Where you could go next, and the order we’d pick

Hill 1: Physical AI and World Model data services

On the May 15 call, you said the signals are converging. That read is right, and the numbers explain why.

The data annotation market that currently generates most of your revenue is roughly $3 to $5 billion in 2026, depending on whether you count tools only or managed services. Mordor Intelligence puts the tools-focused segment at $3.07 billion. Verified Market Research sizes the full managed-services market at roughly $5.1 billion. That is the market you are largely selling into today.

Physical AI, meaning the embodied systems, robotics, autonomous vehicles, drones, and spatial computing platforms that need temporal annotation, trajectory prediction, 3D spatial reasoning, and multi-frame labeling, is an order of magnitude larger. Kaisore Research puts the broad Physical AI market at $81.4 billion in 2025, growing at 33.5% annually. The Robotics Center of Silicon Valley pegs the global robotics market alone at $38 billion in 2026, up 34% year over year. NVIDIA’s Jensen Huang called the broader Physical AI opportunity a $40 trillion labor automation play at CES 2026.

The ratio is roughly 15 to 20 times, depending on definition. Use the full managed-services annotation market as your denominator ($5 billion) and the broad Physical AI market as the numerator ($81 to 108 billion), and the ratio is 16 to 21 times. That is what we meant on the call. The exact multiple matters less than the structural point: you are currently selling into the smaller market while you are already capable of serving the larger one.

You are not at the base of the Physical AI mountain looking up. You are already on it. The robotics pilot underway could ramp to 300 people. You have done autonomous-truck, drone, and sensor-fusion work. Your SOC 2 Type II and ISO 27001 pass the table-stakes compliance gate that eliminates most emerging-market competitors from Physical AI bids. The two top criteria Physical AI data buyers name in 2026 are those exact certifications and “sim-to-real readiness.” You hold one of those two. The other is a packaging question, not a capability gap.

The timing wedge is real and time-bounded. Meta’s $14.3 billion investment in Scale AI created a direct conflict-of-interest problem for Scale’s enterprise customers. Google, OpenAI, Microsoft, and xAI all reduced or paused Scale engagements after the deal closed. Enterprise buyers sharing proprietary training data with Scale are, by extension, sharing pipeline access with a direct competitor. That window will not stay open indefinitely. Scale is rebuilding trust through a government pivot. New alternatives are entering. The neutral, bootstrapped, no-hyperscaler-conflict position is available to SmartOne right now, and the Physical AI sub-segment is where the buyer displacement is most acute.

The Physical AI pivot is not a new market entry. It is making intentional what you are already doing. The missing pieces are a named capability set, a Physical AI buyer map, and a partnership motion. Habib already has a warm connection to Physical Intelligence’s Ali Amin. That is the first call to make.

Hill 4: Defense-space data annotation (Phase 3, conditional)

This was named clearly on the call. “Conditional on a firewall structure, and not the focus of the work ahead.” That framing is exactly right. We are not recommending this as a primary hill.

The demand is real. The Pentagon’s Chief Digital and Artificial Intelligence Office awarded Scale AI a $500 million contract for data analysis and AI-assisted decision-making in May 2026, a fivefold increase from a prior $100 million deal. Defense data procurement is growing, and Scale’s Meta entanglement makes it politically complicated for some defense agencies to continue buying from them at scale.

A Canadian-headquartered firm with Malagasy operations can access US defense contracts, but only through a firewalled US subsidiary structure, with cleared facilities, ITAR compliance, and background investigation timelines that run 12 to 36 months before the first cleared contract can be performed. The structural cost is real. The timeline is 3 to 5 years to be genuinely competitive in this segment.

This is a Phase 3 conversation. If the founders decide to pursue it, the team page shows what activation looks like (Phase 3, Role 3.2). For now: back pocket. The Palantir relationship you already hold is the seed of that conversation when the time is right.

Hill 3: Europe Growth (Phase 2)

This is not a new market entry. You are already in France.

SmartOne has a small business and healthcare team covering Europe today, backed by SMARTONE EU SAS in Paris, where Habib has been president since November 2021. The 2022 to 2023 push did not land because, as Shahysta put it on the call, buyers “had no idea what we were doing.” That was a market-timing problem, not a country exit. The team stayed. The entity stayed. The client work stayed. Safran is a current client.

