Stock the Right Items at the Right Time

AI Demand Forecasting and Inventory Optimization

It's Tuesday morning. Your autumn/winter buy is due to the planning team by Friday, and you're staring at last season's sell-through reports, a spreadsheet of supplier lead times, and a gut feeling that quilted jackets are going to be big. Your junior buyer pulled three years of sales data into Excel. Your merchandising manager flagged that one of your top jacket SKUs went out of stock six weeks before Christmas last year, costing an estimated £40,000 in lost sales. Meanwhile, you have 18% excess inventory in knitwear sitting in the warehouse from a trend that peaked faster than anyone predicted. This is the exact situation AI demand forecasting is built for, and the good news is you don't need a data science degree to use it.

What Demand Forecasting Actually Means for Merchandisers

Demand forecasting is the process of predicting how much of a product customers will want, when they'll want it, and in which sizes, colors, or variants. Traditional forecasting relies on historical sales data, buyer intuition, and static spreadsheet models. These work, up to a point. The problem is that fashion demand is notoriously volatile. A single TikTok video can spike demand for a specific boot style by 300% in 48 hours. A warm October can flatten coat sales for an entire quarter. Static models built on last year's numbers simply can't react fast enough. AI forecasting tools analyze far more variables simultaneously, past sales, returns data, search trends, competitor activity, weather patterns, and social media signals, and they update continuously rather than once a season.

For non-technical professionals, the key mental shift is this: think of AI forecasting less like a calculator and more like a very well-briefed analyzt who has read every sales report, studied every trend report, and never sleeps. You still make the final call. You still apply your knowledge of your customer and your brand. But instead of spending three days manually pulling data, you're spending three hours reviewing AI-generated recommendations and applying your judgment on top. Tools like Edited, WGSN Instock, and Toolio are purpose-built for retail merchandising. General-purpose AI tools like ChatGPT Plus and Microsoft Copilot can also play a supporting role when you're working with your own exported data in spreadsheets or documents.

Inventory optimization is the closely related discipline of deciding how much stock to hold, where to hold it, and when to reorder. Get it wrong in one direction and you're discounting aggressively in the January sale to clear excess stock, destroying margin. Get it wrong in the other direction and you're turning away customers who want your bestselling denim jacket in a size 12. AI tools can model both risks simultaneously, factoring in your supplier lead times, your warehouse capacity, your planned promotions, and your sales velocity by channel. The output isn't a magic number, it's a range of scenarios with probabilities attached, which is genuinely more useful than a single point forecast that's almost certainly slightly wrong.

How AI Reads the Signals You're Already Missing

Edited, another platform widely used in retail merchandising teams, tracks competitor pricing and inventory in near-real time. If three of your competitors have just gone out of stock on a key product type, that's a demand signal, customers who can't get it there will come looking for it with you. Edited's AI surfaces these moments automatically. For smaller retailers and brands who can't afford enterprise platforms, Google Trends, accessed through a browser with no technical knowledge required, gives you free search-demand data by region and time period. Pair that with a ChatGPT Plus conversation about your sales data and you have a lightweight but genuinely powerful forecasting workflow that costs around $20 a month.

Click to enlarge

Choosing the Right Tool for Your Forecasting Need

Not every forecasting need requires an enterprise platform. The right tool depends on your budget, your team size, and what specific problem you're trying to solve. A head of merchandising at a £50M turnover fashion brand has different needs, and a different budget, than a buying manager at a 12-person independent retailer. The table below maps the main tool categories to realiztic use cases, so you can identify where to start without committing to a lengthy software evaluation process.

Putting AI Forecasting to Work This Week

The most common mistake merchandising teams make when they first try AI tools is treating them like a search engine, asking vague questions and getting vague answers. The professionals who get real value out of tools like ChatGPT Plus or Microsoft Copilot in Excel are the ones who bring specific, structured inputs. That means uploading actual data files, giving the AI context about your business (your customer demographic, your price architecture, your key sales channels), and asking for output in a format you can actually use, a table, a ranked list, a scenario comparison. The more specific your input, the more useful the output. Think of it as briefing a new analyzt: the better your brief, the better their work.

