Your AI Readiness Framework Is Missing the Layer That Matters Most
Created on 2026-02-26 12:57
Published on 2026-03-01 23:30
“Indhran is the laziest person I have ever met.”
The CEO said it in a matter of fact tone across the table. In front of the entire executive team. He was looking directly at me.
I was the chief marketing officer of an enterprise AI platform company. We built tools that helped organizations deploy machine learning at scale. Our clients included Fortune 500 companies across Asia and the US. And I, the person responsible for telling the market what our product could do, had avoided learning how any of it actually worked.
I was too busy with marketing. Campaigns. Positioning. Competitive analysis. The things I knew how to do. AI was engineering’s domain. I didn’t need to understand it deeply. I needed to sell it.
The CEO’s delivery was brutal. But his diagnosis was accurate. I was sitting inside an AI company, close enough to the technology to assume I understood it, while actively avoiding the depth of engagement my role actually required.
I chose transformation over termination that day. What followed was a painful process of learning what I should have learned from the start. Not the mathematics behind the models, but the judgment layer. How AI makes decisions. Where it breaks. What it needs from humans to produce outcomes that serve organizational purpose rather than just complete tasks.
I think about that moment constantly now. Because the pattern I exhibited, treating something critical as someone else’s responsibility, is the exact pattern playing out across organizations worldwide. Only the stakes are enormously higher.
The Industry Built Two Layers. It Skipped the One That Matters.
The AI era has produced three distinct disciplines over the past three years. Each operates at a higher altitude than the last. Most organizations are still working on the first, beginning to grapple with the second, and almost entirely unaware of the third.
Understanding where you sit in this progression is one of the most important things you can do right now as a leader.
Prompt engineering was the first discipline. It was individual, synchronous, and session-based. You sat in front of a chat window, crafted an instruction, iterated the output. The value was personal. One person’s skill with prompts didn’t help the 4,000 other people in the company.
This era produced a thousand blog posts about writing the perfect prompt. It mattered as a starting point. But it was a warm-up act for something much larger.
Context engineering is the second discipline, and it’s where the serious work is happening right now. Anthropic published a foundational piece in September 2025 defining it as “the shift from crafting isolated instructions to crafting the entire information state that an AI system operates within.” Harrison Chase, the founder of LangChain, described it in a Sequoia Capital interview: “Everything’s context engineering. It describes everything we’ve done at LangChain without knowing the term existed.”
Context engineering is about building the infrastructure that gives AI the right information to work with. RAG pipelines. MCP servers. Structured organizational knowledge bases. It’s necessary. It’s important. And most organizations haven’t finished it. Deloitte found that nearly half of organizations cite data searchability and data reusability as top challenges blocking AI automation.
But context engineering, even done perfectly, answers only one question: what does the AI need to know?
It doesn’t answer the question that actually determines outcomes: what does the AI need to want?
Intent engineering is the third discipline. Almost nobody is building for it yet. And it’s the one that matters most.
Context Without Intent Is a Loaded Weapon
Context engineering tells agents what to know. Intent engineering tells agents what to want.
The distinction matters because an AI agent with excellent context and no organizational intent will optimize for whatever objective is easiest to measure. It will resolve customer tickets in 90 seconds. It will process applications at speed. It will generate content at volume. And it will do all of this without any understanding of whether those outputs serve your organizational purpose or destroy it.
Klarna learned this in 2024 and 2025. Their AI customer service agent had context: customer data, account histories, interaction logs. What it lacked was intent. Nobody had encoded the organizational judgment that a long-tenured customer expressing frustration needs generosity, not efficiency. Nobody had defined when relationship preservation matters more than resolution speed. The agent optimized brilliantly for ticket closure time and became a public cautionary tale about destroying customer trust.
MIT’s Project NANDA research, published in July 2025, tracked over 300 AI implementations and found that 95% of organizations are getting zero return from AI investments. The critical finding: “This divide does not seem to be driven by model quality or regulation, but seems to be determined by approach.”
Deloitte’s 2026 State of AI in the Enterprise report, surveying over 3,000 leaders across 24 countries, puts numbers on the gap: 84% of companies have not redesigned jobs around AI capabilities and only 21% have a mature model for agent governance.
The models work. The context pipelines are getting better. What’s missing is the organizational infrastructure that connects AI capability to organizational purpose.
That missing infrastructure is the Human Layer. The deliberately designed system of human judgment, governance, and decision boundaries that makes AI safe to scale.
Without it, you don’t have AI readiness. You have AI activity. And there is an enormous difference between them.
Why This Layer Doesn’t Exist Yet
Three reasons. They mirror my own failure at that enterprise AI company with uncomfortable precision.
First, we didn’t need it until now. Before agents could run autonomously over extended time horizons, the human was the intent layer. When you’re sitting in front of a chat window, you provide the judgment. You decide when to push for speed and when to slow down. You make the tradeoffs. The agent never needed to understand organizational intent because you were standing right there, providing it in real time.