You’ve already done work for Safran and other French defense primes. That is a credential French enterprise and government buyers respect. The expansion play uses what you already have.

The market has moved in your direction. On the call you said: “the LLM and the fact that the market is now a little bit more fragmented, that people are used to now building their own toolbox has made it for countries like France to be able to have the gap that they had in terms of AI adoption go a little bit smaller.” The ICP you see in France in 2026 looks like the ICP you saw in the US in 2022 to 2023. That is a concrete, traceable shift, not optimism. The B5 research confirms it.

The research surfaces one finding that the call underplayed: the Palantir connection is stronger than it sounded. Palantir signed a multi-year renewal with France’s domestic intelligence agency (DGSI) in December 2025. A vendor that Palantir trusted with data annotation has already passed Palantir’s security and quality requirements. That chain of trust is a door-opener with French defense primes. You already have the Safran relationship. The B5 research identifies Safran as a warm path for a framework agreement that would let Safran bring SmartOne into new programs without re-qualifying each time.

Three regulatory frameworks activating in 2025 to 2026 are creating structural demand for compliant data vendors in France. DORA entered full enforcement in January 2025, requiring financial sector firms to document third-party AI risk. NIS2 became enforceable in October 2024. The EU AI Act reaches full application in August 2026, requiring documented data governance for high-risk AI systems. Your SOC 2 Type II and ISO 27001 are direct competitive advantages in a procurement environment where those certifications have become gates.

The additional asset you have that no competitor in the comparable set shares: you are both ESSEC alumni. ESSEC’s alumni network has 71,000 members and 125 global chapters. The people buying AI services at BNP Paribas, LVMH, Schneider Electric, and Airbus include ESSEC alumni. A warm introduction through that network compresses the French enterprise sales cycle from 12 to 18 months down to 6 to 9 months. You already have the French entity, the team, and the clients. The activation question is whether you resource this motion deliberately or let it continue ad hoc.

The Europe Growth motion sits in Phase 2 on the team page (Role 2.1). The practical trigger: North American revenue has stabilized enough that a structured French enterprise sales process can be resourced deliberately. Because the team and entity are already in place, the timeline from focused activation to a first signed framework agreement is 6 to 12 months, shorter than a cold-country entry would be.

Hill 2: Commercial enterprise account expansion (now, in parallel)

The immediate mandate on the call: secure revenue as fast as possible. It is the right mandate for the current moment. Converting warm leads to signed engagements, building on the robotics pilot, and pushing existing customers toward annual commitment structures instead of spend-per-month relationships. This work is happening regardless of which other hill gets attention.

The framing to hold onto: this is the floor, not the ceiling. A company that grows to $10 million and $15 million entirely through account expansion and word of mouth has built a better services business, but it has not solved the positioning problem, the concentration problem, or the market-size problem. Hill 4 buys runway for Hills 1 and 3. It is necessary. It is not sufficient long-term.

The specific near-term move that Hill 4 enables: if the robotics deal lands and the pilot converts, you are out of survival mode. That matters because every other conversation, the Physical AI positioning work, the France activation, the platform build, the alignment session, is easier to have when the company is not re-deciding whether it exists every 30 days.

Why this order

The sequencing is not arbitrary. It follows from your actual constraints.

Hill 2 runs continuously because you have no choice. Hill 1 starts immediately in parallel because it requires positioning and packaging work, not new capabilities or capital. You already have the people, the certifications, and the proof points. The job is to make them legible to Physical AI buyers.

Hill 3 waits for Phase 2 because France requires a structured sales motion, and structured sales motions require operational capacity you do not have while in survival mode. Running two separate enterprise sales processes across two continents simultaneously, with no dedicated sales function, while managing a robotics ramp, is the recipe for both processes going slowly and neither closing.

Hill 4 waits for Phase 3 because the structural requirements, the firewall entity, the cleared facilities, the compliance pipeline, take years to build and require capital allocation decisions that depend on the alignment question being answered. You do not start that process in survival mode.

The hill you take first is the one where your existing capabilities are already closest to what the buyer needs, where the market window is open today, and where the cost of entry is positioning and packaging rather than new capital. That is Hill 1.

Hill scorecard

Four hills, scored across the dimensions that matter most at your current stage. TAM figures are 2026 estimates from cited research.