AI Knowledge Management
Top Power University
Note: TPU is a pseudonym. Proprietary metrics and key information were anonymized.
Objective
Implement an immediate, high-impact intervention to stem loss of institutional knowledge, reduce duplication of effort, and streamline operations in the Curriculum Department.
Problem
As identified in the Current State Analysis, TPU faces a critical knowledge bottleneck, with duplicated efforts between various curriculum development groups and loss of institutional knowledge due to Big Power poaching TPU employees.
Solution
My team proposed using GenAI technology, specifically a Retrieval-Augment Generation (RAG) model, to solve this problem by acting as a secure and context-based search engine utilizing TPU’s comprehensive internal knowledge base.
Proposal
TPU departments, including Leadership and the Curriculum Department, adopt GenAI (specifically Microsoft CoPilot) integrated with Microsoft SharePoint to more efficiently and effectively leverage the organization’s explicit and tacit knowledge to analyze, design, develop, implement, and evaluate both curriculum and business reports.
Market Analysis
My team researched whether other organizations had similarly implemented GenAI. Research showed the Centers for Medicare and Medicaid Services had utilized Meta’s Llama 2 Large Language Model (LLM) to more efficiently process vast amounts of complex health documentation and deliver precise, context-based results (Bracken, 2024). An interview by the team with a Director of Customer Experience for a FinTech organization also outlined the organization's use of Microsoft's CoPilot LLM to manage technical writing knowledge management in a similar fashion to the proposal. Research into the use of GenAI for knowledge management in manufacturing environments has also shown benefits, such as improved knowledge sharing and faster training and support for new operators (Freire, 2024).
My team conducted a feasibility study of available Generative AI models to determine the best fit for TPU’s specific constraints.
Evaluation Criteria
We evaluated potential solutions against three non-negotiable constraints.
Data Security: TPU holds proprietary trade secrets (training methods). The solution must be a closed-loop system that does not train public models.
Integration: The solution must work within the existing ecosystem (Microsoft 365 Suite) to ensure user adoption.
Speed to Value: Custom-coded LLM solutions would increase cost and have a high development time.
As part of this proposal, the following LLM options were reviewed:
ChatGPT 4o
Claude Sonnet
Google Gemini
Microsoft CoPilot
Selection
After analyzing different LLM options, Team Linemen selected Microsoft CoPilot.
Rationale
Sits natively on top of the SharePoint infrastructure TPU already licenses.
Integration with SharePoint will streamline implementation.
Offers commercial data protection, ensuring that queries regarding sensitive intellectual property remain contained within TPU's ecosystem.
Automatically transcribes and summarizes recorded meetings, reducing the need for manual documentation.
Lower in cost when bundled with Microsoft 365 Suite, which is already in use by TPU.
Risks
Factual errors from content generation: LLMs can hallucinate if they cannot locate an answer. They will also repeat inaccurate information if that information is in their knowledge base.
To mitigate this risk, Team Linemen recommended approving the tool only for locating and referencing existing content within the organization’s knowledge base, and reviewing any reports generated using CoPilot for content accuracy.
Rapid market changes: CoPilot is one of several GenAI products offered by Microsoft and its affiliated companies. Given the fast-paced development in the GenAI space, newer, more effective tools could emerge, potentially requiring TPU to reassess its chosen platform.
Vendor dependency: Relying heavily on a single vendor (Microsoft) for AI and knowledge management tools could introduce risks related to pricing changes, product discontinuation, or shifts in service models over time.
Cost
Copilot pricing varies depending on an organization’s contract, but standard plans at the time of proposal were approximately $36 per user per month, which includes Microsoft 365 and the full Office suite. This translates to roughly $360 per user per year.
Curriculum Department
Each of the three curriculum teams would require one Copilot account.
The total estimated annual cost: $1,080.
Organization
10 additional accounts across the organization: $4,680.
Productivity
A Microsoft study found that Copilot users completed tasks like researching and summarizing information 29% faster (Microsoft, n.d.).
Sponsors
Team Linemen identified the internal sponsors required for the proposal based on interviews with members of the organization. These sponsors include:
Chief Executive Officer
Chief of Staff
Manager of Legal and Compliance
Director of Information Technology
Process Improvement Manager
VP, Program Delivery and Educational Quality
Project Timeframe
Expected Length
Approximately three months until implementation, with continuous maintenance.
References
Alavi, M., & Leidner, D. E. (2001). Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS quarterly, 107-136.
Bracken, M. (2024, April 17). CMS’s financial office is using LLM pilot to combat loss of institutional knowledge. FedScoop.
Freire, S. K., Wang, C., Foosherian, M., Wellsandt, S., Ruiz-Arenas, S., & Niforatos, E. (2024). Knowledge sharing in manufacturing using LLM-powered tools: user study and model benchmarking. Frontiers in Artificial Intelligence, 7.
Microsoft. What Can Copilot’s Earliest Users Teach Us About AI at Work? (n.d.). Retrieved from https://www.microsoft.com/en-us/worklab/work-trend-index/copilots-earliest-users-teach-us-about-generative-ai-at-work
Storey, V. C. (2025). Knowledge Management in a World of Generative AI: Impact and Implications. ACM Transactions on Management Information Systems.