The Intelligence Network

Most firms are a disconnected brain - AI is the connective tissue

Hi there,

AI is transforming everything, with companies evolving into intelligence networks of interconnected AI systems.

These companies will generate compounding value that differs from traditional network effects, leading to significant implications for the workforce.

The Firm as a Disconnected Brain

What is an operating model? It’s people, process, and technology. Firms decide how to structure based on whether it's cheaper to do a task in-house or pay a supplier.

Digital transformation has enabled firms to build highly automated operating models. Amazon has distribution warehouses, and Netflix streams content to millions.

Even automated firms have a weak point: we automate tasks without automated coordination. Human coordination via meetings and emails is needed for System A to interact with System B.

A pre-transformation firm resembles a disconnected brain with little connective tissue. It hears with dashboards and reports, but can’t combine systems into coherent information. Human workers must connect these parts to provide the correct data for decision-making.

The intelligent firm improves connectivity among the parts. The new approach is to execute tasks automatically and transfer insights between intelligent systems at machine speed. This is the AI, API, AI agent hybrid I’ve discussed recently.

The Engine of Compounding Value

The old method combines resources, like a technology platform, an operations team, and outsourcing business processes. The new approach emphasises network and learning effects that enhance value. Economists describe this as multiplicative or super-additive. Overall value depends on how well a firm’s components perform, connect, and coordinate. Lack of communication between business functions reduces overall value.

This concept isn't new for corporate professionals, who already engage in transformation efforts, cultural resets, and restructuring. The significant change? AI will facilitate this process efficiently, surprising stakeholders. 

Three elements in this compounding mechanism are: shared insight, mutual supervision, and emergent capabilities. Shared insight goes beyond systems communicating; it involves recognising patterns and generating insights that teams might currently analyse to improve effectiveness.

Achieving this requires extensive data. Firms that enhanced data quality years ago may succeed; a comprehensive business context is necessary for leveraging this automated environment. AI wearables document workplace activities and will likely become a contentious worker rights issue, eliminating off-the-record conversations.

This environment resembles an Amazon Warehouse for office workers - unpleasant! Data deletion could be seen as managerial negligence. Users of AI coding find that comprehensive logging enhances AI performance. Erasing data after seven years risks losing crucial shared insight. Expect increased lobbying against data deletion and privacy regulations.

Mutual supervision resembles enhanced audits or peer reviews, with AI agents continuously reviewing operations. With prevalent API-based communication, incident response capabilities will evolve towards self-healing. Monitoring agents will detect discrepancies in quality metrics and investigate without needing preliminary conference calls, leaving humans in complex roles that require real-world problem-solving.

AI governance discussions highlight this mutual supervision layer, converting frameworks and policies into code. Given the prescriptive nature of risk management and compliance, firms will have less room for policy deviations.

Emergent capabilities rely on AI agents to develop new strengths and capabilities previously unimagined by executives. For instance, Stripe uses its data to provide merchant advance loans with insights surpassing traditional financial institutions.

I find boundary changes within emergent capabilities particularly intriguing. Firms can “borrow” capabilities from others via APIs, enabling enhancements. AI agents will learn about publicly available API endpoints and generate new operational models aligned with minimising losses, free from boardroom risk appetites. 

As outsourcing and offshoring decisions become increasingly machine-driven, concepts of national borders and regulations will face significant lobbying pressure from employees of AI-managed firms. 

Any friction stopping AI tools from optimising their actions is a significant issue. My concern lies less with the “paperclip maximiser” AI doom theories and more with aggressive AI-enabled lobbying to roll back legislation to accommodate AI agents doing what they want.

This intelligence network effect diverges from standard network effects, catalysing efficiency and innovation and creating supply-side value. The risk lies in failing to impose adequate safeguards, potentially leading to unwanted optimisation outcomes.

The New Coordination Function

This emerging era has two trends happening simultaneously: the automation of tasks with AI, replacing jobs starting with entry-level tasks and working up the complexity chain as AI capability grows.

I recently wrote about this operating model compression trend. The second is the coordination of automation—the real connectivity between resources in the production function.

This creates and resolves a paradox: a new demand for high-skill human oversight exists. If the firm is an intelligence network, it needs a group of human “gardeners” or “concierges” to complete the human tasks that can’t be automated, yet.

This involves things like architecture, in partnership with AI tools. It includes connecting the intelligent firm to whatever political and regulatory structures still exist. Some human trainers or coaches help refine the interactions between systems and improve the whole system.

Over time, autonomous AI agents are increasingly handling the technology platform of this “intelligent firm,” while people are handling the design, strategy, and complex exception processing.

Determining what labour replacement level will be achieved over the next few years is difficult. Each model release shows that AI can do more complex tasks. Over time, more firms will assign more tasks to AI tools. Only a handful will truly reinvent themselves into intelligent firms.

There is still room for humans in the loop, but the minimum skill level required to stay in the labour force will keep rising. I was only half-joking when I posted a dystopian Substack note about STEM PhD holders being the only folks passing the automated resume screening for job applications in a few years.

The Leader's Choice

So, what do you do if you’re leading a business in 2025? I think the strategic choice is getting clearer.

You need to start thinking about a two-track strategy:

1. Partner with AI for absolutely everything and automate whatever you can inside the legal and regulatory constraints you face, regardless of internal politics or lack of buy-in - you are in an existential challenge, your fiercest competitors don’t care about any of the restrictions you impose on yourself.

2. Think carefully about what enduring monopolistic-type advantages you have in proprietary data, trade secrets, or market share dominance that you can entrench with even more AI-automated workflows. Then, start to design automated end-to-end value chains: this probably requires substantial investments in data quality and doing all of the “foundation work” that got kicked down the road previously because of “other priorities.”

This strategy is going to have to move into the execution phase fast. One of the most significant risks big corporations face today is that with the leverage of AI tools, a tiny team of key subject matter experts can probably AI-first code their way to building peer-level capability in some of your biggest products or services far faster than you ever thought possible.

How will you respond when all of the old corporate behaviours around cost control, layers of management, and micro-managing are the exact anti-patterns that lead to doom in the AI-era?

Best regards,

Brennan

PS: This week, I started making daily video shorts on AI. You can find me on YouTube, Instagram or TikTok.