AI is revolutionizing how we work, making information more accessible and processes more efficient. Yet, as companies grow, so does the challenge of navigating an overwhelming amount of internal data spread across multiple platforms. At SoftServe, we listened to associates’ feedback around that and saw an opportunity to create an internal AI assistant that would simplify navigating corporate knowledge.
by Iryna Tkach, Global Talent Management Director at SoftServe
Named SOFI (SoftServe Intelligence AI Assistant), it is designed to make daily tasks more efficient, helping both employees and managers focus on what truly matters. The creative journey was filled with valuable lessons—some expected, some surprising—that shaped SOFI into the trusted assistant it is today, gaining 2,000 unique users in the first two weeks and handling around 20,000 queries so far. In this article, we’ll share insights on how we brought SOFI to life, from assembling the right team to ensuring a successful company-wide launch.
Assembling a Cross-Functional Team
From the beginning, we knew that creating SOFI required a blend of expertise. In 2024, we conducted an in-depth internal study on the “Manager Experience,” revealing key areas for improvement, including access to information, onboarding support, and ongoing development resources. SOFI became one of the tools we implemented to equip managers with the resources they need to lead effectively.
We assembled a cross-functional team that included tech specialists, product managers, and HR professionals. Each played a vital role:
● Tech Experts ensured seamless AI development, integration, and performance optimization
● Product Managers focused on user experience, making SOFI intuitive and practical
● HR Professionals guided change management, encouraged adoption, and ensured ethical AI use
Bringing together different perspectives helped us create an AI assistant that was not only technically advanced but also user-friendly and aligned with our company’s needs.
A Thoughtful Approach to the Pilot Phase
When we introduced SOFI to a pilot group, the enthusiasm was overwhelming—over 100 associates applied to volunteer in a testing process! However, rather than opening it up to everyone, we carefully selected participants across different roles and departments to ensure diverse feedback. To set the pilot up for success, we created clear guidelines, encouraging structured feedback on usability, accuracy, and overall experience. This intentional approach allowed us to refine SOFI based on real-world use cases and identify potential roadblocks during one month before a full-scale launch.
A separate testing stream involved corporate content owners, with 25 teams testing prompts related to SharePoint spaces they manage. They created common employee queries and evaluated SOFI’s responses. This phase highlighted weak points in our databases, leading to the development of a unified guide for optimizing SharePoint content, which is now an ongoing educational effort for content owners.
Making Prompt Engineering Easy for Everyone
One challenge we faced early on was helping users get the best responses from SOFI. A few years ago, prompt engineering wasn’t widely understood, so we developed easy-to-follow guidelines, workshops, and quick-reference materials. These efforts paid off, as associates quickly learned how to phrase questions effectively, improving their interactions with SOFI (are also other search engines and AI assistants that understand natural language queries) and boosting adoption rates.
Knowing When It Was Time to Launch
Determining SOFI’s readiness for full deployment required a data-driven approach. We tracked key metrics, such as:
● Accuracy rates of responses
● User engagement and satisfaction scores
● The frequency of repeated queries (to identify gaps in information)
By continuously improving SOFI based on these insights, we ensured it met a high standard of performance before rolling it out company-wide. This careful approach minimized risks and increased trust in the assistant from day one.
Overcoming Data Challenges
One of the biggest hurdles was ensuring SOFI had access to clean, organized corporate data. We collaborated with data owners across departments to standardize and structure information, making sure SOFI could provide accurate, reliable answers. This parallel effort of improving data management while developing SOFI was crucial to its long-term success—however, it was not new to SoftServe, as we invested in search infrastructure and information discoverability earlier.
The Impact: Productivity, Employee Experience, and Brand Reputation
Since launching SOFI, we’ve seen tangible benefits across the organization:
● Increased productivity: SoftServe’s associates spend 66,7% less time searching for information and more time focusing on meaningful work.
● Enhanced employee experience: With quick access to company knowledge, people feel more supported and engaged (as SOFI can provide not only vacation balance or benefits lists, but also is able to advice on mentorship or recommend Udemy learning path for individual needs)
● Stronger employer branding: SOFI highlights SoftServe’s commitment to innovation, making it a more attractive workplace for top talent.
Our journey with SOFI reinforced the importance of collaboration, strategic testing, user education, and data governance in AI development. As we continue to evolve and enhance SOFI in 2025, we look forward to seeing how AI will further shape the future of work. By sharing our experience, we hope to inspire other companies to embrace AI-driven solutions for a smarter, more connected workplace.
