AI Playbook
for Inventory Management

You are a demand planner reviewing forecast accuracy for the period.
Data: [PASTE: SKU | Forecasted demand | Actual demand | Variance units | Variance % | Category | Channel]
Analyze:
Overall forecast accuracy — MAPE (Mean Absolute Percentage Error) for the period
Best-performing categories — lowest forecast error
Worst-performing categories — highest error; what drove the miss?
Bias check — are we consistently over-forecasting or under-forecasting? Which categories?
SKUs with >50% forecast error — flag for manual review in next cycle
Output: Forecast accuracy report. End with: top 3 specific actions to improve accuracy next period — not generic advice, specific changes to data inputs, methods, or review cadence.
Statistical Forecast Build

You are a demand planner building a statistical forecast for the next 3 months.
Historical demand data: [PASTE: SKU | Month | Units sold — minimum 12 months of history]
For each SKU:
Identify demand pattern: trend (growing/declining) / seasonal / lumpy / stable
Select appropriate forecasting method based on pattern: moving average / exponential smoothing / trend-adjusted / seasonal decomposition
Generate monthly forecast for next 3 months
Calculate forecast confidence range — upper and lower bounds
Flag SKUs where historical data is insufficient for statistical forecasting (<6 months history or >40% of periods with zero demand)
Output: Forecast table by SKU and month. Method used per SKU. Confidence ranges. Flag list for manual forecast override.
Seasonal Demand Analysis

You are a demand planner analyzing seasonal demand patterns.
Historical data: [PASTE: SKU or product family | Month | Units sold | Year — minimum 2 years of history]
Analyze:
Seasonal index for each month — actual month ÷ average monthly demand × 100
Peak months — which months consistently over-index and by how much?
Trough months — lowest demand periods; inventory and staffing implications
Year-over-year trend — is overall demand growing or declining independent of seasonality?
Seasonal variation by product family — are all categories equally seasonal or do some have flat demand?
Output: Seasonal index table by month and family. Peak/trough calendar. Recommended inventory build-ahead quantities for peak season. Build start date.
New Product Demand Estimate

You are a demand planner building the initial demand estimate for a new product.
Product data: [DESCRIBE: Product description, target customer segment, pricing, channel, launch date, any comparable products currently in your range, competitive products in the market]
Build the estimate using:
Analogy method — identify the closest comparable product in your range; use its launch ramp as a baseline
Market sizing — estimated addressable customers × estimated purchase frequency × estimated attach rate
Cannibalization assessment — will this product take demand from existing SKUs? Quantify.
Channel ramp — how long before each channel reaches full distribution? (retail shelf, ecommerce, direct)
Scenario range: Conservative / Base / Optimistic — with stated assumptions for each
Output: 12-month demand estimate by scenario. Key assumptions listed and rated high/medium/low confidence. Recommended initial stocking quantity based on base case + safety buffer.
Promotional Uplift Forecast

You are a demand planner forecasting the demand impact of an upcoming promotion.
Promotion data: [DESCRIBE: Product, promotion type (price discount / BOGO / feature display / combination), discount depth, channels, start and end date, any similar past promotions]
Historical promotion data (if available): [PASTE: Past promotion | Product | Discount % | Duration | Baseline sales | Promoted sales | Uplift units | Uplift %]
Estimate:
Baseline demand during the promotion period (what would sell without the promotion)
Promotional uplift — units above baseline expected; use historical data if available, or industry benchmarks
Pantry loading estimate — how much of the uplift is forward buying that will reduce post-promo demand?
Post-promotion demand dip — estimated weeks of below-normal demand after promotion ends
Total incremental units over the full pre/during/post window
Output: Promotional demand plan by week. Net incremental units. Recommended inventory build before promotion starts.
S&OP Meeting Demand Pack

You are a demand planner preparing the demand review section of the monthly S&OP meeting.
Data: [PASTE: Product family | Last month actual | Last month forecast | Forecast error % | This month forecast | Next 3 months forecast | Any known demand changes or market events]
Build the S&OP demand review pack:
Performance vs. last month’s plan — families where actual demand deviated >10% from forecast; explain why
Updated forecast — current view of next 3 months by family; highlight changes from last month
Demand risks — events or customer intelligence that could reduce demand below forecast
Demand opportunities — upside scenarios or new business not yet in the plan
Consensus recommendation — proposed demand plan for S&OP team sign-off
Output: One-page S&OP demand pre-read. Each section max 5 bullets. Suitable for distribution 48 hours before the meeting.
Forecast Override Analysis

You are a demand planner reviewing manual forecast overrides applied by the commercial team.
Override data: [PASTE: SKU | Statistical forecast | Override value | Override reason | Who applied | Date applied]
Analyze:
Override bias — are overrides consistently higher or lower than statistical forecast?
Override accuracy — where data is available, compare overridden forecast to actual demand; did the override improve accuracy?
Override by source — which team members or regions apply the most overrides? Are their overrides more or less accurate?
Overrides without documented reason — flag these; undocumented overrides cannot be reviewed or learned from
Recommendations — which categories or sources of override add value vs. add noise?
Output: Override accuracy analysis. Recommendation: which overrides to retain as a process vs. which to challenge or require evidence for.
Demand Sensing Review

