Why Grassroots AI Adoption Needs Trusted Intermediaries

There is a quiet figure at the centre of almost every successful technology story in India, and they are almost never the ones celebrated. They are the intermediary: the common service centre operator who fills in an online form, the shop assistant who sets up a payment app, the schoolteacher who is “good with computers”, the cousin who explains a new phone to the rest of the family. These people are the human interface between technology and the ordinary Indian. And they will decide, more than any product or policy, how AI actually reaches the ground.

This is easy to miss from a distance. From a distance, adoption looks like a curve — a technology appears, people use it, the line goes up. Up close, that curve is made of countless small acts of trust, and almost every one of them passes through a person the user already believes in.

Trust does not scale like software

Software scales effortlessly. Copy it a million times and each copy is identical and instant. Trust does not behave this way. Trust is built slowly, locally, and person to person. It cannot be downloaded. And for most Indians deciding whether to try something unfamiliar, trust is the deciding factor — not features, not price, but whether someone they believe in has vouched for it.

This is why a brilliant tool can languish while a modest one spreads. The modest one had an intermediary — a trusted local person who tried it, understood it, and passed it on. The brilliant one was left to reach people directly, and direct reach, for the unfamiliar and the unconfident, is weak. AI is more unfamiliar and more invisible than most technologies that came before it. Its need for trusted intermediaries is therefore greater, not smaller.

What an intermediary really does

It is tempting to think of the intermediary as merely a helper who clicks the right buttons. Their role is far larger. They translate — not only across languages, but across worldviews, turning an abstract capability into something that makes sense within the life of the person in front of them. They vouch — lending their own hard-earned credibility to something new. They filter — steering people away from what is harmful or fraudulent. And they adapt — figuring out how a general tool can serve a very specific local need.

For AI, each of these functions becomes more important. Translation matters more because AI’s workings are so opaque. Vouching matters more because the stakes of misuse can be higher. Filtering matters more because AI has made deception cheap and convincing. Adaptation matters more because a general-purpose tool means little until someone shows how it applies to this trade, this village, this problem. Strip away the intermediary and AI does not reach the grassroots gently; it arrives as either a mystery or a menace.

The risk of ignoring them

When technology efforts ignore intermediaries, two failures tend to follow. The first is simple non-adoption: the tool never reaches the people it was meant for, because there was no trusted human to carry it the last mile. The second is worse — adoption without understanding, where people are pushed into using something they do not comprehend, mediated by no one who can protect them, and left exposed to error and exploitation.

There is also a fairness dimension. The communities most in need of help are usually the ones furthest from direct digital reach, and therefore the most dependent on intermediaries. Ignore the intermediary, and you do not simply slow adoption; you skew it towards those who least needed the help, and away from those who needed it most.

Investing in the human layer

If intermediaries are this central, then a serious approach to grassroots AI adoption has to invest in them deliberately. That means equipping them with genuine understanding rather than scripts — so they can answer real questions, not just recite instructions. It means supporting them in their own languages and contexts. It means treating them as partners in the work rather than as a distribution channel to be used and forgotten. And it means recognising the responsibility they carry, and helping them carry it well, including the responsibility to protect the people who trust them.

This is patient, human-centred work. It does not produce dramatic launch numbers. But it produces something more durable: a layer of trusted, capable people through whom AI can reach communities in a way that empowers rather than bewilders them.

The face of the machine

For most Indians, AI will never have the face of a company or a product. It will have the face of a neighbour, an agent, a teacher — the trusted person who first showed them what it could do and stayed to help when it went wrong. That is not a limitation to be engineered around. It is the natural grain of how technology and trust travel in this country, and the wisest path is to work with it.

At Svadrishti Foundation we see the strengthening of this human layer as some of the most important work in front of us. Not because it is efficient — it is not — but because it is faithful to how real adoption happens, and because it keeps the people at the grassroots in the hands of others they trust. Getting the machine to the many, gently and honestly, is finally a human task.