Juan Pablo Tejela, CEO and co-founder of Metricool, is a software engineer and technology entrepreneur who has led the company since 2014, helping develop it into a comprehensive platform for managing and measuring digital content. Before launching Metricool, he spent nearly nine years at Optimyth Software, where his responsibilities included product management, application development, user interface design, presales, customer support and online marketing for the Kiuwan software analytics platform. Earlier in his career, Tejela held software engineering roles at Swissrisk, Indra, EADS-CASA and FEDETEC, working across Java, C++, embedded systems, communications technology and Linux kernel development.
Metricool is a Madrid-based social media management platform that allows creators, brands and agencies to plan content, publish posts, manage conversations, analyze performance, monitor competitors and generate reports from a centralized workspace. The platform supports major social networks and advertising channels, including Instagram, Facebook, TikTok, LinkedIn, YouTube, X, Bluesky, Pinterest, Google and Meta Ads, while also offering AI-assisted content creation and optimization tools. Metricool reports serving more than two million users, including organizations such as Forbes, Adidas, Volvo, H&M and Warner Music (WMG ) Group.
Your career began in software engineering, embedded systems, application quality, and product development. How did those experiences shape your approach when you founded Metricool, and what unmet need convinced you that social media professionals required a new type of analytics platform?
Honestly, Metricool started smaller than people think. My wife Laura and I built it in 2015 as a side project, on top of our regular jobs. Bloggers were the biggest thing on the internet at the time, and we noticed they had no decent way to see their numbers in one place — everything was scattered across five different tools, none of them built for someone who wasn’t technical. My background in embedded systems and QA had drilled one habit into me. You don’t trust something you haven’t measured yourself. So we built the simple version we wished existed.
What actually validated the idea wasn’t the bloggers, though — it was the social media professionals who started signing up right after them. They wanted client reports that looked professional, and this is the part that mattered for a side project — they were willing to pay. Bloggers mostly weren’t. That’s the moment I realized we weren’t building a hobby tool, we were building for an underserved profession. We kept our jobs until it became unmanageable, and started hiring properly in January 2017. Everything since has really just been an extension of that first instinct: strip out the noise, keep the analytics transparent, and build something for the person doing the work, not the person buying software for other people.
Metricool has evolved from a focused analytics product into a social media management platform used by more than four million professionals, agencies, creators, and brands. What were the most consequential product decisions behind that growth, and were there moments when you had to rethink the company’s original vision?
A few decisions mattered more than the rest. Going all-in on analytics while everyone else — Hootsuite, Buffer — was focused on scheduling was the first one, and it’s still our differentiator today. The second was refusing to be an “American” tool that treats Spanish and other languages as an afterthought; that gave us a real edge in Europe and Latin America early on. The third, more recent, was deciding social ads couldn’t live in a separate tool from organic content, because no social media manager actually thinks about them separately.
Did the vision change? Constantly, and I don’t think that’s a bad thing. We didn’t set out to build for agencies, instead we built for individual professionals and bloggers. Agencies showed up on their own, told us what they needed (white-label reports, multi-client dashboards, team seats), and we listened and took action. Same with brands later. If I’d kept the”original vision” for Metricool too much, we’d have built a much smaller product for a niche audience. The plan was never sacred. The users largely laid out the roadmap for us.
Metricool’s research found that 96% of social media professionals use artificial intelligence, but 36% do not know or measure whether AI-generated content performs better. Why has AI adoption moved so much faster than organizations’ ability to evaluate its actual business impact?
Because using AI is easy and measuring it properly is work. You open a tool, you get a caption, you post it — no friction at all. Actually knowing whether that caption did anything for you means setting up a comparison, being patient, resisting the urge to declare victory after one good post. Most teams haven’t built that discipline yet, and honestly, why would they have? Until eighteen months ago nobody needed to ask “did the AI version outperform the human version” — a person wrote it, you judged the post on its own merits.
That 36% doesn’t worry me the way it might sound like it should. It’s not that people are careless. It’s that measurement always lags adoption — it happened with paid social, it happened with influencer marketing, it’s happening again here. The tools for using AI improved quickly. The habits for judging it haven’t caught up yet. They will.
Generative AI enables brands to produce more content at a lower cost, but greater reach or publishing volume can create a false sense of success. Which metrics should teams prioritize to determine whether AI-assisted content is strengthening trust, brand health, and customer relationships?
Reach is the metric AI is best at inflating and worst at explaining. If you’re only watching that number, you’ll feel great for months right up until you don’t. What I’d actually monitor are people saving or sharing the content, not just scrolling past it — that’s the difference between “I saw it” and “this was worth keeping.” Are the comments actual conversation, or just more noise under more posts? And, this one gets skipped constantly, does the brand still sound like one person after three hundred AI-assisted posts, or has the voice started drifting post by post without anyone noticing?
That last one is the sneaky one. Trust doesn’t collapse all at once — it erodes quietly, and by the time reach or sales show it, you’ve been losing it for months.
