendliness, answer accuracy, and overall experience. This approach enabled us to refine SOFI based on real use cases and detect potential issues a month before the full launch.

A separate testing stream was created for corporate content creators within the company: 25 teams tested prompts related to the spaces and groups they work with in the internal SharePoint system. They generated typical employee queries and evaluating SOFI’s responses. This process even helped uncover weak points in our corporate knowledge bases.

Working with the data that the AI tool would later access proved to be a major challenge. The clarity and structure of corporate data directly affect the assistant’s ability to find and deliver accurate search results. We spent significant time standardizing and organizing information across SoftServe’s internal platforms. One output of this work was a unified guide for SharePoint content optimization, which is now part of the onboarding process for teams managing internal content.

Prompt Engineering for Everyone

Just a few years ago, prompt engineering — the skill of crafting effective inquiries for AI systems — was little known. Today, it’s an essential competency. One of our main goals (and challenges) was ensuring SOFI returned clear, informative answers to user queries.

We created helpful documentation, ran workshops, and provided cheat sheets to teach employees how to craft effective prompts.

These efforts paid off: specialists quickly learned how to phrase their questions properly, improving their interaction with SOFI (and other natural-language-based search tools and AI assistants), which in turn boosted adoption and usage.

To determine SOFI’s readyness for full implementation, we carefully analyzed usage data and tracked key metrics such as:

• Answer accuracy
• User engagement and satisfaction
• Frequency of repeated queries (to identify content gaps)

This thorough approach helped us reduce risks and build trust in AI assistant from day one.

SOFI’s Results: Productivity, Specialists’ Experience, and Employer Branding

After SOFI’s launch, we noticed several notable benefits across the organization:

• Increased productivity: Employees spend less time searching for information and more time on critical tasks.
• Improved employee experience: With faster access to company knowledge, people feel more engaged and supported. SOFI can provide not only benefit and vacation balance information but also suggest mentors or Udemy courses for personal growth.
• Enhanced employer brand: SOFI showcases SoftServe’s commitment to innovation — a quality appreciated by both current employees and job candidates.

Our experience building SOFI confirmed the importance of cross-team collaboration, strategic testing, user training, and data management in AI-based solution development. In 2025, we’ll continue enhancing this tool and watching how AI transforms our ways of working.