How SmartCat Turned Scattered Project Knowledge Into an Always-Current Source of Truth

In a growing software company, the most important project knowledge rarely lives in one place.

Architecture decisions sit in code repositories. Team information is stored elsewhere. Budgets live in Drive. Delivery context is spread across project tools, documents, and conversations. Some of the most useful information may never be written down at all.

The result is not just messy documentation. New team members take longer to become productive, managers spend time chasing status updates, and leadership lacks a reliable view across projects.

SmartCat built ACE to change that. Instead of asking people to document and maintain project knowledge manually, ACE uses an AI agent to collect it from existing systems, organize it in one knowledge base, keep it current, and answer questions in plain language.

Problem

As companies run more projects across more teams and tools, project knowledge becomes increasingly difficult to find and trust.

The information exists, but it is fragmented across code repositories, Google Drive, Notion, project-management tools, and the knowledge of a few experienced employees.

That creates several recurring problems.

New hires and team members spend days or weeks rebuilding context that already exists somewhere. Delivery managers are interrupted for routine questions because there is no dependable self-service source. Employees outside a project have limited visibility, even when their experience could help identify a risk or reuse an existing solution.

Documentation also becomes outdated quickly. Under delivery pressure, updating a wiki is rarely the highest priority, so architecture notes, team information, deployment instructions, and project status gradually drift away from reality.

This creates a wider business risk. Knowledge stored only in people’s heads does not scale, becomes vulnerable when employees leave, and makes it harder for leadership to make informed decisions quickly.

Results

ACE created an automated workflow for documenting, maintaining, and accessing project knowledge without adding recurring work for delivery teams.

When a new project is added, the agent reviews the available codebase, team records, and budget documents. It then creates a structured project overview covering the architecture, technology stack, team, budget, setup, and deployment process.

What would normally require someone to gather information from several people and systems can be assembled in a single workflow.

The documentation is then maintained automatically. On a scheduled basis, ACE compares the current project information with what is already stored and updates only the parts that have changed. This prevents the knowledge base from becoming another static archive that gradually loses value.

Employees can also query the system in plain language. They can ask who is working on a project, how it is deployed, which technologies it uses, or what budget has been assigned. ACE retrieves the answer from the central knowledge base without requiring the user to access the underlying code repository, Drive folder, or project system.

This opens project knowledge to a wider group of employeesa while preserving access boundaries around the original systems.

ACE can also support planning. Leadership and delivery managers can use the same system to review market developments, assess competitor capabilities, and generate prioritized feature recommendations for a specific project.

The result is more than a searchable company wiki. ACE creates a knowledge layer that supports onboarding, project oversight, cross-team collaboration, portfolio visibility, and strategic planning.

Its integration model also allows additional sources to be connected over time. Jira can add sprint status and delivery velocity, while GitHub can provide contributor activity and pull-request history, extending the project view without changing the underlying architecture.

SmartTip Booster

Before investing in a new internal knowledge base, decide who will keep it current.

When the answer is that project managers, engineers, or team leads will update it when they have time, the system will eventually become unreliable. Better templates and stricter processes rarely solve that problem for long.

The more sustainable approach is to remove manual upkeep from the workflow. Let the system collect changes from the tools people already use and update the knowledge base automatically.

SmartFact Booster

ACE can create, maintain, and answer questions from a complete project knowledge base without anyone manually editing its pages.

People continue working in their existing systems. The agent reads the relevant sources, updates the central record, and makes the information available through natural-language questions.

That shifts documentation from a separate task people must remember to perform into an automated part of project operations.

About the Client

The client was SmartCat itself.

As a software development and IT consulting company running multiple client projects in parallel, SmartCat manages project knowledge across different teams, technologies, delivery processes, and business systems.

The same information needed for onboarding, delivery oversight, and management reporting was often spread across several tools or held by individual employees.

ACE was built to reduce that dependency, shorten the time needed to understand a project, lower the number of routine information requests directed at delivery managers, and give leadership a more consistent view across the project portfolio.

The same approach can be applied to other software and consulting companies dealing with fragmented internal knowledge and documentation that is difficult to keep current.

Technologies Used

• Claude Code for documentation generation and ongoing maintenance
• Claude Desktop and Cowork as the interface for querying the knowledge base
• Model Context Protocol for connecting the agent to company systems
• Notion as the central structured knowledge base
• Google Drive as a source for team and budget information
• Code repositories as a source for architecture, technology, setup, and deployment data
• Jira and GitHub as planned sources for delivery status, contributor activity, and development history