Outer Work: The Knowledge Keeper

Why every company needs someone protecting its memory.

Source: Unknown

While companies are frantically running around figuring out how to ship more and faster with AI, how to create content with AI (maybe you’re even at the stage worrying how to make the content you made less AI slop-y), how to advertise on AI channels (new) where customers are making purchase decisions, how to reduce costs without needing as many humans, somethings been lost. And I can’t stop thinking about it.

Do you remember that book that almost all of us were assigned in our sixth-grade literature syllabus: The Giver by Lois Lowry? A seemingly utopian society that chooses to erase everyone’s memory. They put the context, the history, the color, the pain into one singular person so everyone else can live inside a tidy, frictionless present.

With AI models at everyone’s fingertips, it’s starting to feel a bit like the corporate version of that trade.

In exchange for access to the world’s collective knowledge, companies have stopped tending to their own. The internal truth that actually makes them distinct. It’s the easy path: reach for the same external intelligence everyone else has access to, instead of doing the harder work of building and protecting the intelligence that only you can create.

The truth of The Giver is unmistakable: without memory, life loses meaning.

And so as we go deeper into territory of complete reliance on autonomous tools, agents and realities, I’m encouraging companies to start caring about their memory.

Corporate Memory & Accumulating Knowledge

Companies are drowning in all different types of unstructured information: PDFs, screenshots, emails, decks, looms, Slack messages, internal docs. Some of these haven’t been updated in years (think SOPs and brand guidelines, if you even have those). Some change so frequently, it’s hard to keep up (think regulatory). And that’s just the historical context from the years of internal humans and strategies that form what the company is today. That’s not including the layer of customer data and performance insights that tell interesting and meaningful stories over stretches of time.

The company’s memory, the accumulation of its intelligence, is extremely fragmented. It exists everywhere and nowhere all at once. In documents, in dashboards, in people’s heads. In Slack threads that felt important at the time and impossible to find later.

There is no debating that every company has lots of knowledge. In fact, without it, I’m not sure a business exists at all. The strange thing is that no one is responsible for it. There is no organ that holds it, no function that maintains it. It simply accumulates, fragments, and dies.

Every business has the knowledge. What they’re missing is the steward of it all.

Enter the Knowledge Keeper

The Knowledge Keeper is the person responsible for the company’s memory. They are the connective tissue between what the company has learned and the vision of what it wants to do next. The result of their work provides value to both strategy and execution, machines and people, the internal org and the external customer.

As the recipient of the work nickname “Context Queen,” this sounds like an IDEAL role for someone whose love language is information. But with great knowledge comes great responsibility.

The Knowledge Keeper notices when insights and context get lost between teams, when the information contradicts itself, when AI is hallucinating, and when information is safe to use in customer conversations. For those really curious about this as a real role in their org, I created a generic JD for you at the end of this essay. But first, let’s talk impact of the Knowledge Keeper.

Knowledge for External Use: The More Immediate Need

These are the places where an agent or model is directly communicating with your existing or potential customers. It’s where you have no control over the conversation, which is exactly why it’s more critical to tackle asap.

Right now, companies are experimenting with AI at the edges: using chatbots for customer support, adding chat interfaces to enhanced site searches, testing infinite AI messages in marketing copy, pumping 1:1 personalized emails, and optimizing for AEO/GEO. But model chat conversations are becoming even more embedded into your customer’s decision-making. Shopify recently announced direct checkouts within models like ChatGPT, Gemini, and Copilot. If AI conversations weren’t already influencing commerce, they sure as hell are now.

But here’s the problem, all these “integrations” are shallow. The models pull from whatever data is easily accessible, which is rarely the data that matters most.

The companies with organized knowledge will start to stand out from those without. The gap will compound in customer experience and in revenue. An agent with access to rich, structured, well-maintained knowledge will perform entirely differently than one scraping through fragmented, outdated and insufficient information.

Knowledge for Internal Use: Efficiency, Speed, Quality

I’d be hard pressed if no one at a company has tried using a model to do to work-related tasks. And if they’re doing it even somewhat consistently, I think there comes a time when everyone hits that wall and says, “AI just isn’t there yet.” It’s usually the result of employees creating their own RAG systems to support their function based on… you guessed it… not a knowledge base.

Imagine a new employee asking your knowledge base: “Why did we sunset Product X in 2023?” and getting a nuanced answer with linked PRDs, actual customer feedback and, maybe even, Slack decision threads. Certainly not a hallucination or an “I don’t have access to that information.” Your new employee would be onboarding and decoding while having immediate impact.

Now imagine being able to find all the previous performance tests your company has completed and the insights gained from those tests. You’d be able to have clarity over a testing roadmap and not repeat mistakes of the past.

Hopefully you get the idea, there could be immense impact across nearly almost every function internally in efficiency, speed and quality of work. All of this sounds wonderful but it takes serious documentation and organization to get this up and running, hence the need for the Knowledge Keeper. If you’re a company that has prioritized documentation and infused it as a critical process at the company level (cough, Stripe), you definitely have a 30-second head start.

