What is Stok.ly AI?
Stok.ly AI is a three-tier artificial intelligence system built into the Stok.ly platform. The three tiers are Operational AI for daily forecasting, replenishment and stockout prevention; Commercial AI for margin-driven replenishment, complex transfers and SKU-level profitability; and Executive AI for board-level risk reporting and growth intelligence.
Stok.ly AI is designed to focus on the day to day activities in your business like purchasing and replenishment, but also answer those difficult to answer questions that your executive team need answered.
Unlike AI features added on top of a legacy ERP, Stok.ly AI runs on a purpose-built data lake that pre-computes statistical intelligence every night for every product at every location in your product catalogue.
Roughly 100 fields per SKU, per location — rolling demand, standard deviations, statistical percentiles, safety stock, reorder points, days of cover, stockout risk, gross margin, GMROII and dead-stock signals are calculated and exist as usable values that are retained for 365 days before you ask a question.
When you need answers, the AI reads the pre-calculated data.
It does not calculate under pressure inside a conversation and hope the arithmetic holds.
Stok.ly AI is not an ERP that simply produces a report. Instead, Stok.ly AI calculates in advance, analyses, flags issues pro-actively, makes recommendations on actions and with your approval completes the actions for you.
The Problem: Buying by Memory and Static Thresholds
Manual replenishment means someone deciding, from a sales report or a gut feeling, what to reorder — and a min/max threshold set once, then left untouched until it is obviously wrong. Neither responds to what demand is actually doing. A product that suddenly accelerates blows past a stale minimum before anyone notices. A product that slows down keeps getting reordered at the old rate, tying up cash on a shelf.
“We are massively overstocked and we are having to manually recalibrate real levels, minimums and maximums.”
— Director, Hardware Merchant
The recalibration exercise is the tell.
Any business running min/max at scale eventually finds itself manually reviewing thresholds across every location, because the thresholds have drifted away from reality and nobody has time to keep them honest.
For this business, that used to mean 16 stores of thresholds being adjusted manually by csv.
With Stok.ly AI, that now means a review of sales over the last 30, 60 or 90 days and adjustment in bulk of mins and maxs to plan for 14 days of stock as a minimum and 21 days of stock as a maximum – edited, approved and adjusted for real demand, based on the sales of each SKU at each location, in minutes.
Trust your stock. Transfer with confidence. Control orders. Fulfill every day like clockwork.
How Stok.ly Forecasts Demand
Stok.ly forecasts every SKU at every location using two models.
A 30-day moving average provides the baseline for stable, non-seasonal products.
A trend-adjusted version detects when demand is accelerating or decelerating and scales the baseline accordingly, so a product ramping up is caught early rather than after it has already stocked out.
Why stockout days are excluded from sell-through
This is the single most common error in demand analysis, and correcting it changes purchasing decisions immediately.
Days on which a product had zero stock are excluded from its sell-through rate.
Buy 100 units of two different products.
One sells out in 10 days. The other takes 100 days to clear.
A basic bestseller list shows both as identical: 100 units sold.
In reality, these items are nothing alike but if your team are using basic sell through rates or best seller reports, every downstream calculation — forecast, safety stock, reorder point, GMROII — inherits the error if the correction is not made at source.
How the AI knows how confident to be
Every recommendation carries its evidence with it: literally, how many days of demand history exist for that product at that location. Five days of data still produces a forecast, but the AI says so and hedges accordingly.
A full year of history produces a confident recommendation, because there is a full year of evidence behind it.
Confidence is stated by the AI Assistant as part of the calculation, not implied.
One consistent demand signal, however you sell
Demand is measured at the moment stock actually needs to leave a location, not when a sale is logged.
For a retail till, sale and demand are the same moment someone pays and walks out of the shop with their items.
For a wholesale order taken today and fulfilled in ten weeks, sale and demand are not the same moment.
The forecast is built around the fulfilment date, because that is what determines whether stock was available when it was needed.
This approach evens out sale and demand across businesses trading under very different models.