A practical starting point for any merchandiser is what you might call a 'sell-through audit prompt.' Export your last full season's SKU-level sales data from your retail system, most ERP or POS platforms (including Shopify, Lightspeed, Cin7, and Microsoft Dynamics) let you do this as a CSV or Excel file with a few clicks. Upload that file to ChatGPT Plus and ask it to rank your SKUs by sell-through rate, flag anything that went out of stock before the last four weeks of season, and identify the three product attributes (color, fit, fabric type) most correlated with high sell-through. This is genuinely 30 minutes of work that replaces a half-day of manual analyzis.

For teams already using Microsoft 365, Copilot in Excel is worth exploring first because it requires zero change to your existing workflow. Open your sales spreadsheet, click the Copilot button in the ribbon, and type your question in plain English: 'Which product categories had the highest return rates last season?' or 'Create a chart showing weekly sales velocity for outerwear from September to December.' Copilot writes the formula or builds the chart for you. You don't need to know how pivot tables work. You don't need to know any Excel functions. You describe what you want to understand, and Copilot does the mechanical work. This is particularly useful for buying and merchandising managers who own the strategy but don't want to spend hours in spreadsheets.

How This Applies Differently Across Roles

• Buyers: Use AI to stress-test your buy quantities before sign-off. Upload your draft OTB and ask ChatGPT to model what happens to margin if sell-through drops 10% below plan in your highest-risk category.
• Merchandisers: Use Copilot in Excel to automate your weekly trading reports. Instead of manually updating charts, describe the report you need and let Copilot build it from your updated data.
• Planning Directors: Use AI to run scenario comparisons quickly, 'show me the inventory and margin impact of three different intake phasing options for Q4', so you can walk into a board conversation with modeled options rather than a single plan.
• E-commerce Managers: Use ChatGPT to analyze your returns data alongside sales data. High return rates on specific SKUs often signal a demand forecasting problem, you're selling the wrong sizes or the wrong product to the wrong customer.
• Small Business Owners / Independent Retailers: Start with the free tier of Google Trends plus ChatGPT Plus. Search your key product categories in Google Trends to see demand seasonality, then use ChatGPT to help you interpret what that means for your buy timing and depth.
• Visual Merchandisers: Use AI to connect sales velocity data to your floor layout decisions. If AI shows that a category is selling fastest in its first two weeks on floor, that's an argument for more prominent placement and faster replenishment cycles.
• Wholesale / B2B Merchandisers: Use AI to help you analyze order patterns across your retail accounts. Upload order history and ask for patterns, which accounts over-order and return, which consistently under-order and reorder, and which categories perform differently by account type.

Click to enlarge

Key Takeaways from Part 1

• AI demand forecasting analyzes more variables simultaneously than any spreadsheet model, sales history, search trends, competitor signals, weather, and social data, and updates continuously rather than seasonally.
• You don't need technical skills to use these tools. ChatGPT Plus, Microsoft Copilot in Excel, and Google Gemini in Sheets all work through plain-language conversation with your existing exported data.
• Purpose-built platforms (Toolio, WGSN Instock, Edited) offer the deepest fashion-specific functionality, but general-purpose AI tools are a practical and affordable starting point for most teams.
• The quality of your AI output depends directly on the quality of your input. Bring specific data, real context about your business, and ask for output in a format you can actually use.
• Different roles in a merchandising team have different entry points: buyers stress-test buy quantities, merchandisers automate reporting, planning directors model scenarios, and small business owners start with Google Trends plus ChatGPT.
• AI doesn't replace merchandising judgment, it eliminates the mechanical data work so you can spend more time on the decisions that actually require human expertise: brand direction, customer knowledge, supplier relationships, and trend interpretation.

