For AI-driven startups, the gap between what a model can calculate and what a business can actually execute becomes even more visible. Algorithms can surface insights, but only teams that pair those outputs with real operational context can turn them into decisions that people trust and follow. The fastest-growing companies in this space are the ones that respect both sides of the equation – engineering strength and real-world complexity.

In FMCG, business logic is rarely documented. It sits in people’s heads, historical exceptions, regional habits, inherited constraints, and unspoken rules that “everyone just knows.” In practice, this means the same SKU may appear under one unit of measure in Europe and several in LATAM, or a customer treated as Tier 1 in one region may be Tier 3 in another. Decisions that look simple on a whiteboard quickly split into exceptions the moment real data enters the picture.

This is why adopting new solutions – whether they come from ambitious startups or established global companies – often proves more complex than creating them in the first place. Coding a feature is manageable; translating operational trade-offs into product logic, without losing the nuance behind real-world decisions, is the true challenge.

Technology without context doesn’t work

Supply chain decisions never exist in isolation. Even the most straightforward question, like “who to ship first,” depends on logistics windows, service history, assortment volatility, production rhythms, penalties for late deliveries, channel priorities, and regional calendars.

These choices are shaped by trade-offs – decisions in which improving one operational outcome naturally puts pressure on another. In real operations, this is easy to recognise: pushing service levels up usually means holding more inventory, cutting costs often makes the network less flexible, and giving priority to a vital customer slows someone else down. Supply chains live in this constant balancing act every day, so any technology built for them has to work with that reality, not against it.

Teams that lack industry experience often underestimate how these pressures shape day-to-day decisions. A tool may look flawless in a clean demo environment, but once it encounters shifting hierarchies, conflicting calendars and years of local Excel “fixes,” its logic starts to fall apart. Even the most advanced algorithms feel disconnected when they don’t reflect how operations actually function.

Why domain knowledge changes everything

Domain knowledge means understanding how the industry interprets signals, resolves conflicts between cost and service, and adjusts decisions when materials, equipment or staffing become constraints.

In FMCG, a few decisions are purely technical. Each one sits between global rules and local realities. People who have worked inside this environment know when a rule is essential and when it is an inherited habit. They can trace not just what decision was made, but also why – and design technology that supports that logic rather than flattening it.

This is why the strongest product leaders come from hybrid backgrounds: operations or planning experience combined with product thinking. They bridge the mental gap between business behaviour and system behaviour.

Turning decisions into code: codifying trade-offs

A core challenge in supply chain technology is converting human judgment into structured system behaviour. Decisions that look instinctive – which customer to protect during shortages, which SKU to produce first, when to adjust a forecast – are often long-standing patterns shaped by experience.

These patterns can be expressed in concrete system elements: priority rules, weighting factors, tie-break hierarchies, allocation thresholds, classification tables, and scenario templates. Protecting strategic customers may become a weighted matrix; production choices turn into sequencing logic; forecast adjustments follow specific event- or channel-based conditions.

Codification does not replace expertise. It transforms it into a consistent decision model. When this work is done well, a planning tool becomes a living representation of how the business thinks. Decisions become repeatable instead of personality-driven, and scenario exploration becomes structured rather than intuitive.

It’s important to note that some trade-offs should stay out of the system entirely. These include short-term commercial pushes (“protect this retailer during the campaign”), seasonal exceptions set by leadership, margin negotiations tied to individual contracts, or one-off disruptions such as supplier outages. Encoding temporary or political decisions creates rigid behaviour that outlives the reasons for their introduction. Systems should model structural patterns that repeat over time – not moment-specific exceptions.

Why companies struggle: it’s not the technology

Transformation failures rarely come from weak software; they come from a lack of readiness.

Data is often the first barrier. If product, customer and transaction structures differ by region, if historical data no longer matches current hierarchies, or if calendars conflict, no platform can produce a coherent plan. Technology exposes these inconsistencies; it cannot repair them.

The second barrier is the lack of operational understanding in product teams – something many startups underestimate when entering the supply chain for the first time. Builders must grasp why planners override recommendations, how lead times behave in reality, what service levels mean by channel, and when an optimal plan is operationally impossible. Without that insight, tools struggle to earn adoption.

 What AI-focused startups should remember

1. Domain expertise beats algorithmic novelty. Without it, adoption becomes the bottleneck

2. Understand the trade-offs. Supply chains run on competing priorities – products must mirror this reality

3. Codify judgment, not just data. Decision logic is where true differentiation lies

4. Build for messy reality. Design for shifting hierarchies, calendars, exceptions and local practices

5. Bring people who’ve lived the problem. Hybrid thinkers – part operator, part product builder – will define the future of supply chain tech