The Path Forward

I think it’s pretty obvious why this isn’t within every org right now: this is an incredibly large task. It requires many different skill sets and the tools to support aren’t really here yet. Let’s have grace with ourselves without falling behind.

Because the more interesting evolution is what happens when agents start creating memory, not just consuming it. When the customer support agent captures insights from thousands of conversations and feeds them back into the knowledge base, which then informs the product roadmap. When the sales agent notices shifting language patterns and updates the pitch deck. When the knowledge base becomes a living brain that learns as the company learns.

I am aware of the irony here. We are talking about using humans to organize knowledge for machines that will eventually organize knowledge better than any human could. The Knowledge Keeper is a bridge that may not be needed forever, but I do think it will evolve. Even if AI becomes capable of maintaining and curating knowledge autonomously, someone will still need to decide what matters. What the company is trying to be. These are not optimization problems. They are questions of identity and values. They require judgment that is irreducibly human and conscious.

The company that outsources its memory entirely outsources its identity. It becomes whatever the model thinks it should be, shaped by the same knowledge that shapes everyone else. The sameness we’re already seeing is only the beginning.

So then the question becomes: Will you amplify what makes you distinct or dilute it?

Because without memory, it really is all meaningless. And in an AI-saturated world, your memory might be the only real moat you have left.

---

If you know of tools that are helping to tackle this problem thoughtfully, please send them my way. If you can’t tell, I’m very passionate about this.

And if you know of someone that feels just as passionately about this, please consider sending them this piece.

---

Knowledge Keeper Job Description and Company Phases

I have only seen one company post a JD that reflects the role of the Knowledge Keeper and it was by Ramp. Seems like the role is filled now but here is a generic description that you can use if you are considering someone on your team like this.

Role Overview

The Knowledge Keeper is responsible for owning, organizing, and evolving the company’s internal knowledge ecosystem. This role ensures that what the organization knows—about its brand, product, customers, operations, and strategy—is accurate, accessible, and usable across teams and systems.

This person acts as the connective tissue between functions, enabling teams to create, find, and apply knowledge efficiently. They design systems that support both human decision-making and machine use, ensuring knowledge remains a living, trusted asset as the company scales.

Core Responsibilities

1. Own and Maintain the Company’s Knowledge Base

• Serve as the steward of institutional knowledge across the organization

• Define what information should be preserved, updated, or deprecated over time

• Organize and maintain knowledge across key domains, which may include:

  • Brand: mission, feelings to evoke, copy + design guidelines

  • Product: features, RTBs, functions, selling props, formulations, PRDs

  • Tech stack: infrastructure, decision criteria

  • Marketing: stories, calendars, strategies, angles, advertising, sales materials

  • Partners: agreements, co-brand guidelines

  • Personas: pain points, interests, demographics, behaviors

  • Data: what’s working / what’s not, A/B tests, performance insights, customer experiences

  • Vision: product roadmap, product request backlog, seeding what’s to come thoughtfully

• Establish and enforce standards for documentation, taxonomy, metadata, and versioning

• Ensure knowledge is current, accurate, and discoverable

2. Enable Cross-Functional Use of Knowledge

• Partner with teams across product, marketing, engineering, operations, and leadership to understand how knowledge is created and consumed

• Design systems that enable self-service access to information rather than centralized gatekeeping

• Educate teams on how to contribute to and use the knowledge base effectively

• Integrate knowledge systems into existing workflows, tech stack and onboarding processes

• Treat AI tools, models, and agents as users, ensuring content is structured and optimized for both humans and machines

• Define and track KPIs related to knowledge usage, adoption, and impact

3. Audit, Optimize, and Evolve the System

• Continuously evaluate whether knowledge systems are functioning as intended

• Identify gaps, redundancies, outdated information, and breakdowns in usage

• Sunset irrelevant or low-value documentation thoughtfully

• Stay informed on new tools, platforms, and practices that could improve knowledge management

• Iterate on systems, processes, and governance as the organization grows and changes

When to Implement

Early-stage startups (pre-Series A): The founder is the default Knowledge Keeper. They have the context, make the decisions, and hold the vision. You probably don’t need a dedicated role yet but it’s critical the founder is setting this up from the jump.

Growth stage (Series A/B): This is the inflection point. Teams, information and insights are fragmenting. New people are joining who don’t have the context. Different functions are making contradictory decisions. You need this role.

Established companies: You needed this role yesterday. Your knowledge is already fragmented across a decade of different teams, tools, and strategies. The longer you wait, the harder it becomes.

Sector considerations: Tech companies that built documentation-first cultures (like Stripe) have a structural advantage here. E-commerce and retail brands that haven’t prioritized this are sitting on a massive opportunity, and also a vulnerability. But virtually every consumer sector could benefit from this role.