When the Spreadsheet Lies: Real-Time Forecasting in a Volatile Market

It's Wednesday morning. Your buying team just flagged that a midi-skirt silhouette you ordered 14 weeks ago is sitting at 23% sell-through, and the delivery of another 800 units lands Friday. Meanwhile, your social media manager is texting you because a TikTok from Tuesday night has driven a 340% spike in searches for that exact style on your website. The spreadsheet says you're overstocked. The internet says you're about to sell out. Both are right, and neither one tells you what to do next. This is the scenario where AI demand forecasting earns its keep, not by replacing your judgment, but by processing signals you physically cannot track fast enough to act on.

Beyond Sales History: What AI Forecasting Actually Reads

Traditional forecasting tools, even sophisticated ones, work backwards. They look at what sold last season, apply a growth or decline factor, and project forward. That logic works in stable markets. Fashion retail is not a stable market. A single celebrity appearance, a viral review, or a competitor going into clearance can reshape demand inside 48 hours. AI forecasting tools are built to read forward-facing signals alongside historical data. That means they're ingesting search trend data from Google Trends, social engagement velocity from TikTok and Instagram, competitor pricing changes, weather forecasts, and even macro signals like consumer confidence indices, then weighting all of those inputs dynamically, not just at the start of a season but continuously throughout it.

The practical implication is significant. When you use a tool like Edited, Trendalytics, or NVIDIA's retail AI integrations, you're not just running a better version of last year's buy plan. You're running a model that updates its assumptions as the season moves. If a color that was performing weakly in week two suddenly accelerates in week four because a major influencer wore it, the forecast adjusts, and so do the replenishment recommendations. For merchandisers, this means the conversation shifts from 'what did we think would happen' to 'what is happening right now, and what should we do about it in the next 72 hours.' That is a fundamentally different way of working, and it requires building new habits around how you check, interpret, and act on forecast outputs.

What makes this accessible for non-technical teams is that the better platforms surface these insights through dashboards and plain-language summaries rather than raw data tables. You don't need to understand the underlying algorithm to act on a recommendation that says 'transfer 120 units of this SKU from your Manchester store to your Birmingham location, projected stockout in 6 days.' That recommendation is the output of a complex model, but your job is to evaluate it with your category knowledge and decide whether to execute. The skill being built here is not data science, it is critical evaluation of AI-generated recommendations, which is a different and very learnable skill.

Click to enlarge

Using ChatGPT and Claude for Demand Scenario Planning

Dedicated retail AI platforms are powerful, but they're also expensive and require integration with your inventory systems. For many merchandising teams, especially those in mid-sized brands or independent retail, the most immediately accessible route to AI-assisted forecasting is through general-purpose tools like ChatGPT Plus or Claude Pro. These won't connect to your live inventory data, but they are exceptionally useful for scenario planning, range architecture decisions, and translating raw sales reports into strategic recommendations. The key is knowing how to give them the right context. You are essentially briefing a very fast, very well-read analyzt who knows a lot about retail but nothing about your specific business until you tell them.

AI Forecasting Tools: A Practical Comparison for Retail Teams

The market for retail AI tools has expanded rapidly, and the options vary enormously by price, capability, and the type of team they're built for. A 12-store independent retailer has completely different needs from a 200-SKU e-commerce brand or a department store buying team. The table below cuts through the noise and maps the most relevant tools to the scenarios where they actually make sense, so you can have an informed conversation with your IT team, your finance director, or a vendor without needing a technical background.

Putting It Into Practice: Three Workflows for Monday Morning

The gap between understanding AI forecasting and actually changing how you work comes down to workflow integration. The most effective merchandising teams don't treat AI as a separate task, they weave it into decisions they're already making. The first workflow is the weekly trade review. Instead of spending 45 minutes manually pulling together a narrative from your sales dashboard, you copy your top-line performance data into ChatGPT or Claude, paste in your current stock positions, and ask for a structured summary of what's outperforming, what's underperforming, and what the sell-through trajectory implies for end-of-season stock. You still make the calls, but you arrive at the meeting with a sharper brief in a fraction of the time.

