How Stackpact Calculates Your AI Score

RAG (Retrieval-Augmented Generation) is an AI architecture that enhances large language models by giving them access to external knowledge bases. Instead of relying solely on training data, RAG systems can retrieve and reference up-to-date, specific information.

Stackpact uses RAG to ensure our assessments and recommendations reflect the latest AI tools, techniques, and best practices in software development. This means when we mention a tool like "Claude Code" or "Cursor," we're drawing from current, accurate information rather than potentially outdated training data.

The score is designed to be directionally accurate and actionable rather than a precise measurement. Our assessment methodology was developed by analyzing patterns from thousands of developer interactions with AI tools.

The score is most valuable for: (1) tracking individual progress over time, (2) identifying specific skill gaps, and (3) comparing relative strengths across the four dimensions. We recommend using it as a guide for development, not as an absolute measure of ability.

Stackpact primarily uses Claude by Anthropic, currently one of the highest-performing models for nuanced conversation and technical understanding. We specifically chose Claude for its strong reasoning capabilities and ability to maintain context across complex, multi-turn conversations.

For enterprise customers with BYOK (Bring Your Own Key) enabled, you can use your own Anthropic API key, giving you full control over your AI infrastructure and costs.

We use carefully engineered system prompts that guide the AI through a structured assessment methodology. These prompts ensure:

• Consistent evaluation criteria across all assessments
• Natural, conversational flow that doesn't feel like a test
• Comprehensive coverage of all four scoring dimensions
• Adaptation to individual responses without losing assessment integrity

Our prompts are continuously refined based on assessment outcomes and user feedback to improve accuracy and relevance.

No. Your assessment conversations are not used to train any AI models. We use Anthropic's API with data retention disabled, meaning your conversations are not stored by the AI provider after processing.

Your data is stored securely in our system solely for the purpose of providing your assessment results, tracking your progress, and generating personalized recommendations. Enterprise customers can also opt for self-hosted deployments for complete data control.

We recommend quarterly assessments for most teams. This cadence allows enough time to implement recommendations and develop new skills while keeping pace with the rapidly evolving AI landscape.

However, teams actively focused on AI adoption initiatives may benefit from monthly check-ins, while stable teams might prefer bi-annual assessments. The platform tracks progress over time regardless of frequency.

Yes! Stackpact is designed for seamless integration into your existing workflow:

• Embeddable widget - Add to any internal portal with a single script tag
• SSO integration - Pass user IDs for personalized tracking
• Domain whitelisting - Control exactly where the widget can be embedded
• Custom configuration - Tailor assessments to your tech stack and policies

Yes! Stackpact offers native plugins for both Microsoft Teams and Slack. These integrations allow your team to:

• Take AI skills assessments directly within their communication platform
• Receive automated reports and progress updates in dedicated channels
• Use slash commands for quick assessments (e.g., /stackpact assess)
• Set up scheduled reminders for periodic skill check-ins

The plugins ensure team members never need to leave the tools they already use daily, maximizing adoption and engagement.