Product Intelligence

If a team can capture messy product knowledge with almost no effort, and AI organizes it into decisions, approvals, requirements, owners, risks and open questions that people verify, then the team will find trusted answers faster and stop re-litigating settled work.

Product knowledge is fragmented across documents, research, meeting notes, decisions, designs, approvals, conversations and people's heads. Before acting, teams repeatedly find information, reconstruct why a decision was made, determine what is current, and re-explain context. This prototype tests whether a shared workspace with evidence-backed answers removes that tax.

• Product managers running a product or a cross-functional initiative
• Engineering, design, content, legal and research partners contributing knowledge
• People joining an effort mid-flight who need current context fast

• Workspace — a shared container for a product's or an initiative's knowledge.
• Source material — the raw files, images, notes and links people add.
• Candidate knowledge — something AI believes is a decision, approval, requirement, owner, risk, dependency or question. Not trusted until a person confirms it.
• Trusted knowledge — confirmed items shown on the Knowledge tab.
• Decisions and open questions — first-class objects with owners, status, dates and sources.
• Answer confidence — every answer is marked Known, Uncertain, or Unknown, and never claims certainty the workspace can't support.

• Product / Capability Memory — durable knowledge for a long-lived product or capability.
• Initiative / Problem Memory — knowledge for a specific initiative, problem, experiment, or temporary cross-functional effort. Knowledge outlives the team.
• In this prototype both types behave the same; the distinction is visible but not functional.

• Select an existing workspace from the left navigation.
• Create a new workspace and choose its type.
• Add knowledge — a file, image, note or link — with no classification required.
• Watch AI organize the content into candidate knowledge.
• Review candidates and confirm or dismiss them.
• Ask the workspace a question, in your own words or from suggestions.
• Read a concise answer with a confidence marker and sources.
• Open a source to see the excerpt and why it supports the answer.
• Browse and filter decisions; resolve an open question and record it as a decision.

• Persistent left navigation with workspace switching and workspace creation
• SMS Initiative demo workspace with Flow A and Flow B mock content
• Overview command center: add knowledge, glance metrics, recent activity
• Four capture methods: upload files, add images, add a note, add a link
• Simulated AI extraction of candidate knowledge on every addition
• Human review of candidates before anything becomes trusted
• Ask this workspace panel with suggested and free-form questions
• Answers with Known / Uncertain / Unknown confidence and clickable sources
• Source detail drawer with excerpt and supporting rationale
• Decisions and Open Questions as filterable first-class objects
• Knowledge tab organized by requirements, approvals, owners, dependencies, risks
• Files, Notes and Team views

• People will add knowledge if capture takes seconds and requires no taxonomy.
• Teams want AI to organize, but want humans to decide what is trusted.
• Evidence and traceability matter more than answer fluency.
• Decisions are the highest-value knowledge object to make durable.
• A workspace scoped to one product or initiative is the right unit of memory.

• Who does the confirming, and does review become a bottleneck?
• How much candidate knowledge is too much before review fatigue sets in?
• Do teams trust an Uncertain answer more than no answer at all?
• Is the product vs. initiative distinction meaningful in daily use?
• How should knowledge move between workspaces when initiatives end?
• What happens when sources conflict — who arbitrates?

• Autonomous agents, workflows, and execution on the team's behalf
• Predictive or anticipatory capabilities
• Real permission and access management
• Real integrations with document, design, or chat systems
• Persistent storage — this prototype keeps everything in the browser session
• Analytics, reporting, and admin tooling