Perplexity Is Building a Work Platform

Perplexity launched Computer in late February — a $200/month agentic platform for Max subscribers that orchestrates 19 AI models across complex, asynchronous workflows. The launch was notable. The follow-up is more interesting.

This week, Perplexity announced Skills, a new feature layer inside Computer that lets users build and store reusable workflow instructions as markdown-based specification files. The platform also includes a library of generic skills Perplexity provides directly. When a task starts, relevant skills load automatically based on the query. Multiple skills can combine to handle complex work. The result is a system where a power user defines how recurring task types should be handled once, and Computer executes to that specification consistently — without re-prompting from scratch every time.

Separately, code researchers at TestingCatalog found evidence of a feature currently labeled "Final Pass" in development — a document review mode that appears designed to offer structured, comprehensive analysis of uploaded content, with financial reports and legal filings as the likely primary use cases. No release timeline has been indicated, and the name is almost certainly a placeholder. The intent, however, is clear: a dedicated post-production review layer sitting on top of Computer's existing document handling.

Why Skills Is More Significant Than It Looks

The Skills architecture is not a new concept. Claude Code, Codex, and other developer-facing agentic platforms have offered reusable instruction sets in various forms. What makes Perplexity's implementation worth watching is the audience it's aimed at and the platform it sits on.

Perplexity's core user base skews toward analysts, researchers, and finance professionals — people who repeatedly run variations of the same complex information workflows. Pulling earnings call transcripts. Comparing regulatory filings. Synthesizing research across multiple sources into structured outputs. These are tasks where the friction isn't the AI's capability — it's the overhead of re-establishing context and workflow parameters every single session.

Skills address that friction directly. Define the workflow once, in plain markdown. Load it automatically when relevant. Combine multiple skills for compound tasks. That's a meaningful reduction in the cognitive overhead of working with an agentic system at a professional scale, and it's the kind of feature that turns occasional users into habitual ones.

The distinction between user-created skills and Perplexity-provided generic skills is also worth noting. The generic library functions as an onramp for users who don't want to build from scratch, while the custom layer allows professionals to encode proprietary workflows that reflect their specific analytical frameworks and output requirements. That two-tier architecture is well-designed for a professional audience with varying levels of technical comfort.

What Final Pass Signals About Perplexity's Direction

The Final Pass feature, even at the early evidence stage, tells a clear story about where Perplexity is trying to position Computer relative to existing tools.

Document review — the structured, systematic analysis of a completed document against a set of criteria — is a distinct task from document generation. Most AI platforms handle the latter reasonably well and the former poorly. A purpose-built review layer that can evaluate a financial model for internal consistency, flag unusual language in a legal filing, or assess a research report against stated methodology requirements would fill a gap that currently sends professionals back to manual review processes or separate specialized tools.

For analysts and researchers using AI in their workflows, a reliable document review layer is arguably more valuable than faster generation. Generation speed has improved dramatically across the industry. Review quality — the ability to catch what's wrong, inconsistent, or missing — remains a genuine weak point.

The Broader Strategic Picture

Perplexity's trajectory from AI search tool to full-service digital work platform is now explicit rather than implied. Computer is the infrastructure. Skills is the personalization layer. Final Pass, if it delivers on its apparent intent, is the quality control layer. The three together describe something closer to a professional operating environment than a chatbot with a better search index.

The competition is real and well-resourced. Microsoft Copilot, Google Workspace AI, and Claude's own agentic capabilities are all targeting the same professional workflow market. Perplexity's differentiation has historically been speed and source transparency. Whether Skills and the features following it are sufficient to compete with platforms that have deeper enterprise integration and larger distribution networks is a genuinely open question.

What Perplexity has is a focused audience, a clear product direction, and a pricing model — $200/month for Max — that self-selects for professionals who treat the platform as infrastructure rather than a novelty. That's a defensible starting position. Whether they can build a full work platform on top of it before larger competitors close the feature gap is what the next 12 months will answer.

For marketing and growth teams evaluating AI work platforms, the Skills release is a useful signal: the platforms worth watching are the ones building for repeatability and professional workflow integration, not just raw capability on individual tasks. The question for any platform isn't what it can do once. It's what it can consistently do at scale, within the workflows your team actually runs.

If you want help evaluating which AI platforms belong in your marketing and growth stack — and which ones are still finding their footing — Winsome Marketing's strategists can help you make decisions grounded in how these tools actually perform in professional use.