You can build a simple AI sales forecaster for your startup using only free tools such as Google Sheets and publicly available large language models. This guide shows UK early-stage founders and e-commerce sellers how to combine spreadsheets with AI prompting to generate realistic forecasts based on your own historical data without writing code.
The approach relies on clear data preparation, structured prompts and iterative refinement rather than complex algorithms. It works particularly well for businesses experiencing British seasonal sales patterns such as higher volumes around bank holidays, summer festivals or the run-up to Christmas.
Why a simple AI forecaster helps UK e-commerce sellers
Early-stage founders often need to predict future sales to manage cash flow, set marketing budgets and decide stock orders. Traditional forecasting can feel out of reach without specialist software or statistical knowledge. A practical AI-assisted method using free tools lets you combine your real sales records with reasoned estimates that reflect seasonal patterns common in the UK market.
This approach avoids over-reliance on black-box predictions. Instead, you remain in control by feeding the AI your own numbers and asking it to explain its reasoning. The result is a forecast you can adjust as new data arrives or as external factors such as weather or economic conditions change.
Prepare your British seasonal sales dataset
Start by collecting at least twelve months of monthly sales data in a spreadsheet. Include columns for month, year, total revenue, units sold, and any notable events such as Black Friday, Easter or summer sales. Conversion rates and seasonal patterns vary widely; benchmark against your own data.
Use this example structure in Google Sheets or Excel. Create a tab called "Historical Data" and enter realistic but anonymised figures. For instance, a clothing retailer might record £4,200 in January, £5,800 in April (Easter uplift), £9,100 in July (festival season) and £18,400 in November. Keep the dataset small and clean. Remove any refunds or one-off bulk orders that would distort patterns unless you add a separate column to note them.
Save a copy of the raw data before any analysis so you can always return to the original numbers.
Set up your free spreadsheet template
Open a new Google Sheet (free with a Google account) or use Excel Online. Create three tabs: Historical Data, Forecast Inputs and Forecast Output. In Forecast Inputs, add cells for key assumptions the AI will use, such as expected overall growth rate, the impact of each major UK shopping period, and any planned marketing spend changes.
In the Forecast Output tab, set up columns for future months, baseline projection, AI-adjusted forecast, upper and lower range, and notes. Use simple formulas like =AVERAGE() to calculate historical monthly averages and =GROWTH() for basic trend lines. These built-in functions provide a transparent starting point before AI refinement.
Format all currency cells as British pounds using the pound sterling symbol. Keep the sheet shareable but set viewing permissions carefully if the data is commercially sensitive.
Craft effective prompts for AI forecasting
Copy sections of your data and paste them into a free AI chat tool such as ChatGPT, Claude or Gemini. Structure every prompt clearly. A useful base prompt reads: "Here is monthly revenue data for a UK e-commerce store selling home goods from January 2023 to December 2024. Conversion rates and seasonal patterns vary widely; benchmark against your own data. Please analyse the seasonal trends and produce a monthly forecast for the next twelve months. Explain your reasoning step by step and provide a conservative, moderate and optimistic range for each month."
Paste your actual data after the prompt. Ask the AI to identify patterns such as the average percentage uplift in Q4 compared with Q1. Request that it factor in known UK events including the January sales, spring bank holidays and the pre-Christmas period. Always ask for explanations rather than just numbers so you can judge whether the logic fits your business.
Run the same prompt several times with slight variations. Compare the outputs and take the median values as a balanced view. Record each AI response in a separate sheet tab for later reference.
Combine spreadsheet calculations with AI insights
Transfer the AI-generated monthly figures into your Forecast Output tab. Use the spreadsheet's built-in features to create a weighted average between the simple trend line and the AI suggestion. For example, give 60 percent weight to the AI forecast and 40 percent to the historical average if the AI reasoning seems particularly relevant to your category.
Add columns that automatically calculate a 15 percent buffer above and below the main forecast to give a realistic range. In the notes column, copy the key sentences from the AI explanation that justify each month's projection, such as "November uplift of 65 percent based on historical Black Friday and pre-Christmas buying patterns observed in the supplied data."
Create a simple line chart directly in the sheet to visualise historical actuals against the new forecast. Update the chart each month as real results come in so you can see where the forecast was accurate or needed adjustment.
Refine and maintain your forecaster over time
Treat the forecast as a living document. At the end of each month, add the actual revenue to the Historical Data tab and rerun the AI prompt with the updated series. Note which assumptions proved correct and which external factors, such as unusually warm weather affecting garden product sales, were not captured. Over time you will improve the quality of your prompts by including these observations.
Keep prompts consistent in structure so comparisons remain valid. If your product range changes significantly, create a separate forecast sheet for the new category rather than mixing dissimilar data. Review the entire model every quarter to confirm it still reflects current trading conditions.
Document any changes you make to the weighting or to the prompt template so that another team member could understand the method if required.
Practical limitations and sensible use
This method produces indicative forecasts rather than precise predictions. It cannot foresee sudden economic shifts, new competitor activity or changes in platform algorithms. Treat the output as one input among several when making decisions about stock, hiring or marketing spend. Cross-check the figures against your own knowledge of the business and against cash-flow projections.
Free AI tools may change their capabilities or terms of service. Always verify the current rules on the platform you use. Similarly, any financial decisions should be reviewed against the latest guidance available from GOV.UK, HMRC or Companies House as thresholds and requirements change.
The value comes from the discipline of recording data, explaining assumptions and regularly comparing forecasts with reality. Many small UK e-commerce teams find this structured approach improves planning even if the final numbers are adjusted by hand.
Frequently asked questions
Can I build a sales forecaster without using Python or paid software?
Yes. Google Sheets combined with a free large language model such as ChatGPT or Gemini is sufficient. The spreadsheet handles calculations and charting while the AI analyses patterns when given clear prompts and your own sales data.
How much historical data do I need for a useful UK seasonal forecast?
At least twelve months of monthly figures is the practical minimum. Twenty-four months allows the AI to observe repeating patterns such as the pre-Christmas peak and post-January lull that are typical for British e-commerce businesses.
How do I include British seasonal events in the forecast?
List the events and their historical impact in your prompt. Provide past sales figures for months containing Black Friday, Easter, summer bank holidays and the Christmas period. Ask the AI to calculate average uplifts from your data rather than using generic percentages.
What should I do when actual sales differ from the AI forecast?
Add the new actual figure to your dataset each month and rerun the prompt. Record the reasons for any large variance in your notes. Over time you will refine both the data and the prompts, which usually improves future accuracy.
Is this type of forecasting suitable for very early-stage startups?
It works well once you have several months of real sales. Before that, the method can still help by letting you test different scenarios with assumed baseline figures while you gather initial trading data.
Related reading: AI tools for recruitment and HR in early-stage UK startups, Use AI to boost website conversion rates for British online stores, How to use AI for preliminary UK trademark and brand name searches. Browse all ai for business guides or apply to Loomworks.