The second workflow is range planning support. When you're building a buy for next season, AI tools can help you stress-test your assumptions. You can describe your planned range architecture to Claude Pro, the split between entry, core, and fashion lines, your planned color palette, your depth per option, and ask it to identify gaps or risks based on current trend signals. This works best when you combine it with data from a trend intelligence tool like Trendalytics or Edited. The AI doesn't replace your buying instinct; it adds a layer of market signal checking that would otherwise require an analyzt or a costly trend report subscription.

The third workflow is exception management. Every merchandising team has more SKUs to monitor than hours in the week. AI tools can dramatically compress the time it takes to identify which items need attention. If your business uses Microsoft Copilot within Excel, you can highlight a table of SKU performance data and ask Copilot to identify the ten items most at risk of not clearing before season end, based on current sell-through rate and weeks remaining. That takes seconds rather than the 30-minute manual triage most teams currently do. The output is a shortlist, you still apply your judgment about whether to act, and how.

Click to enlarge

How This Looks Different Across Roles

• Buyers: Use AI scenario planning to stress-test option depth decisions before placing orders, ask 'what if this style underperforms by 30%' and model the markdown exposure before committing.
• Merchandisers: Automate the first draft of weekly trade commentary using Copilot or ChatGPT, cut report prep time by 40-60% and redirect that time to action planning.
• Store managers: Use AI summaries of your store's top and bottom performers to brief your team at Monday morning stand-ups without needing a report from head office.
• E-commerce managers: Use Google Gemini within Google Merchant Center to identify which product categories are gaining search momentum and adjust your featured collections accordingly.
• Planning managers: Use Claude Pro to model seasonal OTB (open-to-buy) scenarios, describe your budget, current commitments, and trend signals, and ask for a recommended reallocation.
• Marketing and merchandising cross-functional teams: Use AI to align promotional timing, ask ChatGPT to analyze which product categories historically respond best to which promotional mechanics, then map that against your current slow-movers.
• Small business owners and independent retailers: Start with the free tier of ChatGPT or Claude, even without live data integration, you can use it to interpret your own sales data and get structured recommendations faster than any spreadsheet.

Handling the Signals AI Can't See

Key Takeaways from Part 2

1. AI forecasting tools read forward-facing signals, social velocity, search trends, competitor pricing, not just historical sales, which makes them fundamentally different from traditional planning tools.
2. General-purpose AI tools like ChatGPT Plus and Claude Pro are immediately usable for scenario planning and trade analyzis without any technical setup or IT integration.
3. The skill to build is not data science, it is critical evaluation of AI recommendations, knowing when to act on them and when your own context overrides the output.
4. A weekly 40-minute habit, checking top SKUs and running a brief AI analyzis on your trade data, can prevent the majority of reactive markdowns.
5. Different roles interact with AI forecasting differently: buyers stress-test orders, merchandisers automate reporting, store managers get faster briefings, and e-commerce teams act on real-time search signals.
6. Multi-store reallocation decisions are one of the highest-value use cases for conversational AI, paste in your stock positions and sell-through by location and you get a structured transfer recommendation in under two minutes.
7. Always flag what the AI cannot see: planned promotions, supplier issues, local events, or competitor activity that isn't in any dataset but is in your head.

It's Thursday afternoon. Your weekly sell-through report just landed in your inbox and one SKU, a mid-weight quilted jacket, is sitting at 34% sell-through with six weeks left in the season. You ordered 800 units based on last year's numbers. A cold snap is forecast for next week, but your markdown calendar is already locked. You need a fast read on whether to hold, discount, or transfer stock to higher-performing stores. This is exactly the moment AI demand forecasting stops being a concept and becomes a decision-support tool you actually use.

Turning Forecast Data Into Inventory Decisions

Most merchandisers already have the raw ingredients for better decisions: sell-through rates, store-level sales history, weather patterns, promotional calendars, and supplier lead times. The problem is speed. Pulling those threads together manually takes hours you don't have on a Thursday afternoon. AI tools, including ChatGPT Plus, Microsoft Copilot, and specialized retail platforms like Edited or Relex, can compress that analyzis from hours to minutes. You paste in your data summary, describe your constraints, and ask a specific question. The output isn't a magic answer, but it's a structured starting point that's faster and more consistent than gut feel alone.

