Upstart Takes On One of AI's Biggest Problems

In today's AI-driven world, more information is generated daily. The average knowledge worker already spends roughly 2.5 hours a day, nearly a third of the workday, searching for information. That is in addition to time spent dealing with confident, but unreliable AI generated answers. Cognatum launched today as the first Intelligence Management System for regulated enterprises and the solution to that problem.  
 
Every AI system a company uses today needs a source of information. Often times, it is data from the internet. In more advanced systems designed for enterprise use, it may be that company’s documents, policies, and data. The AI system answers questions as if it already knows which content is accurate, current, and approved. It generates answers even when they are not correct or based on usable information. IDC reports that the average knowledge worker wastes 2.5 hours daily on these tasks. That is nearly a third of the workday spent searching for information or trying to verify AI results, at a cost of more than $5 million a year for an organization of a thousand people.  

In 2018, the International Organization for Standardization published ISO 30401 and, for the first time, gave the world a formal, standardized system for enterprises to manage their company’s knowledge and information assets. This is a knowledge management system (KMS). The ISO standard set requirements for how a company acquires, applies, retains, and retires what it knows. Regulated industries, including banking, insurance, life sciences, aerospace, healthcare, quickly adopted it as the reference. 

With the advent of AI, the standard's core assumption became obsolete. 

ISO 30401 was written for a time when only people read documents. Today, the new primary consumer of your company's knowledge is an AI machine. 

Right now, inside a Fortune 500 company somewhere, an AI just gave someone an answer. No one knows who approved that answer. No one knows when it was current. No one knows where it came from. And if a regulator asks tomorrow, someone has to explain it. That is a problem for most businesses.  

It is not, however, just an AI problem. It is a knowledge management problem and it is an enterprise problem. Existing KMS systems were not designed to manage AI-driven knowledge. 

Cognatum is the Intelligence Management System a category built on ISO 30401 foundations and extended for the AI age that has come next. 
 
Cognatum's bidirectional connectors reach more than 50 enterprise systems, including SharePoint, Box, Salesforce, HubSpot, Snowflake, Confluence, Microsoft Teams, Jira, ServiceNow, and Workday. This enables users to work the way that they already work while providing them with access to the knowledge that they need.  

Maintenance as a Loop, not a Project. 

This is the shift underneath the shift. When the consumer of your knowledge was a person, a knowledge base could be a project stood up, populated, forgotten. When the consumer is a machine, running thousands of times an hour against the same knowledge base, the whole ISO 30401 lifecycle has to run continuously, KMS-wide, or the answers rot in place. 

Cognatum runs that lifecycle as a continuous loop: eight steps, four phases, one governing rule. 

• Intake · Capture and Structure.  Ingest from every source, no re-authoring. Normalize into question, answer, context, source. AI finds and drafts the entries; humans need never touch a form. 

• Refine · Clean.  Update, dedupe, reconcile, archive continuously, across the whole corpus. AI flags duplicates, conflicts, and staleness in real time. This is the step that makes an AI answer defensible. 

• Elevate · Enrich and Improve.  Tags, relationships, metadata; a living knowledge graph. Rewrites proposed from what usage actually shows isn't landing. AI builds the graph; AI proposes the improvements. 

• Govern & Activate · Approve, Deploy, Reuse.  A single human sign-off in Microsoft Teams - logged who, what, when. Approved knowledge made live everywhere at once, written back to the source systems. Served identically to people and AI. AI queues and summarizes, publishes, writes back - but never approves so humans remain in the loop and data remains governed by SMEs.  

AI runs the loop continuously. Humans hold the gate. 

Above the Repositories You Already Have 

The loop runs above the repositories a company already uses: SharePoint, Box, Salesforce, HubSpot, Snowflake, Confluence, departmental drives. Nothing migrates. Nothing moves. And every answer ever given can be replayed exactly as it stood on the day it was served not just what it says today, but also what it said then. The whole system runs as a single-tenant appliance, with an optional air-gapped deployment. No cross-tenant data path. Full data sovereignty. 

From the Founders 

"Every enterprise AI conversation eventually arrives at the same question, and it isn't about the model," said Perry Robinson, CEO of Cognatum, who previously co-founded unicorn Medable, Inc. "It's about finding the answer and knowing the answer is right. That's a knowledge management problem and knowledge management stopped being a project the day the primary consumer became a machine. It's a loop now. The AI runs it. Humans hold the gate. Get that right, and finding and proving the answer becomes a feature." 

Cognatum was built by Scott Getchel, Head of AI, who started the project to keep track of his own work before it quietly spread across the company. "What people tell me in every conversation is that their knowledge lives in fifteen different systems and they need fifteen logins to get at it," Getchel said. "The answer isn't a sixteenth system. It's one loop, running above all of them, that keeps track of what's current, who signed off on it, and where it came from. Once that loop is running, the humans and the AI both stop guessing." 

Common Roots. Collective Intelligence.