Metricool found that 66% of respondents involve AI in at least half of their content, while six in ten believe AI-generated content matches or exceeds human-created work. How should brands design experiments that fairly compare AI-assisted and human-created content without confusing correlation, platform algorithms, and content quality?
Keep everything fixed except the one thing you’re testing — same brief, same platform, same day and time, same person doing the final edit — and only swap who wrote the first draft. That’s it. Most “AI vs. human” comparisons floating around don’t do this. They compare an AI post from Tuesday against a human post from three months ago on a completely different topic, and then draw a conclusion.
Run it in pairs, run it for weeks, not for one post — a single post’s performance is mostly luck and algorithm mood that day, and you need enough volume to see past that. And be honest about what “matches human work” actually means. If AI content gets similar impressions but worse comment quality or lower saves, that’s not a tie, that’s a warning sign. A lot of that “six in ten” confidence, I suspect, is people looking at the surface number and stopping there.
Where does automation currently add the most value for social media teams, and which responsibilities, such as establishing brand voice, managing communities, or responding during a crisis, should remain firmly under human control?
The boring, high-volume stuff — scheduling, first drafts, resizing a video for six platforms, pulling a monthly report — that’s where automation is basically free time given back to a team. Nobody should be doing that by hand in 2026.
The line for me is judgment. Voice, community, and especially anything that’s a crisis — those need a human. Not because AI can’t draft something reasonable, it often can, but because the cost of one wrong tone in a sensitive moment is public and immediate, and a brand’s reputation isn’t something you get to A/B test. I’ll go back to something I’ve said before: being half-present or badly present on social media is like having a broken shopfront window. AI drafting a crisis response for a human to review in thirty seconds,great. AI hitting publish on that response by itself,never.
Metricool Studio allows users to generate reports, compare performance, analyze competitors, and identify posting patterns through natural-language prompts. As AI shifts from generating content to interpreting performance, how will this change the daily role of social media managers and marketing analysts?
It pushes the job upward. A huge chunk of a social media manager’s week still goes into pulling numbers into a slide and writing the sentence that explains what happened. If you can just ask “why did engagement drop last week” and get a real answer back in seconds, that whole chunk of the week basically disappears.
What’s left is the actual valuable part: deciding what to do with the answer. Which format to double down on, how to explain it to a client who doesn’t care about the mechanics, when the data is telling you something your gut disagrees with. I don’t think the role shrinks,it splits people into two groups pretty fast: the ones who were doing real analysis, and the ones who were mostly doing data entry with extra steps. Only the first group grows from here.
Metricool has adopted the Model Context Protocol, allowing AI assistants, automation platforms, and custom agents to interact with social media data and workflows. What opportunities does this create, and how can companies prevent autonomous systems from publishing inappropriate content, misreading analytics, or taking actions that damage a brand?
We were one of the first social media platforms to build on Claude’s MCP, and the reason is simple. It means Metricool’s data doesn’t just live inside our own interface anymore. A brand’s own agent, an automation platform, whatever a team is already using,they can now pull performance data or trigger a workflow directly, without us having to build every single integration ourselves. That opens up workflows we haven’t thought of yet, built by the people closest to the data.
The risk has to be designed in from day one, not patched on afterward. Practically publishing should default to a human approval step, not silent autonomous posting, especially anything client-facing. Permissions need to be scoped tightly such as an agent built to pull a report should not be able to trigger a post, period. And every action an agent takes needs a trail a human can check afterward. The protocol gives you the connection. The guardrails are still a decision every company has to make on purpose.
Metricool’s research suggests that traditional public interactions can decline even as overall engagement increases, driven by less visible actions such as clicks, swipes, views, and link taps. How should brands adapt when some of the most meaningful signals of attention are increasingly invisible to the public?
Likes and comments were never a great proxy for attention, and they’re getting worse. Most of the genuine interest today happens somewhere nobody else can see.Someone swipes through every slide of a carousel, watches a video to the end, screenshots something to send to a friend privately. None of that shows up publicly. All of it is real attention, arguably more valuable than a reflexive double-tap.
So brands need to get comfortable treating completion rate, swipe-through, and click-through as core numbers, not footnotes buried under likes. The practical shift is asking a different question altogether: not “how much engagement did we get” but “how much attention did we actually earn,” knowing most of that second number will never be visible to anyone outside your own dashboard.
Looking ahead, do you expect AI to remain a collection of assistants used by social media professionals, or will it develop into an operational layer that continuously plans campaigns, allocates resources, interprets performance, and recommends decisions across every platform?
An operational layer, and sooner than most people expect. The “ask AI one thing at a time” model is a phase, not the destination. What we’re actually building toward is AI sitting underneath the whole workflow such as watching performance across every platform continuously, catching what’s working before a human would notice, and recommending the next move.
Recommending, not deciding. I don’t think the end state is AI running campaigns with nobody watching. I believe it’s AI doing the relentless monitoring no team has the hours for, and handing a person a short, useful list instead of a wall of dashboards. The professionals who do well in that world won’t be the ones fighting the layer, they’ll be the ones who get good at pointing it out.
Thank you for the great interview, readers who wish to learn more should visit Metricool.
