A new test shows an Artificial Intelligence (AI) forecast beat the old sales plan in 15 of 15 live rollouts, cutting errors by 32% on average [1]. In the same month, climate experts warned a giant El Niño storm pattern could cost the world nearly a trillion dollars [2][3]. My read: Both numbers will land in the same planning meeting, but a better math model can only fix one of them.

The big picture:

The test checks something small but useful. It puts an AI forecast up against the number a company already trusted on its own data [1]. That old number could be a spreadsheet, a planner’s best guess, or the shared goal from Sales and Operations Planning (S&OP).

The AI won every time. It cut the error rate by 14% to 56% [1]. That is a real win, but it only fixes the normal noise a model can learn from the past.

A giant El Niño breaks that past pattern. Risk experts at Verisk Maplecroft warn this year’s ocean heat is already the strongest on record [2]. They say a repeat of the 1997 to 1998 El Niño could wipe out trillions of dollars [2].

The part I keep circling: Teams view a better forecast as a safety net. But a tighter math model does not sit between a copper mine and its shipping port.

By the numbers

  • 32% — Average error cut: The AI forecast beat the old metric across 15 live tests, with drops ranging from 14% to 56% [1].
  • 5.7 trillion US dollars — Past storm cost: The price tag of the 1997 to 1998 El Niño, which Verisk Maplecroft says this year could repeat or top [2].
  • $986 billion — 2027 loss guess: Oxford Economics says near-term global losses could hit a trillion dollars, growing to $7.2 trillion over six years [3].
  • 70% — Less manual work: Planning work fell by 70% on average, and one firm saw three-quarters of its forecast need zero human edits [1].

What I’d watch:

The leaders closest to the risk are already splitting these two jobs. The parent company behind retail stores T.J. Maxx and Marshalls told investors it relies on holding extra goods to match seasonal items with actual weather shifts [5].

That is a physical buffer answer, not a math answer.

Gartner’s October planning summit points at the exact same gap. Meeting notes warn that teams chase a better forecast, only to watch factories panic when a monthly plan hits a weekly schedule [4]. This causes fake demand spikes and rush orders [4].

What I am watching next:

  • The buffer line: Whether the budget that pays for an AI upgrade also pays for the extra space a record El Niño needs.
  • The risk map: Which sourcing teams have priced Latin American copper, coffee, and soy again before the December peak [2].
  • The ticking clock: How fast a firm can act once a signal moves, which Gartner flags as the gap between seeing a problem and fixing it [4].

The catch

I could be wrong to frame these as a single story. A firm that only sells stable, local goods may never feel the El Niño storm [1]. For them, an accuracy upgrade is a clean win.

The catch is what the test actually proves. Its wins come from shrinking errors on past trends the model has already seen. A record weather shock is, by definition, outside that past data.

Accurate math and physical buffers solve entirely different problems. A plan that pays for only one leaves the business exposed on the other.

At a glance

  • The Big Shift: A new test shows an AI forecast beat the old plans in 15 of 15 live rollouts. At the same time, risk experts warn a record El Niño storm will peak in December.
  • Why It Matters: Forecast accuracy is often sold as a safety net. But fixing past math errors does not protect a supply chain from a massive physical shock.
  • What I’d Watch: How planning teams divide their budgets between math and physical space.
  • The buffer: Held space and extra goods that absorb a physical shock no math model can see.
  • The clock: How fast a firm can act once a signal moves, which sets the true cost of a delay.
  • The risk map: Which supply lines into Latin America have been priced again before the December peak.
  • The Catch: The AI accuracy gain is real but limited. Better math and more physical buffer solve different problems, so a plan that pays for only one is exposed on the other.

Related reading

Sources

[1] DemandForecast.ai — “The Demand Forecasting Accuracy Report 2026” (October 2026) — https://demandforecast.ai/resources/forecast-accuracy-report [2] Verisk Maplecroft — “Record El Niño poses growing risk to Latin American supply chains, commodity output and infrastructure” (September 30, 2026) — https://www.maplecroft.com/solutions/supply-chain-risk/insights/latin-americas-vulnerability-to-super-el-nino-poised-to-upend-key-supply-chains/ [3] Oxford Economics — Super El Niño global cost outlook, via SupplyChainBrain (September 8, 2026) — https://www.supplychainbrain.com/articles/44808-super-el-nino-could-cost-global-economy-72t [4] Gartner — Supply Chain Planning Summit session agenda (October 5–6, 2026) — https://www.gartner.com/en/conferences/emea/supply-chain-planning-uk/sessions [5] Supply Chain Dive — “TJX CEO: Distribution model will help weather El Niño” (September 18, 2026) — https://www.supplychaindive.com/news/tjx-ceo-distribution-model-will-help-weather-el-nino/830665/