The KnowledgeOps Manifesto
The most valuable thing your firm owns cannot be found on your balance sheet. It lives in your senior partners' judgment, your institutional memory of what works, the pattern recognition your best people developed over fifteen years of client work. For most of your history, that knowledge has been valuable but fragile — dependent on individuals, difficult to transfer, impossible to scale.
AI has made knowledge deployable for the first time. The firms that figure this out will do things their competitors cannot: scale without adding headcount, deliver consistency without losing nuance, and build services that are genuinely defensible. The firms that don't will find their growth permanently capped by the oldest constraint in professional services: you can only grow as fast as you can hire and train.
KnowledgeOps is the discipline for getting this right. And like every discipline worth taking seriously, it didn't appear from nowhere — it follows a pattern we've seen before.
History Repeats
Two decades ago, software delivery was organized around heroic individuals. The person who knew the deployment scripts. The developer who held the architecture in their head. The operations engineer who got called at 2am because no one else knew how the system worked.
It wasn't sustainable — and everyone knew it. But the problem didn't have a name, and without a name it didn't have a solution.
Then a set of practitioners began describing a different way: shared ownership, automated pipelines, continuous delivery, infrastructure as code. They called it DevOps. It transformed an industry — not because the technology was new, but because the discipline finally matched the complexity of the work.
A decade later, machine learning hit the same wall. Data scientists were running experiments no one could reproduce. Models were trained by individuals who then left. The name that emerged — MLOps — brought the same discipline to the same kind of chaos.
We are at that same turning point. But this time the asset is not just code or data.
It is knowledge at scale.
The expertise in your senior partner's head. The judgment call that closes a difficult client situation. The institutional memory of what worked, what failed, and why. The pattern recognition that took fifteen years to develop.
That discipline is KnowledgeOps.
The Problem
Your firm's most valuable knowledge moves informally — through partner review, shadowing, and the memory of whoever has seen this before. A few people use AI well. The firm as a whole doesn't get smarter. You know the pattern:
- You essentially sell knowledge and expertise, but your best attempt at internal knowledge sharing and cross-pollination is quarterly off-sites.
- John leaves and a significant amount of know-how walks out with him.
- If you're ahead of the curve, you've onboarded some AI tools to tame this — but other than small local wins, the work has the same bottlenecks still.
This is the gap KnowledgeOps is meant to close.
The Solution: The SCALE Framework
For the first time, the expertise that makes your firm valuable can scale beyond the people who hold it — not to do the same work with fewer people, but to do work that wasn't previously possible. These are the five convictions behind KnowledgeOps.
| Choose this | Over this | |||
|---|---|---|---|---|
| S | Systematize | A firm that grows faster than its headcount | over | a firm limited by who it can hire |
| C | Compound | A firm that gets smarter with every engagement | over | a firm that relearns the same lessons |
| A | Accelerate | Standards that let the whole firm move fast | over | improvisation that creates risk at scale |
| L | Liberate | Partners who grow their practice | over | partners who protect their bandwidth |
| E | Expand | New revenue that didn't exist before | over | incremental improvements to what already works |
Practices
These values are achieved by following well-designed and rigorous processes and tooling to capture and evolve knowledge across the firm.
Acquire & Shape Pull knowledge into your system deliberately, then structure it before it can be used. Raw capture isn't enough — the real work is making implicit expertise explicit: defining the context, surfacing the reasoning, naming the patterns. This is where senior judgment becomes organizational infrastructure.
Example: Instead of letting everyone run their own separate AI note-taker, roll out a firm-wide one and ensure all transcripts are summarized and committed to a knowledge base for further analysis and usage.
Craft & Review Turn shaped knowledge into reusable artifacts — frameworks, workflows, templates, decision guides — then test them before they spread. The goal is tools that live in how work gets done, not documents that live in a folder. Review is what separates a knowledge base from a knowledge system.
Example: Implement human-in-the-loop processes to examine meeting transcripts and craft prompts for future client meeting prep, workflows and templates for client call to proposal processes, or reusable prompts for junior staff role-play client engagements. All of these feed into the firm's knowledge base for distribution and usage.
Approve & Integrate A human makes the final call. Then approved knowledge gets embedded — not as documentation to consult, but as infrastructure that runs in the tools and workflows people use daily. This is the step that turns a knowledge project into a knowledge operating system.
Example: A group designated by the firm reviews all staged prompts, workflows, and proposed integrations and, after testing, approves them for deployment. These are rolled out as ready-to-use artifacts wherever staff works — ChatGPT, Claude, or custom agents built by the firm.
Operate & Monitor Staff uses the system in live engagements. Areas of improvement are surfaced: What's compounding? What's drifting? What new patterns are emerging from your most recent work? This is where KnowledgeOps earns its defining property: it is a closed loop. Monitoring feeds directly back into Acquire, making the firm measurably smarter with every engagement — not through training days, but through the work itself.
Example: A junior consultant and a senior partner prep for a client meeting using the same prompts, then share their feedback on the prompts to a centralized inbox. That inbox is integrated into the knowledge base and accumulates feedback for the next round of improvements.
The Journey
Firms don't arrive at KnowledgeOps all at once. They move through five predictable stages — and most are stuck earlier than they think:
- Stage 1 — Individual Experimentation: A few motivated people use AI well, but the learning stays local. No coordination, no shared output, no firm-wide benefit.
- Stage 2 — Standardization & Governance: Shared prompts, frameworks, and guardrails replace one-off improvisation. The firm starts accumulating knowledge instead of losing it.
- Stage 3 — Firm-wide Enablement: Approved knowledge assets become accessible across the organization. Junior staff can access senior judgment without waiting for a senior person.