The key shift in mindset is treating AI as an analytical partner rather than a reporting tool. Most teams use spreadsheets to describe what happened. AI helps you model what might happen next, given a set of conditions you define. You can feed it a plain-language summary of your stock position, your remaining weeks on the floor, a planned promotional event, and a weather forecast, and ask it to reason through risk scenarios. It won't access your live ERP system unless you're using an integrated platform, but it will apply structured retail logic to the information you provide, consistently, without fatigue, at any hour.

This matters especially for smaller buying and merchandising teams who don't have a dedicated data science function. A two-person menswear team at a regional chain has the same access to ChatGPT Plus ($20/month) as a 50-person planning team at a national retailer. The difference is knowing how to ask precise, context-rich questions. Vague prompts produce vague outputs. Specific prompts, with real numbers, real constraints, and a clear decision you're trying to make, produce outputs you can actually act on before end of day.

AI Tools for Demand Forecasting: A Practical Comparison

Applying Forecast Thinking to Real Buying Decisions

Notice what that output does. It doesn't just give you an answer, it structures the trade-offs in a way that maps directly to how a buying meeting actually runs. You can paste that summary into a Teams message, present it in your Thursday review, or use it as the backbone of a markdown recommendation email to your director. The AI hasn't accessed your systems. You gave it numbers and context; it applied retail logic and returned a decision framework. That's the workflow.

The same approach works for open-to-buy planning, newness allocation, and size ratio decisions. If you're building a buy for next season and want to stress-test your depth on a hero style, describe your sales history, your supplier minimums, and your risk tolerance, then ask the AI to reason through upside and downside scenarios. You're not replacing the buyer's judgment, you're giving it a faster, more structured input to work from. The decision still sits with you. The analyzis gets there faster.

For teams using Microsoft Copilot inside Excel, the workflow gets even more direct. You can highlight a sell-through table and ask Copilot to identify which SKUs are at risk of end-of-season carryover, or to flag styles where the rate of sale has decelerated over the last three weeks. No formulas needed. You describe what you're looking for in plain English, and Copilot surfaces the patterns. This is particularly useful during range review season when you're working across dozens of categories simultaneously and need to triage quickly.

How This Applies Differently by Role

• Buyers: Use AI to stress-test depth decisions before committing to supplier minimums, model sell-through scenarios at different intake quantities.
• Merchandisers: Use AI to triage weekly sell-through reports and generate ranked lists of SKUs requiring action, faster than manual review.
• Planning managers: Use AI to document forecast assumptions in structured templates, creating an auditable record of why decisions were made.
• Store managers: Use AI to flag local stock imbalances and build transfer request summaries with supporting sell-through data.
• Small business owners (independent retail): Use ChatGPT to apply professional buying logic to decisions that would otherwise be made on instinct alone.
• E-commerce merchandisers: Use AI to analyze page performance data and match it to stock availability, identifying where low stock is suppressing conversion on high-demand lines.

Key Takeaways

• AI demand forecasting tools are accessible to non-technical retail professionals right now, no coding, no IT team required.
• ChatGPT Plus, Microsoft Copilot in Excel, and Google Gemini Advanced are the most practical starting points for buyers and merchandisers.
• The quality of your AI output depends almost entirely on the quality of your input, include real numbers, real constraints, and a specific decision.
• AI doesn't replace buying judgment; it compresses the time it takes to structure a decision so your judgment has better material to work with.
• Store transfers, markdown timing, open-to-buy depth, and size ratio decisions are all strong use cases for prompt-based AI analyzis.
• specialized platforms like Edited and Relex offer deeper automation but require enterprise investment, general AI tools cover most day-to-day analyzis needs.
• Saving effective prompts as reusable templates is one of the highest-ROI habits a merchandising team can build right now.
• The skill being built here isn't technical, it's knowing how to brief AI the same way you'd brief a sharp analyzt who's new to your business.