The End of the Standardization Bargain

Decoupling Scale from Uniformity in the Age of AI

For most of the last century, sameness was the price of scale: if you wanted to deliver a product or service to thousands of people, you had to standardize it. Litigators didn’t draft custom interrogatories or complaints from scratch for routine disputes. They used master form sets, sending fify standard questions where only ten applied, because there was no incentive to tailor every set. Universities didn’t construct a distinct curriculum for every student. Law firms hired associate classes, organized lawyers into practice groups, established billable-hour structures, created standard forms, and developed repeatable workflows. I think of it as the standardization bargain: we accepted uniformity because uniformity made scale economically possible.

That old arrangement, built around mass production and uniform delivery, is breaking down under pressure from software networks and frontier AI. What was once an economic necessity is becoming an operational choice.

The fundamental strategic question facing executives and managing partners is no longer about mass automation, but about customization: What happens when AI makes nonstandard, highly personalized workflows as cheap to manage as standardized ones? For decades, we treated scale and standardization as identical concepts. AI is decoupling them. And when you decouple scale from standardization, the economic assumptions that dictated how we organize institutions begin to fall apart.

What If Customization Becomes as Cheap as Uniformity?

Consider higher education. For centuries, the most economical way to teach 100 students was to put them in a room together. They heard the same lecture, read the same textbook, moved at the same pace, and took the same examination.

There are some valid pedagogical reasons for shared experiences. But there was also a prosaic economic reason the system looked this way. Individualized instruction was impossibly expensive.

Now imagine a system where every student has an AI tutor capable of observing what that student understands, identifying specific misconceptions, generating exercises at the exact level of difficulty required, changing explanations when one fails, and doing so continuously. A hundred thousand students might use the underlying platform, but no two would receive the same education. That is a fundamental decoupling of scale from uniformity, and mass education starts to mean something entirely unexpected

The interesting strategic question is not whether a university should “adopt AI.” The question is: Which parts of our institution are standardized because uniformity improves learning, and which are standardized only because personalization used to be too expensive? We rarely needed to distinguish between the two before. Now we must.

Law Firms and Legal Departments Made the Same Bargain

The modern law firm is an extraordinarily successful solution to a historical coordination problem. Expertise had to be assembled in one place. Young lawyers had to be trained. Institutional memory had to be stored. Complex transactions required dozens of workers. Clients needed access to specialized niches. Quality required layers of supervision.

All of this was expensive to coordinate across human beings, so we built large institutions. Law firms created associate classes, practice groups, partnership tracks, libraries, knowledge-management systems, billing departments, and risk-management structures. Those mechanisms solved real coordination problems under twentieth century and early twenty-first century constraints by building mass assembly lines designed to scale billable hours.

Corporate legal departments made their own version of the bargain. To keep administrative coordination from collapsing into chaos and to control outside spend, they created preferred-provider panels, outside-counsel guidelines, standard playbooks, matter taxonomies, and rigid reporting templates.

How many of those features exist because they are inherently necessary to excellent legal work, and how many exist simply because they were once the cheapest way to coordinate human labor and control costs at scale?

AI does not need to replace lawyers for that question to reshape the market. Suppose a boutique team of five lawyers can dynamically assemble research capability, drafting capacity, project management, document review, data analysis, and workflow automation on demand, rather than maintaining those overhead capabilities permanently inside a 500-lawyer organization. The important development would not be that five individuals became “10x lawyers.” It would be that the minimum viable institution got dramatically smaller.

Optimizing Legacy Scaffolding

Much of today’s enterprise AI discourse starts from a lazy assumption: the institution remains unchanged, and AI is simply plugged into the existing machinery. How can AI help us process standard interrogatories faster? How can AI improve our knowledge-management portal? How can AI automate billing intake? How can AI increase realization rates? How can AI help the law department manage outside counsel?

These are optimization questions. They assume the current organizational structure survives, and that software exists merely to make the traditional assembly line run faster. That is the trap of legacy scaffolding: using new technology to reinforce structures designed for constraints that no longer exist.

Imagine instead that software orchestration makes it cheap for a corporate legal department to generate a tailored playbook for every transaction, a customized reporting structure for every internal business client, and a specific instruction set for every matter. Historically, that level of variation would have been administrative lunacy. The coordination costs would have destroyed the department.

That is no longer true. The strategic future of AI in operations is less about automating standardized workflows than about making nonstandard workflows economically manageable.

Necessary Friction vs. Waste Friction

This is where Prudent AI parts company with the move-fast-and-break-things crowd. When coordination costs fall, the temptation is to remove all friction immediately. But some institutional friction is waste, while other friction is protection.

Peer review is slow. Legal risk assessment is slow. Strategic deliberation is slow. Developing professional judgment in junior talent is slow. Due process is deliberately slow.

The mistake is removing friction without first asking what function that friction was performing. Waste friction exists solely because human coordination, documentation, and data transfer used to be slow and expensive. Necessary friction exists to enforce deliberation, catch rare catastrophic errors, preserve institutional memory, or protect fundamental rights.

AI gives us the capability to eliminate both instantly. Prudence consists of knowing the difference. Before dismantling an inefficient process, you must know everything that process was quietly maintaining.

Renegotiating the Bargain

The twentieth-century professional career was itself standardized: join an institution, acquire credentials, climb the hierarchy, accumulate seniority, and let the institution supply the tools, clients, reputation, and security.

AI and global networks are starting to chip away at that monopoly on scale. An individual professional can increasingly assemble a temporary, highly potent stack of AI capabilities, specialized human collaborators, and targeted tools tailored to the specific problem in front of them. The career becomes less about climbing a permanent corporate ladder and more about repeatedly configuring the right system for the task at hand.

As you evaluate your organization’s long-term strategy, put these diagnostic questions on the whiteboard:

  • What are we standardizing only because personalization used to be expensive?
  • What are we centralizing only because coordination used to be expensive?
  • What requires a large organization only because expertise used to have to reside permanently under one roof?
  • Which of our slow processes are slow because human coordination is inefficient, and which are slow because deliberation is valuable?
  • If we were designing this organization today, assuming frontier AI and network orchestration from the start, which parts of our current structure would we never build?

We have spent the last few years asking how AI will perform the work we already do. The far more consequential realization is that AI alters the underlying economics that caused us to organize the work this way in the first place. The old standardization bargain is up for renegotiation. Prudent leadership starts by deciding which terms are worth keeping.

Prudent AI governs consequential operational commitments under uncertainty. As you dismantle legacy scaffolding, eliminate waste friction rapidly, but protect the necessary friction that preserves judgment and institutional resilience.

Accelerate reversible learning. Pace irreversible commitment.

Dennis Kennedy – CC BY 4.0 license

[Originally posted on DennisKennedy.Blog (https://www.denniskennedy.com/blog/)]

DennisKennedy.com is the home of the Kennedy Idea Propulsion Laboratory

DennisKennedy.Blog is part of the LexBlog network.