- Stage 4 — Integrated Operations: Knowledge systems are embedded in the live workflows through which client work is actually delivered. AI agents handle the repetition; humans handle the judgment.
- Stage 5 — Product & Service Innovation: Systematized knowledge becomes a competitive asset that can be packaged, licensed, or sold. New revenue that didn't exist before.
Frequently Asked Questions
These are the questions we hear most often. They are not technical questions — they are strategic ones.
"We already have people using AI tools. Why do we need anything else?"
Yes — and it's also the most common stall point. Individual experimentation is valuable proof. It is not a strategy. Without standardization, you have 20 different workflows, no shared learning, and mounting inconsistency. The question isn't whether your people are using AI. It's whether the firm is.
"We sent out some guidelines about AI use. That helps us make sure everyone uses AI consistently and safely, right?"
Consistency and safety do not come from a policy document. It is a consequence of systems built around your knowledge base: shared prompts, approved workflows, defined guardrails, and clear ownership of what gets reviewed before it reaches clients. Standards have to live in systems, not in PDFs.
"We've tried knowledge management before. It didn't stick. Why is this different?"
Because the previous attempts produced documents, not infrastructure. A SharePoint full of PDFs is not a knowledge system — it's a filing cabinet that nobody opens. The difference with KnowledgeOps is that the knowledge base is embedded in the tools people already use to do their work. It's not something you consult separately; it's something that runs in the background of how work gets done. The other difference is the loop: previous knowledge management was a capture exercise with no feedback mechanism. KnowledgeOps is designed to improve continuously, which means it gets more useful over time instead of going stale.
"We have early AI adopters, but most of our team hasn't started. How do we change that?"
This is not an adoption problem — it's a bandwidth and infrastructure problem. Your team isn't resistant. They're at 150% capacity with client work and won't find time to experiment. The solution is to bring systematized workflows to them, not to ask them to build their own. Make it easier to use the system than to work around it.
"How do we integrate AI into how client work gets done — without risking them thinking 'I could have gotten the same info from ChatGPT'?"
This is where the operating model changes. AI amplifies the structure that you introduce to it. If AI use in client-facing functions is ad hoc, sporadic, and based on disorganized information, then AI amplifies that. If you have a well-organized, self-improving knowledge base powering your internal and client-facing workflows, then AI amplifies that rigor — and the opportunity for growth that comes with it.
"Will use of AI change what makes us valuable to clients?"
This is the right question, and most firms ask it too late. AI that erodes judgment, homogenizes output, and removes senior expertise is a threat. AI that removes low-value work, frees senior attention, and makes firm knowledge more consistent and accessible is a competitive advantage. KnowledgeOps is a framework for doing the second thing deliberately.
"How does use of AI eventually lead to new services or new revenue?"
In the KnowledgeOps era, your firm is no longer a collection of people — it's a proprietary operating system. That reframe changes everything: your valuation, your scalability, what you can offer clients, and what competitors can replicate. The firms that reach this stage discover that their systematized knowledge base is itself a product — one they spent years building and competitors cannot buy. Their operating system becomes a product in its own right: tools, training, licensing, or entirely new service lines that didn't exist when they started.
"Would AI help us cut headcount or reduce costs?"
KnowledgeOps is a growth framework. If you're looking for a 10% efficiency gain to trim your cost base, you don't need a new discipline — you need better processes. KnowledgeOps is for firms that want to grow what's possible: more clients served at the same quality, new services that didn't exist before, revenue that isn't capped by how many people you can hire. If the prize you're after is a bigger pie, this is for you.
"How do we protect our IP or client data while gaining efficiency from using AI?"
This is the right question to ask first. The answer depends entirely on how the system is designed and how sensitive those are. Your IP and client data should stay in systems your firm controls, with access policies your firm defines, as much as possible. Knowledge-Ops has a clear take here: reduce the friction of taking safe actions. For example:
- Provide employees with business / enterprise accounts and make sure data controls set to prevent data training on your usage.
- Ensure your IT and legal teams are comfortable with data usage and retention policies your AI tool uses and communicate that to the staff.
- If your data is very sensitive, provide the option of using a privately deployed AI tool ideally with controls for redacting PII / PHI or other sensitive information.
- If the level of sensitivity is very high, consider hosting open models using on-prem servers so that no data leaves your premises.
The same governance you apply to client files applies here. Done correctly, a well-structured knowledge base is actually more secure than the current reality — where sensitive context lives in individuals' inboxes, personal notes, and memory.
"Who owns this internally? Do we need to hire someone?"
Someone(s) needs to own it — but it doesn't have to be a new hire. The firms that do this well designate a knowledge steward team: a group representing various functions and needs from across the firm. The team will be charge of getting educated and driving the majority of AI tooling work necessary for knowledge-ops to be rolled out effectively. The team would also work with various technical and business stakeholders to help with identifying new opportunities for AI intervention, writing prompts, and integrating into new tools.
"How long does this take, and how disruptive is the transition?"
This is an ongoing process. The firms that struggle are the ones that try to build everything before deploying anything. Start narrow, deploy fast, expand from proof. There are always low hanging fruits like AI helping with writing and language tasks. Those will get people started and excited while you build out the more sophisticated workflows and processes.
"How do we know if it's working?"
A few signals are unambiguous. Junior staff output quality rises without additional senior review time. Onboarding time drops. Senior partners report spending less time on work that shouldn't require them. New client engagements start faster because the relevant context already exists.
"Where do we start using AI effectively?"
Not with tools. Not with a strategy deck. With one question: where is the most valuable knowledge in your firm, and where does it most often fail to reach the people who need it? Start there.