You are a supply chain analyst reviewing short-term demand signals for replenishment decisions.
Short-term demand data: [PASTE: SKU | Rolling 4-week actual demand | Rolling 4-week forecasted demand | Week-over-week trend | Any known events (promotions, new listings, competitor OOS, weather)]
For each SKU showing a significant signal:
Classify the signal: genuine demand shift / promotional pull-forward / stockout at customer / one-time order / noise
Assess whether to update the near-term replenishment plan based on the signal
Flag signals that are inconsistent with the longer-term forecast — may indicate a broader demand change
Recommend: adjust replenishment now / monitor for one more week / escalate to demand planner
Output: Demand signal review. SKUs requiring replenishment adjustment today. Escalation list for demand planning review.
Customer Order Pattern Analysis

You are an inventory analyst reviewing customer ordering patterns to improve forecast accuracy.
Order history: [PASTE: Customer | SKU | Order date | Quantity | Order frequency (how often they typically order)]
Analyze:
Order regularity — do customers order on a predictable cycle or erratically?
Order size variability — are order quantities consistent or highly variable?
Lumpy demand — customers who order infrequently but in large quantities (hard to forecast)
Lead time sensitivity — customers who change orders inside your replenishment lead time
Segmentation: forecast with confidence / require longer horizon commitment / safety stock buffer needed
Output: Customer order pattern segmentation. Recommended forecasting approach and safety stock policy per segment.
Long-Range Demand Plan

You are a supply chain planning manager building the 12-month demand plan.
Inputs: [DESCRIBE: Business growth targets (revenue or volume), planned new product launches, planned product discontinuations, key market assumptions, sales team bottomup input (if available)]
Historical baseline: [PASTE: Product family | Last 12 months actual by month]
Build the 12-month plan:
Baseline extrapolation — project last year’s actuals forward with trend adjustment
Growth adjustments — apply business growth targets by family
New product additions — add launch ramp estimates for planned new products
Discontinuation removals — fade out demand for products being discontinued
Reconcile top-down (business target) vs. bottom-up (sales input) — flag gaps >15% for alignment meeting
Output: 12-month demand plan by product family and month. Top-down vs. bottom-up reconciliation. Assumptions summary. Items requiring commercial team alignment.
Forecast Accuracy Improvement Plan

You are a demand planning manager building a plan to improve forecast accuracy.
Current state: [DESCRIBE: Current MAPE by category, key sources of forecast error identified, current forecasting process and tools, team size and capability] Target: Reduce MAPE from [CURRENT %] to [TARGET %] within [TIMEFRAME].
Build an improvement plan covering:
Root cause analysis of current forecast error — which categories drive the most error and why?
Data improvement opportunities — additional data inputs (POS data, customer inventory, leading indicators) that would improve accuracy
Process changes — improved collaboration with sales, marketing, and customers
Tool or model improvements — better statistical methods, machine learning, or planning software
Quick wins (30 days) vs. structural changes (90+ days)
Output: Improvement plan with actions, owners, and expected MAPE improvement from each. Timeline. Total expected improvement if all actions executed.
Collaborative Forecasting with Key Customers

You are a supply chain manager preparing for a collaborative forecasting session with a key customer.
Current situation: [DESCRIBE: Customer, products they buy, current forecast accuracy for this customer, any recent service failures, planned promotions or events]
Build the collaborative session agenda:
Performance review — how accurate was the last shared forecast? Where did we miss?
Customer demand outlook — what does the customer expect to sell over the next 13 weeks? What events or risks do they see?
Inventory review — what is the customer’s current stock level at their DC? Any risk of out-of-stock or overstock at their end?
Promotional calendar alignment — upcoming promotions we need to plan jointly
Agreed forecast — document the consensus forecast and ownership of any disagreements
Output: Collaborative forecasting agenda + template for capturing agreed forecast. Format for sharing with customer in advance of meeting.
Intermittent Demand Handling

You are a demand planner reviewing SKUs with intermittent or lumpy demand.
Data: [PASTE: SKU | Monthly demand for last 12 months (include zeros) | Unit cost | Lead time (days)]
For SKUs with intermittent demand (many zero-demand periods):
Calculate demand frequency — % of periods with non-zero demand
Calculate average demand when demand occurs (ADI — average demand interval)
Classify: smooth / erratic / intermittent / lumpy (use Syntetos-Boylan classification if possible)
Recommend forecasting approach: Croston’s method / moving average of non-zero periods / min/max with manual review
Recommend stocking policy: stock and replenish / stock-to-order / do not stock (order on demand only)
Output: Intermittent demand classification table. Stocking policy recommendation per SKU. Estimated inventory reduction from reclassifying do-not-stock items.
Demand Plan Risk & Opportunity Register

You are a demand planning manager maintaining the demand risk and opportunity register.
Current demand plan: [PASTE: Product family | Current 3-month forecast | Last updated]
Risks and opportunities: [PASTE or DESCRIBE: Event | Type (risk/opportunity) | Products affected | Potential volume impact | Likelihood | Timeline | Source]
For each item:
Quantify the demand impact — units and $ revenue at risk or opportunity
Assess probability — high (>70%) / medium (30–70%) / low (<30%)
Expected value = Impact × Probability
Recommend: build into baseline forecast / hold as upside or downside scenario / monitor
5 Owner and next review date
Output: Risk/opportunity register ranked by expected value. Total demand at risk (downside). Total upside opportunity. Items to raise at next S&OP meeting.