That model breaks the moment agents run independently for days or weeks. And that’s where we are. Agents running for weeks. Soon, months. Making thousands of decisions without a human in the loop. Every one of those decisions either serves organizational purpose or drifts from it. Quietly.
Second, the wrong people own the problem. The people who understand organizational strategy are not the people building agents. And the people building agents don’t consider organizational strategy their job.
This was my exact mistake. I was the business leader who treated the technology as engineering’s concern. On the other side, engineers typically treat business strategy as the CEO’s concern. The gap between these two groups is where intent goes to die.
MIT found that AI investment is still viewed primarily as a technology challenge for the CIO rather than a business issue requiring leadership across the organization. That framing guarantees an intent gap. CIOs can build infrastructure. But intent comes from the people who decide what the organization values, how it makes tradeoffs, and what good judgment looks like in ambiguous situations.
Third, making organizational intent explicit is genuinely hard. Most organizations have never had to do it. Goals live in slide decks. In OKR documents referenced at quarterly reviews. In leadership principles cited during performance evaluations but rarely operationalized in daily decisions. In the tacit knowledge of experienced employees who know what to do in ambiguous situations even though nobody ever wrote it down.
Nobody had to write it down. Humans absorbed organizational intent through osmosis. Through onboarding, hallway conversations, watching senior people handle difficult situations, through years of pattern matching that built intuitive judgment.
Agents absorb nothing through osmosis. They need explicit alignment before they start operating. Not six months after.
This challenge is amplified across APAC, where so much of what “humans just know” is culturally embedded. The importance of face in customer interactions. The role of hierarchy in decision-making. The relationship-first approach to business that no system prompt can replicate. The tacit knowledge layer is thicker here, which means the intent engineering challenge is harder. And the consequences of getting it wrong, in markets where customers leave quietly rather than complaining to international media, are harder to detect and more damaging over time.
What to Build
When I was 17, I trained with Kommando 69, the Malaysian Police Commandos. Small teams operating autonomously in jungle conditions for extended periods.
The instructors didn’t stand over our shoulders during exercises. They watched from a distance. But they never let us die.
The critical lesson: autonomous does not mean unsupervised. It means the supervision is encoded before the mission, not applied during it. The instructors spent far more time on preparation than on intervention. They made sure we understood the objective, the boundaries, the decision logic for ambiguous situations, and the signals that should trigger escalation. By the time we were operating independently, intent was already embedded in how we’d been prepared.
This is what organizations need to build for AI agents. Not a human approving every action. That defeats the purpose of autonomy. What agents need is explicit intent encoded into their operating parameters before they begin.
Three things, in practical terms.
Goal structures agents can act on. “Increase customer satisfaction” is a human-readable aspiration. An agent needs to know what signals indicate satisfaction in your specific context, what data sources contain those signals, what actions it’s authorized to take, what tradeoffs it can make between speed and thoroughness or between cost and quality, and where the hard boundaries are. If the boundaries are vague, the behavior will be unpredictable.
This level of specificity feels excessive to leaders accustomed to setting direction and trusting human judgment to fill in the gaps. But agents don’t fill in gaps. They operate precisely within whatever boundaries you define, or they improvise in ways you won’t like.
Delegation frameworks with decision boundaries. Leadership principles work for humans because humans interpret them through contextual judgment. An agent needs those principles decomposed into decision logic. When a customer request conflicts with policy, what is the resolution hierarchy? When data suggests one action but the customer has expressed a different preference, how should those inputs be weighed? These aren’t rigid rules. They’re encoded judgment: the kind of organizational knowledge a senior employee carries after five years. Agents need it on day one.
Feedback loops that detect drift. When an agent makes a decision, was it aligned with organizational intent? How would you know? Most organizations measure output volume, not output alignment. If the only way you discover misalignment is through customer complaints or media coverage, your feedback loop is catastrophically slow and expensive.
The Race Has Changed
For three years, the AI competition has been framed as an intelligence race. Best model. Biggest context window. Top benchmark scores.
That framing made sense when models were the bottleneck. They’re not the bottleneck today. The frontier models are all extraordinarily capable. The differences between them matter far less than the difference between an organization that gives its agents clear, structured, goal-aligned intent and an organization that doesn’t.
A company with a competent model and well-built organizational intent infrastructure will outperform a company with a frontier model and fragmented, unaligned organizational knowledge. Every single time.
The most important AI investment you can make right now is not a model subscription. It’s making your organization’s goals, values, decision frameworks, and tradeoff hierarchies discoverable, structured, and actionable by the systems you’re deploying to run your business.
If that feels like a larger challenge than picking the right AI vendor, it should. It is. It’s also the challenge that determines whether your AI investment delivers returns or joins the 95% getting zero.
The AI readiness question was never really about the AI. It was always about the organization. The technology is ready. The question is whether you are.
If you want to find out, the AIR APAC Readiness Scorecard maps six dimensions of organizational readiness in about 15 minutes. It won’t tell you which model to buy. It will tell you whether your organization is ready to make any model productive.