What “Locally Led Adaptation” Means When You Introduce AI

Locally Led Adaptation (LLA) is often described as a shift—of power, decision-making, and ownership—closer to communities living with climate risk.

On paper, the idea feels straightforward.

In practice, the moment artificial intelligence enters the picture, LLA becomes far more complicated.

Because AI, by design, tends to centralize data, models, expertise, infrastructure, and often authority. Even when deployed with good intentions, AI systems quietly reshape who gets to decide what counts as “good” adaptation.

So the real question isn’t whether AI can support Locally Led Adaptation.

It’s whether AI can exist without slowly undoing it.

The Comforting Myth of “AI for Farmers”

Many climate technology tools proudly describe themselves as “AI for farmers” or “AI for communities.” The framing is reassuring. It suggests proximity, usefulness, and inclusion.

But in practice, proximity is not the same as power.

During work with Adaptation Clinics and advisory tools like Tia Apa, one pattern surfaced repeatedly: farmers rarely wanted instructions. What they wanted was conversation—options, explanations, and the ability to weigh advice against their lived realities.

Field facilitators often played a critical role here. They interpreted recommendations, contextualized risks, and sometimes ignored digital advice altogether when local conditions didn’t align.

That friction was not a failure of technology. It was a signal of agency.

Locally Led Adaptation is not about who receives information. It’s about who gets to shape it, question it, and reject it.

When AI systems are designed elsewhere, trained on abstracted datasets, and deployed locally with fixed logic, adaptation becomes something that happens to communities rather than with them.

That isn’t locally led. It’s locally delivered automation.

When Local Knowledge Becomes a Casualty

One of the most underestimated tensions between AI and LLA lies in how we treat local knowledge.

In Adaptation Clinics, farmers often describe risk narratively:

• land that behaves differently after repeated floods
• pests that appear when winds shift direction
• seasons that no longer follow calendars

This knowledge is contextual, experiential, and often non-linear. It is shaped by loss, memory, and adaptation over time.

AI systems, on the other hand, prefer clean inputs: structured variables, historical consistency, and standardized categories.

When we force local knowledge to fit AI systems, we don’t just lose nuance—we lose trust.

Treating local insight as “training data” flattens what makes it powerful. Locally Led Adaptation asks something harder of technology: not extraction, but respect.

In practice, this often means hybrid systems—where AI supports decision-making rather than replaces it, and where human judgment remains central, visible, and legitimate.

Disagreement with AI, in this context, is not an error. It is a sign that adaptation remains locally grounded.

Power Hides in Defaults

AI rarely announces its authority outright. Power tends to hide in subtler places: default settings, risk thresholds, and “recommended” actions.

In climate adaptation, these defaults matter deeply. A wrong recommendation can mean crop failure, income loss, or long-term vulnerability.

During advisory work, field staff often asked a simple but important question: “What if this advice doesn’t fit this farmer’s reality today?”

That question reveals whether adaptation is truly locally led.

LLA requires the right to override technology, to adapt its logic, and to say, “This does not apply here.” If local actors cannot contest AI outputs, decision-making has already shifted—quietly and efficiently.

An AI system that cannot be questioned is not empowering. It is controlling.

Strengthening Institutions, Not Bypassing Them

Another common temptation in climate tech is to go directly to individuals—apps for farmers, dashboards for households, and automated advice at scale.

But Locally Led Adaptation is not only about individuals. It depends on local institutions: facilitators, extension workers, cooperatives, and community structures that mediate trust and accountability.

In the most effective cases, AI did not replace these actors. It reduced their cognitive load, supported better conversations, and left final decisions where they belonged.

When technology bypasses institutions, adaptation becomes fragile. When it strengthens them, it becomes resilient.

What LLA With AI Actually Demands

From field experience, a few principles stand out:

• AI should support local judgment, not replace it
• Communities must influence how systems evolve, not just use them
• Local institutions matter more than algorithmic precision
• Slower, trust-based pilots often outperform rapid scaling
• Adaptation is social first—technology second

If AI makes climate programs faster but less democratic, something fundamental has gone wrong.

The Uncomfortable Question

As AI becomes more embedded in climate adaptation, we need to ask:

Are we decentralizing intelligence—or just redistributing interfaces?

Locally Led Adaptation is not a feature you add to technology. It is a discipline that forces designers, institutions, and funders to give up control.

AI does not naturally do that.

And that tension—between efficiency and agency, scale and legitimacy—is not a problem to solve once.

It is the work.