If you sell physical products, you need to guess future demand. You’re walking a tightrope between two undesirable outcomes, and keeping your balance has a huge impact on future revenue. If you understock your product, you’re losing sales, but overstock means using storage and shelf space inefficiently.

Stocking depends on a huge number of variables, but most companies make stocking decisions based on last year’s performance. AI tools are changing this approach, with advanced models able to pick up on thousands of small variables and market fluctuations. So, how can new AI tools potentially transform your supply chain planning? Here are seven signals, including where the data comes from and how to get it to work for you.

1: CRM Pipeline Velocity

This refers to how fast your sales deals are moving. When deals move faster and close sooner, you can expect more orders in the weeks ahead. When deals stall, near-term orders soften. AI tools can track these shifts across thousands of opportunities at once and translate them into predicted stocking needs per SKU.

You’ll need to feed your AI tools the following kinds of data:

• Stage entry timestamps
• Deal amounts
• Win rates
• Sales cycle length
• Conversion history

The trouble is that this commercial data often sits across fragmented systems. Deal records live in the CRM, call transcripts in the conversation intelligence tool, account data in the enrichment provider, and the rest in the revenue operations stack.

Platforms like GTM AI unify these sources into one context and intelligence layer, so your forecasting model reads from a single clean feed instead of a patchwork that breaks when one system changes.

2: Promotion Impact

Every price change or scheduled promotion changes demand, but the impact of promotion is something that planners frequently overlook. For an AI model to account for promotion calendars, it should be fed the following data:

• A structured calendar of list prices
• Discount events
• Coupon codes
• Bundle offers

The AI learns the elasticity for each SKU, meaning the percentage change in units sold per percentage change in price. Once the model knows a 15% promo is running next week, the resulting spike stops looking like unexplained demand and is built into the forecast.

Research from Javad Feizabadi, on hybrid forecasting methods confirms that incorporating promotion variables outperforms forecasts based on past sales alone.

3: Logistics Lead Times

A forecast is only useful when you can act on it. So if your suppliers are slow, you need to account for the delay. If a supplier’s lead time drifts upward, your reorder point needs to shift along with it. AI models are adept at adjusting to changes in the supply chain. If you feed AI models the following data, they’ll be great at accounting for these kinds of fluctuations:

• Carrier ETAs
• How long containers sit at the port before unloading
• Supplier shipment confirmations
• Customs clearance data

Combining planner expertise with AI forecasting improves reliability for products with volatile demand patterns, such as fashion and high-tech items with short life cycles.

4: Competitive Catalog Shifts

When a rival launches a comparable product or cuts prices, it affects your demand curve almost immediately. In some industries you might have hundreds of competitors, so keeping tabs on them manually is difficult.

AI agents can help, using web-scraping tools and marketplace APIs to catalog competitor SKUs, prices, and stock status as they change.

5: Macro Indicators

When you’re focused on bringing specific products to market, it’s easy to lose sight of the bigger picture. But if you fail to account for factors like consumer confidence, unemployment, and housing starts, you’ll never be able to accurately predict your market.

AI models can keep track of many macroeconomic indicators at once, given the right data, such as the feeds offered by central banks and national statistics offices.

AI Helps Unify Demand Planning

No single signal is enough on its own. The real gain comes from feeding several into one model and letting it weigh them against each other. Start with two or three feeds you already have clean access to, check the lift against your current forecast, then expand from there once the approach has proven itself.

If you’re interested in learning more about how AI is affecting supply chain management, see our other blog posts.