The Invisible Work of Building a Social Intelligence Function
When I started working with social listening, I didn't actually know what social listening was. I was a young padawan building a big agency's digital marketing department from scratch in my hometown in Brazil, mostly by Googling things as I went. Part of my job was monitoring campaigns and messages on social media – we had only recently started tagging conversations so we could assess how the campaign had performed and shape our clients' next strategy. At the time, I genuinely thought it was…customer service with extra steps.
Booleans? Never heard of them.
Life pivoted hard since then: I fell headfirst into social intelligence, changed continents, and spent years building - and rebuilding - services, teams and technologies. I won't pretend I have all the answers, but I do have a few opinions. So, here's what I wish someone had told young padawan-me about building social intelligence capability across an organisation.
A tool is not a capability
If you're a regular follower of The SI Lab, you already know that one of the easiest traps is confusing having a social intelligence platform with having social intelligence capability. You can have an excellent technology, a sophisticated team and thousands of users with access and still struggle to get social intelligence into the decisions that matter. The capability lives in everything around the tool: how people are enabled to use it, how it enters their day-to-day workflow, how they know that what they're looking at is meaningful, and, most importantly, why they should make time for it at all.
That last one is particularly relevant in a big company. Social intelligence isn't just competing with other, long-established research methodologies; it's competing for attention with who-knows-how-many other platforms, dashboards and processes already fighting for a slot in someone's browser tabs. It might be the most important thing in the world to you— but to someone in Marketing, R&D or Insights, it's one more tool to figure out.
Which brings me to my favourite hill to die on: democratisation is not the same as distribution. Giving someone a login does not make them capable of turning social data into an insight, anymore than giving me access to a database makes me a data engineer. If we want people to use social intelligence consistently, we need to build the structures, training, support and use cases that make it useful in the context of their day-to-day work.
Build for people, not just for use cases
Coming from an agency, and being a marketer myself, I assumed marketers would instantly get what I was showing them. Then a client once asked me during a meeting "What does neutral sentiment mean?" I remember being briefly scandalised — how is this not obvious? It's not positive, not negative, it's right there in the name, and then, more humbling: hold on… what does neutral actually mean? And it's not just sentiment. Ask ten people what "engagement" means and you'll get ten answers depending entirely on who's asking.
That question has followed me around ever since, as a reminder to put myself in someone else's shoes. Because I was once that person too, the one who didn't know what Booleans were, or how exactly engagement was calculated. So now, when I train teams and the newer generation of social listeners, I try to get across one thing: nothing is obvious. You'll be dealing with very different stakeholders, and "neutral," "engagement" and "results" mean different things depending on who's in the room. Social intelligence is unusually transversal — even more than most research practices. Marketing wants one thing, innovation another, communications another, consumer insights another.So you almost have to productise the capability around different jobs-to-be-done, and then sell it, repeatedly, in each team's own language.
And there's no single "social intelligence user" to sell to, either. There are the people building queries and taxonomies, the technical social listeners, the insight professionals turning data into stories, and the stakeholders who only touch social intelligence when they need an answer to one specific question. They all need different things. So "enablement" can't mean the same for everyone. We also tend to underestimate how much insight skills matter on the technical side: even someone mainly building dashboards works differently once they understand how insights are actually consumed, structuring a dashboard around a question rather than a data source.
The invisible infrastructure is the real product
Here's the underrated, unglamorous stuff that actually makes or breaks capability: methodology, standardisation and governance. I know, it’s boring, it’s not sexy, nobody writes case studies about those (sadly). But this is where organisations win or lose. Too much centralisation and SI becomes a service desk that everyone resents., but too much decentralisation and every market builds its own version of the truth, and now you can't compare anything to anything.
Sometimes the best approach, in my opinion, though probably not the easiest, is a hybrid model. Global provides the standards, frameworks, technology, benchmarks and best practices; markets and entities keep the flexibility to adapt to their own reality. Global gives you a best-in-class point to benchmark from; local gives you relevance. I would argue that you genuinely need both.
At Nestlé, for example, we've built global dashboards around key categories and macro topics that serve the whole organisation and act as a best-in-class structure. Local teams can either use these in their day-to-day work or take that same structure and apply it to their own brands and markets if they choose to. And when we enriched our global dashboards with search trends, we also showed markets how to do the same locally. Everyone benefits from the standard, without being boxed in.
It also means thinking about where social intelligence sits within the broader insights ecosystem. Social should not become an island: If your social intelligence lives cut off from the rest of the insights ecosystem — the other research, the knowledge repositories, the toolkit people already open by default — it will always feel like extra work. Connecting it to where insights already live isn't a nice-to-have; it's how SI stops being a novelty and becomes infrastructure.
AI doesn't change the need for judgement
Naturally, we can't talk about capability today without touching AI. It's making social data far easier to query and interpret, stripping away barriers that once took tons of effort. Much of the friction I spent years fighting is finally (hopefully!) melting away. But there's a caveat: remove the friction of using social intelligence without building guardrails around it, and we'll democratise access faster than we democratise good judgement.
In a strange way, AI makes capability more important, not less.
So, what does it really take?
If you've read this far waiting for the tidy checklist, I'm sorry to disappoint. Ultimately, I don't think the test of social intelligence capability is how many people have access to the data. It's how often the organisation knows when to listen, knows what to do with what it hears, and, most importantly, actually changes something because of it. And that only happens with senior buy-in. Social intelligence has to be embedded in the insights ecosystem, the toolkit, and the decision-making rituals of the business. If it isn't, it will forever be"extra work", and the moment budgets or attention get tight, it'll be the first thing dropped, or quietly replaced by whichever AI tool happens to be easiest to click.
Young-padawan me had no idea the hardest part of social listening would have so little to do with listening and everything to do with all the invisible parts around it.