Skip to main content
AI Features

AI in Unpack

What AI Can Do in Unpack

Unpack integrates artificial intelligence throughout the retrospective workflow to help your team reflect more deeply, surface patterns faster, and leave every retro with clear, actionable outcomes. AI in Unpack is designed as a co-pilot: it suggests, drafts, and highlights, but your team always makes the final call.

Here is a summary of the AI-powered capabilities available across the product:

  • Card Coaching — Helps team members write clearer, more specific, and more actionable feedback cards during the reflect phase.
  • Smart Grouping — Analyzes card content and suggests thematic groupings so your team can move through the grouping phase faster.
  • Discussion Guide — Generates a structured discussion guide with talking points, suggested questions, and time allocation recommendations for the facilitator.
  • SMART Action Item Drafts — Proposes specific, measurable, achievable, relevant, and time-bound action items based on the themes that emerged during the retrospective.
  • Theme Insights — Identifies recurring themes and sentiment patterns across multiple retrospectives, giving teams a longitudinal view of their improvement journey.
  • Facilitator Co-pilot — A chat-based assistant available to facilitators during live retros for real-time guidance, conflict resolution tips, and on-the-fly suggestions.

Powered by Anthropic Claude

Under the hood, Unpack routes AI requests through OpenRouter and defaults to Anthropic Claude as the language model. Claude was selected for its strong performance in nuanced text analysis, its ability to follow structured instructions, and its emphasis on safety and helpfulness.

OpenRouter acts as the gateway layer, which means Unpack benefits from high availability and the flexibility to incorporate model improvements as they become available. Your team does not need to configure anything — the AI is ready to use out of the box on supported plans.

Unpack does not train AI models on your retrospective data. Your content is sent to the AI model only when you explicitly trigger an AI feature, and it is not retained by the model provider after the response is generated.

Privacy and Data Handling

Your retrospective data is sensitive. It often contains candid team feedback, honest assessments of processes, and sometimes interpersonal observations. Unpack takes this seriously and has built its AI integration with privacy at the forefront.

What data is sent to the AI

When you trigger an AI feature, Unpack sends only the minimum context needed for that specific task. For example:

  • Card coaching sends the text of the individual card being coached.
  • Smart grouping sends the card titles and content within a single retrospective.
  • Theme insights sends aggregated, anonymized theme summaries from selected retrospectives.

Unpack never sends personally identifiable information such as email addresses or user names to the AI model. Card content is transmitted without author attribution, preserving the anonymous-by-default nature of the platform.

Data retention

AI responses are stored within your Unpack workspace so your team can reference them later. However, the underlying AI model provider does not retain your prompts or responses beyond the immediate request-response cycle. Unpack's data processing agreement with OpenRouter ensures your content is not used for model training or improvement.

Enterprise controls

Organizations on the Enterprise plan can configure additional controls around AI usage, including the ability to disable specific AI features at the organization level, require facilitator approval before AI suggestions are shown to the team, and audit AI feature usage across all retrospectives.

AI as a Co-pilot, Not a Replacement

Every AI feature in Unpack follows a human-in-the-loop design pattern. The AI suggests, and your team decides. This principle is reflected across the product in several ways:

  1. Explicit triggers — AI features are never automatic. A team member or facilitator must deliberately activate each AI capability.
  2. Accept or reject — Every AI suggestion can be accepted, modified, or dismissed entirely. Nothing is applied without human approval.
  3. Transparency — AI-generated content is always clearly labeled so your team knows what was written by a person and what was suggested by the AI.
  4. Editable outputs — Accepted AI suggestions become regular content that can be edited, moved, or deleted just like anything else in the retro.

The best retrospectives combine AI efficiency with human judgment. Use AI to handle the mechanical parts — grouping, drafting, summarizing — so your team can focus on the conversations that matter.

Plan Availability

AI features are available on the Pro and Enterprise plans. Teams on the Free plan can run full retrospectives with all core features but will not have access to AI-powered capabilities.

Here is a breakdown of AI feature availability by plan:

  • Free — No AI features. Full access to manual retrospective workflow.
  • Pro — All AI features included: card coaching, smart grouping, discussion guide, SMART action item drafts, theme insights, and facilitator co-pilot.
  • Enterprise — All Pro AI features plus organization-level AI controls, usage auditing, and priority processing.

AI feature usage on the Pro plan is subject to a generous monthly allowance per team. Enterprise plans include unlimited AI usage. See the Plans & Pricing article for detailed limits.

Getting Started with AI

If your team is on a Pro or Enterprise plan, AI features are enabled by default. You do not need to install anything or configure API keys. Each AI feature is accessible from within the relevant phase of the retrospective workflow.

To explore each feature in depth, continue reading the articles in this category. We recommend starting with Card Coaching to see how AI can improve individual contributions, then working through Smart Grouping and Discussion Guide to understand how AI supports the facilitation flow.