Product Strategy — PRISM

Product Definition

PRISM is an enterprise SaaS application for running Algolia prospect research and search audits at company scale. It should make the current expert-operated skill pipeline usable by a broader field organization while preserving the depth, evidence discipline, and report quality that made the original workflow valuable.

Why SaaS

The skill system is powerful but operationally fragile:

• It depends on local paths and environment variables.
• It produces file-based state rather than first-class product state.
• It requires expert orchestration judgment.
• It has limited user/account authorization.
• It is hard to monitor, retry, compare, govern, and roll out broadly.

PRISM should convert it into:

• deterministic execution
• durable storage
• typed contracts
• workflow visibility
• repeatable validation
• authenticated and authorized access
• enterprise-ready audit history
• reusable intelligence across accounts and verticals

UX Direction

The next product direction is chat-first/copilot-first, not traditional navigation-first.

The copilot should be the user's primary interface for:

• starting an audit
• asking what PRISM knows about an account
• seeing which modules are running or blocked
• drilling into evidence
• asking "why does this matter?"
• generating AE-facing call prep
• refining deliverables
• deciding what to publish/share

The UI should still expose structured state, but the primary mental model is not "click through 20 tabs." It is "work with the PRISM copilot over a grounded audit state."

Product Boundary

PRISM is not only a report generator. It is a research and audit operating system:

• Research collection
• Browser evidence capture
• Scoring
• Business case synthesis
• Sales play generation
• ABX package generation
• Factcheck and gate verdict
• Copilot explanation and follow-through

Future Components

Scout

Scout should be incorporated as a crawl and WAF/JS bypass layer where current collection is weak, especially:

• blocked corporate pages
• JS-rendered career or newsroom pages
• competitor newsroom enrichment
• IR/PDF discovery
• pages where BuiltWith or SimilarWeb are stale/incomplete

Scout should not replace all data collection. It should be a targeted crawl substrate with token/data-volume controls.

Discovery OS

Discovery OS should become the translation layer between audit findings and rep behavior.

It should transform findings into:

• calibrated hypotheses
• permissioned insight moves
• first-call openers
• objection pre-maps
• branch logic based on buyer reaction
• MEDDPICC field updates
• next-meeting hooks

This belongs after the baseline audit state is reliable.

Opinionated Product Risk

Do not design the copilot as a generic chat wrapper over documents. The differentiator is that the copilot understands the audit workflow, module state, evidence tiers, source provenance, and sales motion.

The copilot should be able to say:

• "This claim is websearch-only, so don't put it in the leave-behind."
• "The financial module is degraded because SimilarWeb/Yahoo data is missing."
• "This is a Golden Angle because a competitor has verified Algolia proof."
• "This hypothesis is high-impact but high-risk if wrong."