November 16, 2026
AI’s Data Problem Needs Robots
Last week, an Italian startup called Mosaico raised €7.5m to tackle one of AI's least glamorous problems: getting the right data into the machine in the first place. Not the sexy bit. Not the demo-day bit. The plumbing: data engineering.
And honestly, good.
Because while half the market is still acting like sticking 'AI' on a slide deck is a strategy, the grown-ups know the real problem is far more boring and far more valuable. If your data is patchy, expensive, slow to collect or unusable in the real world, your AI ambitions are basically a very expensive mood board.
According to Sifted's report on Mosaico's funding round, the company is building robotics software to improve data collection for AI systems. That matters more than most founders, CTOs and hiring teams realise. Source: Sifted, 'Exclusive: Italian robotics startup Mosaico raises €7.5m to tackle AI's data problem'.
AI does not have a model problem
Let me be slightly provocative: most firms do not have an AI model problem. They have a data operations problem wearing an AI costume.
The models are getting cheaper, faster and more accessible. Open source is improving. Tooling is everywhere. Every second person on LinkedIn is apparently an AI strategist now, which should worry us all a bit.
But high-quality, relevant, well-structured data tied to actual workflows? That is still hard. Very hard.
This is especially true in sectors where the physical world gets involved. Robotics, infrastructure, industrial settings, logistics, energy, heritage environments, field operations. Messy places. Real places. Places where the data does not arrive neatly labelled in a pristine CSV file like a gift from the gods of machine learning.
Mosaico's bet is simple and smart: if AI depends on data, then improving how that data is captured is not a side quest. It is the game.
The unsexy layer is where value hides
Investors love a grand narrative, and AI has supplied enough of those to last us until the next hype cycle. But under every big AI promise sits a stack of unsexy dependencies:
- data collection
- data quality
- data labelling
- workflow integration
- infrastructure
- governance
Most companies still underestimate how painful those layers are. They want the chef's kiss AI outcome without buying the ingredients.
It is a bit like wanting Premier League football with Sunday league fitness. Ambition is lovely. Reality still has opinions.
That is why this raise is interesting beyond robotics. It signals that serious people are putting money behind tools that solve foundational constraints, not just flashy interfaces.
And in hiring terms, I would pay very close attention to that.
What this means for data engineering and tech hiring
Whenever a market shifts, hiring lags behind the truth. Companies keep hiring for yesterday's org chart while trying to build tomorrow's capability.
We're seeing that already in AI.
Many businesses say they need AI talent. Fair enough. But when you dig a little deeper, what they actually need is one of three things:
- better data engineering
- stronger platform and cloud capability
- people who understand how software meets physical or operational systems
In other words, the bottleneck is often not a superstar prompt engineer in a puffer gilet. It is the less glamorous, more commercially valuable talent that can make data usable at scale.
If you are hiring into AI-adjacent teams, ask yourself:
- Do we need someone to build models, or someone to make our data fit for purpose?
- Have we over-indexed on innovation theatre and under-invested in infrastructure?
- Who owns the messy middle between operations, data and product?
- Are we hiring for headlines or outcomes?
Those questions save time, money and a surprising amount of executive embarrassment.
The rise of hybrid technical operators
One of the biggest hiring gaps over the next few years will be hybrid talent.
Not just pure software engineers. Not just pure data scientists. And not just domain people with a passing interest in AI because they used ChatGPT twice and now feel unstoppable.
I mean people who can operate across boundaries:
- data and infrastructure
- software and robotics
- product and operations
- commercial goals and technical reality
These are the people who spot where value gets stuck. They are rarely the loudest in the room, but they are often the reason things actually work.
For mission-led startups and scale-ups especially, this matters. You do not have the luxury of bloated teams or endless experimentation. You need people who can connect systems, make decisions and build things that survive contact with reality.
That requires a more mature hiring approach than simply chasing whatever title is trending this quarter.
Founders: stop hiring the story, hire the constraint
If there is one lesson here, it is this: hire against your biggest operational constraint, not your most exciting strategic narrative.
That sounds obvious. It is not. Plenty of businesses still recruit based on what sounds impressive to investors, boards or the internal Slack channel.
But if your constraint is dirty data, fragmented systems, poor telemetry, weak cloud architecture or disconnected teams, then hiring an AI Lead without fixing the substrate is like buying a racing car for a road made of soup.
A better framework is this:
The Constraint-First Hiring Check
- What is slowing delivery right now? Be brutally honest.
- What technical capability is missing? Not what sounds cool. What is actually absent?
- What role would remove friction fastest? Think leverage, not prestige.
- What outcomes should this person own in 12 months? If you cannot answer that, pause the search.
Simple. Not always comfortable. Very useful.
Why this matters beyond robotics
Mosaico may sit in robotics software, but the underlying lesson travels well.
The next wave of AI value will not just come from better models. It will come from better systems around the models. Better collection. Better integration. Better decision-making. Better teams.
That has implications for:
- CTOs deciding where to place technical bets
- COOs trying to operationalise AI without creating chaos
- Heads of People building capability in a noisy market
- Founders trying to avoid expensive hiring mistakes dressed up as innovation
The firms that win will not necessarily be the ones shouting loudest about AI. They will be the ones quietly fixing the bottlenecks everyone else is ignoring.
That is usually where the money is. And, incidentally, where the best hiring opportunities sit too.
What to do next
If you are building in AI, robotics, data or infrastructure-heavy environments, do these three things this quarter:
- Audit your bottlenecks. Identify whether your real problem is models, data, infrastructure or execution.
- Re-write at least one job brief. Strip out the hype and define the actual capability needed.
- Map your hybrid talent gaps. Look for roles sitting awkwardly between teams. That awkwardness is often your growth constraint.
And if you are six months into a search wondering why the 'perfect AI hire' has not materialised, there is a decent chance the market is not the issue. The brief is.
Mosaico's raise is a funding story on the surface. Underneath, it is a signal. AI's future will be built by companies willing to solve hard, unfashionable problems. The data problem is one of them.
Turns out the people fixing the pipes may end up owning the building.
Back to news