AI Features

RWRD integrates AI throughout the platform, from onboarding data extraction to ongoing advisory intelligence. All AI calls are routed through secure server-side Edge Functions -- the AI API key is never exposed to your browser.

AI Chat Sidebar

The AI Chat is accessible via the chat icon in the dashboard header. It supports four display modes:

Normal (modal): A floating dialog window centered on screen with a chat history area and input field at the bottom.

Sidebar: Click the dock icon in the chat header to pin the chat as a 380px-wide panel on the right side of the screen. The dashboard content shifts left to accommodate it, so you can read AI responses while viewing your metrics. Click the dock icon again to detach back to modal mode.

Fullscreen: Click the expand icon to fill the entire viewport with the chat interface. Useful for longer analytical conversations.

Minimized: Click the minimize icon to collapse the chat to a floating button in the bottom-right corner. The button shows a message count badge so you know conversation history is preserved. Click the button to restore.

The chat header shows the current connection status: "AI Connected" when the server-side AI service is configured and responding, or "Pattern matching mode" when the AI service is unavailable (in which case RWRD falls back to built-in pattern-matching responses for common questions).

A Clear button resets the conversation history. The chat preserves messages across open/close cycles within the same session.

How AI Context Works

Every message you send to the AI chat includes a comprehensive snapshot of your current dashboard state. The AI does not just answer generically -- it has access to your actual data.

The system prompt sent with each message includes:

• Health Score -- Your current composite health score
• Organization identity -- Company name, current user name and role
• Summary counts -- Total objectives, financial metrics, risk KRIs, and critical/at-risk item counts
• Status distribution -- How many items are green, yellow, and red in each domain (OKR, Financial, Risk)
• Top 5 priority items -- Ranked by status urgency with impact assessment
• Critical items -- All items in red/critical status
• Key objectives -- Your top 5 objectives with progress and status
• Key financial metrics -- Your top 5 financial metrics with values and variance from target
• Cross-domain analysis (when available) -- Top systemic drivers, vulnerabilities, and interdependency insights
• Strategic context (when available) -- Purpose, vision, mission, values, pillars with alignment counts, themes, and unaligned objective counts

This means the AI can reference your actual metric names, values, and statuses. When it says "Your Employee Turnover Rate KRI is in the red zone at 18% against a threshold of 12%," it is reading real data from your dashboard.

Executive Summary

The AI Executive Summary is a one-click feature available from the Health Score card or the AI Report button in the dashboard header.

When triggered, it sends a structured prompt asking the AI to provide:

1. Health Assessment -- One sentence on overall organizational health with the health score
2. Top 3 Priorities -- The three most urgent items requiring executive attention, with specific metric names and values. The AI prioritizes items with high systemic importance when cross-domain analysis is available
3. Key Recommendation -- One actionable recommendation the CEO should act on this week

The summary uses a larger response budget to allow for a more detailed analysis.

The response is specific to your data -- no generic advice. Every metric name, value, and status reference comes from your actual dashboard state.

Strategic Context in AI

When strategic context has been configured (purpose, vision, mission, values, pillars), the AI system prompt includes a STRATEGIC CONTEXT section that enables:

• Values-aligned recommendations -- The AI can flag when a recommended action might conflict with a stated company value
• Pillar-aware prioritization -- The AI references which strategic pillars have at-risk objectives and which have no OKR coverage
• Alignment gap identification -- The AI highlights unaligned objectives and pillars with zero linked OKRs
• Strategy-grounded responses -- When you ask about priorities, the AI frames its response in terms of your stated strategic direction

If no strategic context exists, these features are invisible -- the AI operates normally without strategic framing.

Example Queries by Use Case

For a CEO reviewing the dashboard:

• "What needs my attention this week?"
• "Give me an executive summary"
• "Which risks could cascade across the organization?"

For a CFO monitoring financial health:

• "Which financial metrics are trending in the wrong direction?"
• "Show me industry benchmarks for our sector"
• "What is driving the gap between our current ratio and target?"

For a risk manager:

• "Which KRIs are in the red zone?"
• "What has the highest systemic influence?"
• "Which risks could cascade?"

For an OKR owner:

• "How are my key results tracking?"
• "What should I focus on to get back on track?"
• "How do I add a new key result?"

General navigation help:

• "How do I add metrics?"
• "Explain the health score"
• "What do the colors mean?"

When your industry is set, an additional query appears: "Show me industry benchmarks." The AI provides industry-specific key metrics, benchmarks, common risks, and recommendations.

AI Limitations

Response time: AI responses typically take 5-15 seconds. If the service times out, a helpful error message appears with a suggestion to try again.

Fallback mode: When the AI service is unavailable (network issues, service outage), the chat falls back to a built-in pattern-matching system that provides helpful but generic responses to common questions. The chat header displays "Pattern matching mode" to indicate this state.

Data scope: The AI sees a snapshot of your dashboard state at the moment of the query. It does not have access to historical data, audit logs, or data from other organizations. It cannot make changes to your dashboard -- it is read-only and advisory.

Security: All AI calls route through secure server-side functions. API keys are stored server-side and never exposed to the browser. Your dashboard data is sent to the AI for context but is not permanently stored by the AI service.

Not a substitute for judgment: The AI provides analysis and recommendations based on your data, but it does not replace executive judgment. Treat AI suggestions as one input among many in your decision-making process.