Table of Contents
The Real Challenge is Coordination
We’ve known how to solve many of the world’s biggest challenges for decades. Energy abundance, clean water, global nutrition, access to education. The blueprints exist. The constraint isn’t ideas or information - it is execution, which breaks down when coordination fails.
Why? Because incentives are misaligned. Power structures are entrenched. Teams and systems operate in silos. Each local actor optimizes their own objective function, often at odds with the whole.
This isn’t a new problem. It’s the oldest one in complex systems. We don’t fail because we lack intelligence. We fail because we can’t align our intent and actions across the layers of people, tools, and processes required to make something real. These forces act as invisible spores of organizational slime mold, decomposing the best intentions of otherwise hard working, intelligent, motivated people.
This is where AI becomes interesting, not as a source of new knowledge, but as a tool for better coordination.
Shallow Automation Misses the Point
Most people approach GenAI as a way to automate tasks. Engineering uses it to quickly generate documentation. Customers use it to summarize the same documentation to avoid reading it. Since no one reads the docs, the company creates a chatbot trained on them to provide customer support. All shallow wins. None of this fundamentally changes how the organization actually works.
As Jules White writes in The AI Labor Playbook, this isn’t about doing the same work faster:
This is not just about doing the same work faster; it’s about doing better work, bolder work, and work that wasn’t possible before.
If refocused, AI can reveal something deeper: emergent coordination logic. When embedded across the workflows of a business, not just bolted onto the edges, it starts to surface patterns no individual can see.
Reengineering and Constraints: Old Problems, New Tools
This challenge isn’t new, but echoes ideas from decades ago. Dr. Michael Hammer was a computer science professor at MIT who championed the concept of Business Process Reengineering (BPR) in the 90’s. In his HBR article Reengineering Work: Don’t Automate, Obliterate, he argued that automating existing processes is insufficient, and that true transformation requires radical redesign - obliterating outdated practices and leveraging technology to rethink how work is done for dramatic improvements.
The usual methods for boosting performance — process rationalization and automation — haven’t yielded the dramatic improvements companies need. In particular, heavy investments in information technology have delivered disappointing results — largely because companies tend to use technology to mechanize old ways of doing business. They leave the existing processes intact and use computers simply to speed them up…
It is time to stop paving the cow paths. Instead of embedding outdated processes in silicon and software, we should obliterate them and start over. We should “reengineer” our businesses: use the power of modern information technology to radically redesign our business processes in order to achieve dramatic improvements in their performance.
In the 80’s, Eliyahu Goldratt introduced the Theory of Constraints in his book The Goal: a concept that argued every system has a single weakest link, its constraint, improving anything else is wasted effort. Most organizations focus efforts on the place of the latest embarrassing screw-up, or the target of the squeakiest wheel. They’ll spend enormous efforts, time and cost, but realize little holistic benefit.
Even organizations that look for constraints still haven’t found theirs, because it’s not visible in a dashboard. It’s buried in handoffs, assumptions, and unowned process steps under the guise of tacit knowledge or simply the way things have always been done.
AI as Coordination Discovery Engine
These earlier frameworks diagnosed the problem. Today’s AI tools, when properly designed, give us a shot at solving it. AI, paired with the right architecture, has the potential to move us out of this pattern of reactive optimization and into a mode of continuous flow, where the real constraint is identified, surfaced, and actively addressed in real time. Embedded across the system, AI isn’t limited to automate rote tasks. Instead it can objectively observe, find what’s actually holding things back, and without ego make recommendations.
Rather than being a cost-cutting scheme to replace the people, AI surfaces where the real bottleneck lives, so the work can flow. It notices the process friction no one owns. The delays between legal and sales. The conflicting objectives between finance and ops. The redundant steps across engineering and product. It connects what’s actually happening, not just what’s written down, or what people say they do.
This flips the typical GenAI deployment mindset. Most companies look for places to automate. But the bigger opportunity is to embed AI across your workflows and use it to surface where coordination is failing.
The Model Context Protocol (MCP) plays a crucial role here. It removes glue code between systems and allows AI agents to operate with shared context across tools like Slack, Notion, Jira, and Salesforce. Instead of writing brittle integrations, you can create composable workflows that adapt in real time. Agents can access and update shared context as they evaluate how work flows (or doesn’t) across systems.
Beyond looking across systems, the OpenAI agents project highlights several patterns that are directly useful in uncovering and improving coordination:
- Agent as Judge: Evaluates outputs or decisions made by other agents. Ideal for flagging where a workflow broke down or misaligned.
- Agent as Mediator: Intervenes between agents or stakeholders in conflict, helping resolve friction and propose compromise steps.
- Agent with Handoff: Transfers context and tasks from one agent to another. Useful in long-running or multi-phase workflows where context continuity is key.
- Agent as Critic: Reviews reasoning or workflow outputs against desired principles or patterns.
DeepMind’s Era of Experience frames this shift well, outlining how AI agents can learn not by imitating us, but by acting, observing, and constantly adapting across long-running environments. Through continuous observation, contextual analysis, and tuned reward systems, AI can surface and resolve coordination breakdowns holding organizations back.
For example, in a product launch workflow, an Extractor Agent pulls activity from meetings, tickets, and chat logs. A Judge Agent flags where scope decisions are repeatedly made without product alignment. A Mediator Agent suggests inserting a checkpoint earlier in the sprint cycle. None of this is predefined, but it emerges from how agents observe and collaborate via shared context.
From Automation to Elite Performance
These patterns let us model not just automation flows, but coordination logic. One agent can extract workflow data, another can judge where friction exists, and a third can mediate a solution. This architecture isn’t hard-coded. It’s declarative, dynamic, and adaptive.
In practical terms, this means you can build a coordination audit pipeline with agents. Pull documents, threads, and tickets from across tools. Let one agent build context, another analyze for breakdowns, and a third propose workflow improvements. All of it flows through shared context.
AI doesn’t just make workflows faster. It serves as a guide to how workflow should change. This isn’t just faster execution. It’s a better way to operate.
It’s not about replacing people. It’s not about being clever. It’s about seeing across the system and finding a better way to coordinate.
As Jules White warns,
If your primary goal is to reduce your human workforce, you’ll likely end up doing the same—or less—while your competitors expand capacity, move faster, and outcompete you.
The future of AI in organizations is not just speed. It’s system-level intelligence. It’s discovering how to execute differently.
That’s how you get out of POC purgatory. That’s how you move from automation to transformation. That’s how you make progress.