Run retail, wholesale and ecommerce together and this approach to sale and demand holds consistently across all three models.
G2 Customer Reviews
What do you like best about Stok.ly – Inventory-Centric Cloud ERP?
“we have been a user for the last 4 years and we cannot think of life without it for our retail business. it makes listing to shopify and other marketplaces easy, keeps inventory accurate online and the POS is easy to use. accurate inventory across all sales channels is the big win for us.”
Order management and inventory ERP software for growing retail, wholesale and distribution businesses.
Four Numbers, Recalculated Every Day
The forecasting engine outputs four fields per SKU per location. Together they replace the weekly export, the memory and the stale minimum.
| Signal | What it tells you | How it is calculated | How the buying team uses it |
|---|---|---|---|
| Reorder point | The exact threshold at which to place a purchase order today. | (Trend-adjusted forecast daily demand × lead time in days) + safety stock | When available stock falls below this number, order now. |
| Safety stock | The buffer needed to absorb demand variability at a 95% service level. | 90-day demand standard deviation × √(lead time in days) × 1.645 | Set and forget. It recalculates daily as variability changes. |
| Days of cover | How long available stock will last at the current rate of demand. | Available stock ÷ trend-adjusted daily demand | Sort ascending to see imminent stockouts. |
| Stockout risk | A single urgency score from 0 to 1. | Composite of current availability, forecast demand and demand variability | Sort descending for today’s purchasing priority list. |
On the 1.645. That is the z-score for a 95% service level — the standard for a business that intends to be in stock on 19 of every 20 demand days. It is a choice, not a law of nature, and it is stated here because a buyer should be able to interrogate it. Where lead time has not been configured on the product record, safety stock returns null and the AI flags the product for setup rather than quietly guessing.
Our replenishment process is currently manual, error-prone and time-consuming. Between the three of us, we’ll hopefully gain 70 to 80% of our time back.
Statistical Replenishment vs Min/Max: What Actually Changes
| Min/max replenishment | Stok.ly statistical replenishment | |
|---|---|---|
| Where the threshold comes from | A person, once | Actual demand variability, daily |
| How it responds to a demand spike | It does not, until someone notices | Safety stock increases automatically |
| How it responds to a slowdown | Keeps ordering at the old rate | Safety stock reduces, releasing capital |
| Treatment of stockout days | Counted as low demand | Excluded from sell-through |
| Service level | Implicit and unknown | Explicit: 95% |
| Maintenance overhead | Periodic manual recalibration | None |
| Lead time handling | Usually baked into the minimum | A separate configured input |
Stok.ly still supports location-specific min/max rules, cascading replenishment logic, shortage allocation priority and weighted distribution for businesses that want rules-based control. The statistical layer is not a replacement for operational governance; it is a replacement for guessing the numbers that feed it.
Cross-Location Stock Balancing
Stock sitting overstocked in one location while another runs short is one of the most common and least visible sources of lost margin in a multi-location business.
The cash is already spent; the stock is sitting in the wrong location.
Stok.ly’s AI classifies every product as fast or slow-moving per location, compares stockout risk and days of cover across locations simultaneously, identifies exactly where the imbalance exists, and can recommend or directly create the stock transfer to fix it.
The question it answers is not “which location is lowest” but “where is this stock most likely to sell, and where is it currently doing nothing”.
You may wish to ask “where will this stock make the most margin?” or generate the highest GMROII.
The sequencing matters commercially: rebalancing existing stock before buying more is the difference between moving cash and spending it.
Ask a competitor’s AI to check for a margin return at the SKU before creating a transfer or recommending a purchase order — is their AI intelligence joined up?
Ask a Question in Plain English. Get an Answer. Let It Act.
Stok.ly’s AI assistant does not stop at giving you the answer.
Where the next step is clear, it can take it — creating a purchase order, raising a stock transfer, starting a manufacturing run, or updating pricing and attributes.
Stok.ly AI works on an “Analyse, Recommend and Act” pattern.
Make better decisions and let Stok.ly AI complete the manual tasks for you.
