Commercial Architecture Must Begin Before Product Launch

There's a version of building a new venture where commercial strategy waits until the product is ready, on the theory that you can't sell what doesn't exist yet. I've come to think that ordering is backwards for anything trying to enter a genuinely difficult market, and T57 — a pre-launch AI-native food ecosystem, where I lead sales and strategy — has been the clearest test of that belief I've had.

Why the commercial questions can't wait

Food and agricultural trade is fragmented and opaque, and it's hardest on the people with the least leverage — emerging-market producers and smaller enterprises who lack the infrastructure to reach serious buyers across borders reliably. Building a platform for that world raises commercial questions that are just as hard as the technical ones, and answering them late is expensive in a specific way: a product built without a clear view of which markets it's for, at what price point, through which partners, tends to end up generically useful and specifically compelling to nobody.

The questions that had to be answered before the product was fully ready: which markets first, and why those and not others? At what price, in economies with meaningfully different purchasing power — a single global rate is close to meaningless across markets this different? Through which partners, since credibility in unfamiliar markets almost always travels through relationships that already exist rather than through a platform's own reputation, which hasn't been earned yet? Waiting until launch to answer these means launching into markets chosen by default rather than by design.

What leading this actually involves day to day

In practice, my role has meant designing the commercial architecture ahead of the platform itself: the market-entry sequence, a pricing model calibrated to local purchasing power rather than one global figure, the partnership strategy, and the sales-operations infrastructure — research, workflows, follow-up, pipeline discipline — that has to exist before the first real deal can move through it. None of that waits for the product to be feature-complete, because a commercial operation that only starts thinking about these questions at launch is starting months behind where it needs to be.

The other significant part of the role is working day to day with the AI/ML team, and the rule I've held to there is that business questions come first: what do we actually need to know about a market, to what standard of confidence, decided before any model gets built to answer it. That ordering keeps the research honest and the engineering effort pointed at something a real commercial decision depends on, rather than at whatever's technically interesting to build.

What Gulfood actually tested, and what it didn't

At Gulfood 2026 in Dubai, I led an eight-person team through a deliberately structured activation — a digital signing workflow and a daily performance cadence, rather than the more traditional approach of collecting a stack of business cards and hoping some of them turn into something later. The team came back with more than 150 signed letters of intent and memoranda of understanding.

I want to be precise about what that number means and doesn't mean, because it's easy to let a large number imply more than it should. Letters of intent and memoranda of understanding are not revenue. They are not signed contracts. They are not customers in any sense a business would normally use that word. What they are is a genuine test of whether the commercial proposition lands with real traders when it's put in front of them directly, under no obligation, at a major industry event where they have plenty of other options competing for their attention. On that specific, narrower test — does this land — the answer at Gulfood was clearly yes. That's meaningfully different from proof of a working commercial model at scale, and I'd rather be explicit about the distinction than let the number do work it hasn't earned.

The discipline pre-launch work actually requires

Pre-launch commercial work is, more than anything else, an exercise in intellectual honesty. Every plan and projection at this stage is necessary to have — you can't operate without some model of what you expect to happen — but every one of them is also a hypothesis, not a proven fact, and the discipline is keeping a clear, actively maintained line between what's been validated by real evidence and what's still belief dressed up as a plan. That line gets blurry under pressure, especially the pressure to present confident numbers to people who want reassurance. Defining the business question honestly before the research starts — what would actually prove or disprove this belief — is, I've found, the cheapest and most effective quality control available at this stage, especially with AI-assisted research in the loop, where it's easy to generate confident-sounding analysis that answers a question nobody carefully specified in the first place.

Why purchasing-power pricing is harder than it sounds

One specific piece of this work deserves more detail, because it's easy to state as a principle and genuinely hard to execute well: pricing calibrated to local purchasing power rather than a single global rate. The naive version of this is straightforward currency conversion, which solves almost none of the actual problem, because the relevant comparison isn't "what does this cost in local currency" — it's "what does this cost relative to what a business in this market would normally pay for a comparable service, and relative to what margin structures already look like in that market's existing trade relationships."

Getting that right required genuinely understanding each target market's existing commercial norms, not just its exchange rate — which is research work, not a spreadsheet formula, and it's exactly the kind of question I insisted the AI/ML team treat as a business question to answer carefully rather than a parameter to estimate quickly. A pricing model that looks sophisticated but is built on a shallow understanding of local purchasing power is, in practice, worse than a simple flat rate, because it creates a false sense of precision around a number that's still fundamentally wrong for the market it's supposedly calibrated to.

Where this stands, honestly

T57 remains in its pre-launch phase. Commercial planning continues alongside product development, fundraising and partnership discussions, and I'm not going to describe any of that as further along than it is. The Gulfood activation was a genuine, meaningful test of the proposition. It was not a launch, and the work between a strong pre-launch signal and an operating commercial business is still substantially ahead.