Cognita Journal: July 2026
To get our newsletter in your inbox, subscribe.
On July 4, 2024, I sat in my recliner by the big front window of our Texas Hill Country cabin and filed the documents to establish Cognita Knowledge Management, LLC.
The company I had worked for was circling the drain. Many of my U.S.-based co-workers had been laid off and I expected to get an email any day telling me that the company had filed bankruptcy and my job there would end.
Ernest Hemingway said it best: bankruptcy happens “slowly, then all at once.” The writing had been on the wall for some time, and I had been methodically applying for other knowledge management positions, but to no avail. I had to do something different.
So I declared my independence, put together a business plan and used the last couple of months of employment to get the pieces in place.
(By the way, the choice of July 4 as the filing date was coincidental; I didn’t know the Texas Secretary of State’s office would actually process the documentation and give me July 4, 2024 as the date of formation. Independence day indeed!)
One of the bits of advice I got when I first started talking to other people about my new venture was to give myself two years to get established. It would take that long to really determine whether my venture had legs, or was just a way to waste time.
Well, here we are, two years later. Lately, I’ve been thinking about what worked, what didn’t work, where I am at this point, and where it seems I’m going. My verdict: yeah, this thing has legs.
I’ve published a book, been on podcasts, been invited to speak on webinars and in-person conferences and most importantly, helped clients do knowledge management better, saving them money and grief along the way.
I’m currently working with two clients. When I think about the day-to day work of Cognita KM and the commitments I have outside of work, I’m busier than ever. But I’m okay with being a little busier. I’ve run the numbers, and as we’re heading into the second half of the year, I’ve figured out that I can handle one additional client without cutting in my ability to deliver quality results.
(If you’ve been thinking about improving your KM operation, building a KM strategy or just doing an assessment, get in touch with me soon to get something on the books before the Q4 rush.)
One last thing before we get back to our regularly scheduled KM programming: none of this would have happened without the support of a ton of folks, including Mark Brody of Brohawk Consulting, Brad Shaw at livepro, Fred Stacey and Darren Prine at Cloud Tech Gurus, Stephen Pappas at eGain, Mitch Pautz and Sam Chan at UCSF IT, the fine folks at CX Accelerator and Support Driven, and last but certainly not least, my wife Cheryl and my family near and far.
Going to Chicago
I wasn’t able to make it to Contact Center Week in Las Vegas due to personal commitments (one of them being a Rush show that was rescheduled). My next conference appearance will be the Support Driven Expo in Chicago.
This time, I’ll be on the stage as a speaker instead of sitting in the audience. I’m excited to talk about how customer support organizations can tackle the specter of shadow knowledge.
The Illusion of the Automated Org Chart
You’ve no doubt seen the countless stories (or maybe even encountered them in your job): companies that treat AI as a glorified headcount reduction tool. And you’ve probably also seen the result: seduced by promises of immediate operational savings, a company decides to dismantle their human support structures, which destabilizes the very asset that makes AI function—their organizational memory.
A recent article in Fortune highlights shows that 80% of business executives who have piloted autonomous technologies reported subsequent workforce reductions. Even worse, they made these cuts regardless of whether the technology was actually generating clear financial returns. As the Gartner report quoted in Fortune points out, chasing organizational value solely through headcount reduction is a short-sighted strategy that leads down a path of severely limited returns.
When you lay off experienced professionals under the assumption that an LLM can seamlessly replicate their subject matter expertise, you don't actually eliminate the labor. You merely transform documented expertise into undocumented, fragmented shadow knowledge.
The reality is that agentic AI isn’t some sort of independent program running on a server. In reality. it’s a bunch of processes and subprocesses working together. It’s an end-to-end operational workflow. If your underlying knowledge, ingestion pipelines, re-ranking, and context layers are brittle and unoptimized, your fancy AI bot will encounter terminal delays and hallucinate under high concurrent traffic. AI can’t magically retrieve what’s been forgotten, and it can’t reason over a broken Sharepoint site.
To build a resilient enterprise, stop looking at AI as a replacement for human intellect. Build those AI systems, but only after you have deployed frameworks (like our MVP KM™ approach) that treat knowledge as an evolving, format-first infrastructure capable of fueling both human insight and machine intelligence.
Breaking Silos via the Open Knowledge Format (OKF)
One of the biggest hurdles I’ve consistently encountered in my KM career is the fragmented nature of corporate knowledge. It’s been trapped inside proprietary content management systems, isolated intranet wikis, scattered shared drives, OneNote, stickies, chat threads and agent brains.
Getting from there to a “single point of truth” has been my goal in every KM initiative I’ve worked on. And now there is something that could help get down that road faster. My former employer Google Cloud has introduced the Open Knowledge Format: a pragmatic, vendor-neutral shift away from heavy, centralized "knowledge platforms" and toward a lightweight, decentralized architecture.
OKF relies on flat Markdown files and simple YAML frontmatter to act like a lingua franca that separates knowledge content from restrictive tool ecosystems. (If you don’t know what YAML is, don’t worry, I didn’t either. According to Wikipedia, it’s a “human-readable data serialization language. It is commonly used for configuration files and in applications where data is being stored or transmitted.”
Why is this a big deal? Because it’s a first stab at creating a common knowledge format that can be used by AI and humans alike across systems.
Knowledge as Code: Because an OKF bundle lives as flat files within a directory structure, it integrates corporate documentation directly into engineering and development workflows. This allows organizations to apply software engineering rigor—such as standard Git workflows, pull requests, peer reviews, and seamless rollbacks—to core business assets.
Curing "Wiki Rot" via Division of Labor: Traditional corporate wikis fail because human maintenance is tedious, leaving dependencies and cross-references to decay over time. The OKF explicitly leverages an "LLM-wiki" pattern: automated AI agents handle the bookkeeping—crawling data schemas and updating links—while humans elevate to high-level curators and editors. This is where approaches like KCS® can shine.
Slashing RAG Infrastructure Costs: Enterprises currently spend massive development cycles building custom retrieval pipelines (RAG) to assemble context from fragmented legacy systems. Because OKF pre-structures data relationships into a standardized graph using explicit Markdown cross-links, AI agents can traverse and reason over a repository exponentially faster, bypassing bespoke pipeline costs.
Metrics That Matter: From "Containment" to Frontline Force Multipliers
If your company is still measuring the success of its knowledge management strategy by looking at "Bot Containment" or "Deflection Rates," you’re optimizing for the wrong outcome. At least, that’s according to Varun Sharma’s article recently published by HDI Service and Support World.
Most CX pros know that “containment” is an obsolete approach that damages customer experience. Trapping a frustrated user inside a circular, unhelpful chatbot loop might satisfy an internal deflection dashboard, but it ultimately accelerates downstream customer churn. Forward-thinking organizations are transitioning to Verified Resolution Rates (VRR), a closed-loop KPI that measures absolute goal completion and dynamically turns real-time chat interactions into fresh, self-service knowledge.
Even with AI in the picture, 20% to 30% of all technical tickets reach Tier 2 support or higher. These “escalations” or “elevations” are rarely a resource issue. Instead, they’re a sign that you have critical knowledge gaps, poor information access, or ineffective KM frameworks.
When frontline staff can’t easily get high-quality, embedded documentation, they’re forced to hand off tickets, multiplying labor costs and inflating Mean Time to Resolution (MTTR). By embedding accurate, real-time context via knowledge directly at the Level 1 support tier, organizations can dramatically reduce escalation dependency, maximize existing personnel resources, and elevate customer happiness.