Questions the AI Answers From Pre-Calculated Data Points
Which items do I need to order today? (Sorted by stockout risk, descending.)
Which products will run out first? (Sorted by days of cover, ascending.)
What is the reorder point for this SKU at this location, and how was it calculated?
Where is demand accelerating fastest this month?
Which products have had no demand for more than 30 days, and at which locations?
Should we transfer these items or discount and sell them off?
Which SKUs have manually set minimums significantly above or below their statistically correct safety stock?
Where am I overstocked relative to demand, and where should that stock go instead?
It will push back on a bad purchase order
Shown a draft purchase order, the AI reviews it against everything it holds on each line — margin, stockout-corrected sell-through, what is already in stock, how fast it is actually moving — and provides feedback to help you create better Purchase Orders.
These items are slow-moving and low-margin.
Are you sure about these quantities.
Have you considered these other lines instead, where demand is accelerating and cover is already tight.
This is the behaviour that distinguishes an advanced AI Assistant from a AI reporting tool.
Running continuously in the background
Beyond the assistant you ask directly, an AI business intelligence layer analyses inventory, suppliers, pricing, purchasing, returns and warehousing continuously, surfacing findings as cards organised by business area.
Each card carries a specific insight, a recommended action, and the ability to execute that action on the spot.
Analyse, Recommend and Act
How This Compares to Cin7, Brightpearl, Orderwise and Linnworks
Positions below reflect vendor-documented capability as at 27 July 2026.
- Cin7. ForesightAI forecasts demand up to 24 months ahead, calculates reorder points, recommends safety stock and generates purchase orders, using ABC segmentation, and recommends at least six months of history for good accuracy. It is a paid add-on for Core and Omni, priced on gross merchandise value. Cin7 does not document a pre-computed intelligence layer, a defined statistical service level, or exclusion of stockout days from sell-through — and Ask Me Anything, its conversational tool, is documented as an insight and guidance tool rather than one that commits transactions.
- Brightpearl. Forecasting and replenishment live in Inventory Planner, a connected Sage for Retail product, which forecasts from sales history, seasonality, stockouts and trends and can push approved purchase orders into Brightpearl. Sage Copilot adds a prioritised insight panel — priority replenishments, purchase-order value review, overdue stock orders, missing cost prices — but it is a Premium-only beta, several insights require ABC analysis enabled, and it depends on synchronisation between two products rather than reading one live record.
- Orderwise. Data-driven forecasting using historical sales, seasonal trends, min/max levels and lead times, with automated purchase planning. Its public material does not identify this as AI or machine learning. MyForterro AI is invitation-only and proof-of-concept.
- Linnworks. No demand forecasting capability of this kind; the product’s centre of gravity is multichannel listing and order routing.
None of the four recalculate reorder point, statistical safety stock, days of cover and stockout risk automatically overnight for every SKU at every location, and none combine that with an assistant that can act on the finding directly.
That is the specific claim, and it is the one worth testing in a demo when reviewing these competitors.
How to Test Any ERP’s Replenishment AI, Including Ours
Use the same anonymised data set and the same questions with every vendor, and insist on a live workflow rather than slides.
- Show the ten SKUs most likely to stock out in the next 30 days, by location — and explain why each is at risk, with the underlying transactions and the data timestamp.
- Show the safety stock for a specific SKU and the formula that produced it, including the service level assumed.
- Confirm whether stockout days are excluded from the sell-through rate. Ask to see it demonstrated on a product that sold out.
- Recommend a purchase order that respects lead time, minimum order quantity, case size, price breaks and a cash limit.
- Check whether stock should be transferred from another location before anything is purchased.
- Forecast a new SKU with no history, using a comparable product, and explain the relationship chosen.
- Change a lead time and show exactly how the reorder point and recommended actions change.
- Introduce stale stock, missing cost prices and a promotion, then see whether the AI warns you or answers confidently from poor data.
Score on time taken, number of clicks, whether the answer is traceable to source records, and whether the insight becomes an editable draft without rekeying.
Frequently Asked Questions
Does Stok.ly calculate safety stock automatically?
