The real company behind the World Model Data Services tagline: what you have built, why it is harder to see from the outside than it should be, and what changes when you close the packaging gap.
The positioning on your website says World Model Data Services. That is where you are going. The 2026 revenue says something narrower: high-accuracy human annotation at scale, anchored heavily in Amazon's AI development pipeline, with smaller pockets of robotics, LLM prompt work, and model evaluation growing at the edges.
That is not a contradiction. It is a company mid-pivot. The Physical AI thesis is real, and on the call you said it plainly: the signals are converging. But the marketed story has gotten ahead of the packaged revenue by about 18 months. You are presenting a destination while selling a journey.
The specific gap: your autonomous-truck, drone, and sensor-fusion work is real. The robotics pilot underway is real. The temporal annotation and 3D spatial reasoning capability is real. But buyers who visit the website or look up SmartOne cannot tell what is an active, scalable practice and what is an aspirational positioning line. The job is not to rebuild the identity. It is to close the distance between what you say and what you can demonstrate.
You put it well on the call: "sometimes this diversity, of what we did might hinder our capacity to clearly say to the market, you know, this is what we do, because saying we do everything is, I don't think it's ever a good thing to, to say." That is not a critique of the breadth. It is an acknowledgment that breadth needs a spine. The Physical AI spine is already there. It just needs to carry the weight.
You founded SmartOne in 2012 as a French-market call center in Madagascar. The original business had nothing to do with AI. It was about language, geography, and a lower-cost delivery model for French-speaking markets. You were about to shut it down.
In 2016, a meeting with an Amazon AI researcher changed the direction of the company. Amazon needed annotation volume. You had trained people, a management infrastructure, and a quality discipline already in place. Within three months you went from a test to 100 people working full-time on Amazon's AI data pipeline. The call center was gone. SmartOne was an annotation company.
For the next six years, you grew on Amazon demand, word of mouth, and what you called "just luck and, and some ad hoc like, outreach." No GTM motion. No sales infrastructure. Revenue re-decided every 30 days, one anchor customer accounting for the majority of it. The business worked. It also created the dependency you are now trying to reduce.
The Montreal move in 2023 was the first deliberate repositioning. You brought in a PhD CTO and a US-based VP of Sales. On the call you were honest about what happened: "we naively thought before coming here that, we would just get the technical expertise and, and turn into a product company." Neither hire stayed. Shahysta returned as CEO in January 2026 with a single mandate: secure revenue as fast as possible.
That is where the company sits today. Fourteen years of operational depth, a decade of enterprise accuracy discipline, a 1,000-person workforce that knows how to do this work, and a packaging story that still reads like a services bureau that discovered AI three years ago. The story and the capability are not in the same place.
The assets that make SmartOne genuinely differentiated are not on the website in any useful form. Here is what you actually have:
None of these appear in the way a buyer reads a company. They are real. They are valuable. They are invisible.
ZoomInfo lists SmartOne at $210.1M in revenue. You told us $6M on the call. That gap is not a problem to hide. It is a story to tell.
Third-party aggregators like ZoomInfo generate revenue estimates by scraping LinkedIn headcount and multiplying by industry averages. A company with 1,000 employees in an AI data services category gets assigned a revenue estimate calibrated to US or EU-headquartered firms with US or EU salary structures. SmartOne's Madagascar delivery model collapses that assumption. The formula is wrong because the business is not what the formula assumes.
The framing worth testing: own the gap. ZoomInfo's number is built on headcount-times-industry-average; your number reflects the actual cost structure of a Madagascar-anchored business. That gap is your story to tell, not a hole to defend. And then you lead with the list: 1,000 trained full-time annotators, a decade of enterprise accuracy discipline, SOC 2 Type II, ISO 27001, Palantir-adjacent work, French defense prime work, and a robotics pilot underway.
That is a stronger room presence than quietly hoping the buyer does not look it up. The ZoomInfo gap is actually evidence of how radically mis-measured this category of company is by outsiders. Owning that framing is an offensive move, not a defensive one.
The goal is not to explain away the gap. It is to make the gap the first line of the credibility argument.
This is a reframe worth pressure-testing with you. Is owning the gap something you want to do, or would you rather just correct ZoomInfo quietly?
That is not a confession. It is the clearest possible statement of where the work is. The capability exists. The presentation does not match it yet. That is a fixable gap, not a structural one.