Your workforce is the largest unproductized asset in your business.

Roughly seventy percent of what an organization actually knows is in the heads of the people doing the work. The other thirty percent is in the documents, the dashboards, the wikis, the CRM. Knowledge management products serve the thirty percent that was already written down. The seventy percent that was never written down is the asset this page is about.

The asset has commercial value. A beauty advisor at a large retail counter knows which brand a customer almost picked, why the customer switched, and which ingredient is trending months before social listening catches it. A senior consultant at a large firm knows what actually works on an SAP deployment, as opposed to what is written in the playbook. A pharmacist on a chain floor knows which drug interactions get flagged and which patient concerns recur. None of this reaches the corporate center on a useful cadence. The knowledge dies at the end of every shift and is never compounded.

Knowledge as a Service is the category that converts that asset into a continuously refreshed intelligence layer. The unit of input is a fifteen to forty-five minute structured AI-guided interview with the human who already holds the knowledge. The unit of output is a queryable layer that distributes into three endpoints: the client's own internal dashboard, a scoped brand-partner API, and an agent retrieval feed for MCP-compliant clients. Internal decision, downstream revenue, and AI grounding from one corpus, refreshed weekly.

Six categories of tool already exist around the workforce. Each one is built for a layer above or below the one KAAS targets, and each one fails at this layer for a specific reason.

The moat is the integration of three stages into one supply chain. Any single stage already has incumbents. Capture alone is what Outset sells. Codify alone is what Qualtrics sells. Monetize alone is what data brokers sell. Remove any one stage and the economics, the defensibility, and the strategic positioning all collapse. The chain is what creates the category.

When the output is a deliverable (report, dashboard, study) the client's mental model is “tool I bought.” When the output is infrastructure with three endpoints generating a P&L line, the client's mental model is “business unit I run.” The shift moves three things at once. Executive sponsorship moves from VP-level to C-level because business units report to C-level. Budget moves from a departmental research line to a strategic business-unit line. Renewal stops being a question, because shutting down the layer means shutting down a revenue line.

The pricing tells the same story. A report prices in the low six figures. A SaaS seat prices per user per month. A data layer that monetizes brand partners prices against the downstream revenue it enables, with an enterprise floor. In a vertical with a multi-billion dollar adjacent AI market, capturing even a small percentage of that adjacent spend lands the deal in a band that is unreachable from any pricing model built around tools, seats, or reports. It works only because the buyer is purchasing infrastructure.

The commercial chain (Gather, Codify, Monetize) needs a technical chain underneath it that holds at a hundred conversations and at the scale of a national workforce. The five-layer technical moat lives on platform. The synthesized version a buyer needs to know on this page is below.

Every fact in the corpus carries three independent timestamps. learned_at is when the system observed it. valid_at is when the fact became true in the world. invalid_at is when the fact stops being true, set explicitly by a contradicting conversation or by half-life decay if no refresh confirms it. The bi-temporal model lets a brand partner replay last May's state of the corpus against today's and get the correct answer for each. That alone is what makes the layer underwritable in a contract.

Every fact also carries a sparse segment vector: a projection over the segmentation axes that matter for the vertical (tier, geography, store format, customer segment, brand partner, demographic). A brand partner's API key projects onto a specific brand_partner value and surfaces only the facts whose segment vector intersects that projection. Every other partner gets the parallel scope. The same corpus serves the entire set of brand partners on the shelf. The data path is the access control. An unauthorized scope sees nothing, and has no signal that anything was hidden. The security model behind that is the same one running on every Trust & Security page on this site.

The agent-facing contract surface is an MCP server. The server exposes a small tool set: query_facts, traverse_relations, pivot_dimension, replay_as_of, cite. Claude, Copilot, and any MCP-compliant client consume the layer through this interface. The session ticket carries the security scope and the scope cascades into which tools the LLM can see, which graph nodes it can traverse, which fact rows it can read, and which verbatim chunks resolve when it cites. Multi-tenant retrieval with per-segment scope on one corpus is a structurally different shape from per-user agent memory. That difference is the next twelve months of category IP.

The buyer is the COO, the CFO, or the Chief Strategy Officer. The conversation is “new revenue line,” never “better employee insight” and never “better customer research.” HR is the wrong room: HR reframes the project as engagement, which destroys the executive sponsorship and the pricing. Research is the wrong room: research reframes the deliverable as a study, which collapses the third endpoint. The COO is the right room because the COO is the only operator with authority to convert the layer into a brand-partner API.

We refuse a few things on principle. Military applications. Surveillance of named individuals. Re-identification of respondents. Selling raw transcripts to third parties. Training third-party LLMs on customer data. Synthetic-respondent augmentation passed off as real capture. The architectural enforcement behind each refusal is on the refusal protocol page.

Beauty retail is the first vertical. The shape of the opportunity is tens of thousands of advisors at the counter across many markets, with a brand-partner ecosystem already paying the retailer for shelf space. The first KAAS client in that vertical arrives with a captive distribution layer. The adjacent AI-in-beauty market is multi-billion dollar and growing at high double-digit CAGR. Capturing a small percentage of that adjacent spend frames the deal in the infrastructure band rather than the tooling band.

The same playbook runs in consulting (senior consultant benches at the major firms), restaurants (multi-brand operators with thousands of crew), pharmacy (large chain networks of hundreds of thousands of pharmacists), legal services, financial services, and industrial field operations. The total addressable market is tens of billions in current research, R&D, and consulting spend, none of which captures the organic brain layer KAAS captures.

Specifics on the vertical sequence, the wedge sizing, and the adjacent landscape are on the market page. The competitive analysis of each adjacent category is on the category map.

Every other firm in the broader category starts with data already in hand and ends with insight. KAAS starts with humans and ends with infrastructure. The supply chain runs from a trained worker through a structured AI-guided interview, into a codified intelligence layer, and out through three monetization endpoints. That inversion is the entire business.

The category is Knowledge as a Service. The acronym is KAAS. The first vertical is beauty retail. The architecture and the security model are the same chassis the rest of the site documents. The next page you should read depends on the question you arrived with. If you want to see where each adjacent category falls short, read the category map. If you want the market sizing, read the market page. If you want the vocabulary, read the glossary.