Yes. Safety stock is recalculated daily for every SKU at every location from actual demand variability, using the 90-day demand standard deviation multiplied by the square root of lead time in days multiplied by 1.645, the z-score for a 95% service level. It is not set manually once and left to go stale. A product that becomes spiky receives more buffer automatically; a product that stabilises receives less, freeing capital.
How does the AI decide when to reorder?
Through a reorder point that combines three things: trend-adjusted forecast demand, your configured lead time, and statistical safety stock. The calculation is (trend-adjusted forecast daily demand multiplied by lead time in days) plus safety stock. When available stock drops below that number, today is the day to place the order for stock to arrive before a stockout.
What is the difference between min/max replenishment and statistical demand forecasting?
Min/max uses thresholds a person set once and must periodically recalibrate by hand. Statistical forecasting derives the threshold from what demand is actually doing and updates it daily. Min/max cannot respond to a demand spike until someone notices; statistical safety stock increases automatically. Min/max has an implicit and unknown service level; Stok.ly’s is explicitly 95%. Stok.ly still supports min/max rules for businesses that want them, including location-specific thresholds and cascading replenishment actions.
Does the AI account for stockout days?
Yes. Days on which a product had zero stock are excluded from its sell-through rate. This corrects the most common error in demand analysis: a product that sold out in ten days and then sat at zero for weeks is not treated as a slow mover. Because the correction is made at source, every downstream figure — forecast, safety stock, reorder point, GMROII — inherits the corrected number.
What is the stockout risk score and how do I use it?
It is a composite score between 0 and 1 combining current availability, forecast demand and demand variability, recalculated daily per SKU per location. Sorting descending gives the buying team the day’s purchasing priority list in one action. Sorting days of cover ascending gives the same list framed by urgency in days rather than risk.
Can the AI actually create purchase orders and transfers, or does it just tell me what to do?
Both. Ask a question and get an answer, or ask it to act — creating the purchase order, the stock transfer or the manufacturing run directly, with your approval. It shows the full workings before it asks, and nothing is committed until you confirm.
Can the AI move stock between locations automatically?
Yes. It classifies every product as fast- or slow-moving per location, identifies where stock is overstocked relative to demand and where it is short, and can recommend or directly create the rebalancing transfer. It will also check whether a transfer solves the problem before recommending a purchase, which is the difference between moving cash and spending it.
Will the AI tell me if a purchase order I have drafted is a bad idea?
Yes. Shown a draft purchase order, it reviews each line against margin, stockout-corrected sell-through, current stock and demand trend, and will say if items are slow-moving and low-margin, question quantities, and suggest alternative lines where demand is accelerating and cover is already tight.
Can we train the AI on our own purchasing and forecasting rules?
Yes, through a Model Context Protocol knowledge base built specifically for teaching Stok.ly’s AI business-specific rules. You write training articles describing your purchasing criteria, forecasting adjustments, range integrity definitions and operational processes, and the AI applies them when answering questions and executing tasks. Competing ERP AI features are configured rather than taught.
What happens if lead time is not configured for a product?
Safety stock returns null for that product and the AI flags it for setup rather than substituting a guess. This is deliberate: a silently assumed lead time produces a confident-looking reorder point that is wrong, which is worse than no number at all.
Can the buying team get these numbers outside the AI assistant?
Yes. All replenishment signals are available conversationally through the AI assistant and via API endpoints, including for Power BI, so a buying team can have a morning dashboard as well as a conversation. The figures are identical because they come from the same pre-computed fields.
How does Operational AI handle wholesale orders fulfilled weeks after they are placed?
Demand is measured at the point stock needs to leave the location, not when the order is logged. For a wholesale order taken today and fulfilled in ten weeks, the forecast is built around the fulfilment date, because that is what determines whether stock was there when it was needed. Businesses running retail, wholesale and ecommerce together therefore get one consistent signal rather than three that disagree.
Contact Information
All our sales, support and development team are located in Hereford and Cheltenham in the U.K. Please submit the contact form and we will contact you within the same business day.