Months of stock4.0
0Limit 34
Cash sits in the warehouse
Some products are over-ordered and hold far more than three months of sales, often with sales already falling.
AGENTIC AIFor distributors, importers and retailers
Purchase Planner Agentic AI
Know what to buy, how much and when for every SKU. The agent reads your sales, stock, open orders and supplier proformas, and your buyers make every decision.
Order nowStock-out + Understock
225SKUs
Review order planCheck a proformaExport plan
Obsolescence risk
456
High + medium risk · 173 to watchAbove 3 months on arrival
0
No order overfills stockData issues
22
Logged for fixing in the ERPValue forecast
Planned purchases, next 12 months (FOB)
USD 48.6M
Aurex23.4M
Kelvo14.8M
Norda9.9M
Solis0.6M
The problem
Planning purchases by hand leads to too much of some stock and too little of the rest
A small purchase team has to decide, every month, what to order from each brand across a thousand SKUs. When that work is done by hand, the same four things go wrong.
Sold last 3 months412
In stock0
Coming0
Best sellers run out
At the same time, products customers still want have nothing on hand and nothing on the way.
Ordered in August165
Spreadsheet says order+165
Orders on the way are forgotten
Nothing tracks an order between signing and shipping, so goods already coming get ordered again.
R-BG410PUK6X-4 GBKBrandRBG410PUK6X4GBKERP
Proformas are checked by eye
Each brand sends its own layout and codes. Nobody checks every line against stock, forecast and the last price paid.
The solution
Purchase Planner AI is an AI agent that plans your purchase orders for you
An AI agent that turns your ERP data into a purchase plan for every SKU, and checks each supplier proforma before you sign it.
Sales history21 months, with seasons
Current stockWarehouse on hand
In transit & open POsWhat is already coming
Supplier proformasPDFs, scans, any layout
Purchase Planner AI
1
Cleans and checks the dataEvery gap goes to a data issues log
2
Forecasts every SKU3, 6 and 12 months, units and value
3
Plans the orderShould we buy, how much and when
4
Checks the proformaAgainst plan, stock, limit and price
Order planQuantity and order date per SKU
Early warningsStock-out, overstock, obsolescence
Proforma verdictsIn line, stop, still short, new
Why agentic AI
Agentic AI turns a manual, error-prone process into clear, checked purchase decisions
Purchase planning means thousands of numbers re-entered and re-checked every month. Small mistakes in that work become overstock or empty shelves. The agent does the reading, checking and arithmetic, and your team makes the decisions.
Manual planning
Spreadsheets, rules kept in memory, proformas checked by eye
- Orders already on the way get ordered again
- Limits checked after the order is signed
Purchase Planner
Agentic AI
Agentic AI
Agent-assisted planning
Clean data, checked orders and a clear reason for every recommendation
- Open orders counted automatically
- "Stop" shown before anyone signs
Reads your files directly
ERP exports and supplier PDFs, with no re-typing
Checks every number
Against stock, open orders, your limit and the last price
Explains every answer
One plain sentence per SKU that a buyer can verify
How it works
The agent works inside your current buying process, so nothing about how you buy has to change
It adds analysis at two points, the monthly order plan and the proforma check. Negotiation, signing, purchase orders and the ERP stay exactly as they are.
01Your team
Export month-end data
Stock, sales and goods receipt reports, uploaded or dropped in a folder.
02Agent
Check and clean the data
Returns, negative stock and bundle codes handled the same way every time.
03Agent
Build the order plan
What to buy, how much and when, per brand, with the reason for each SKU.
04Brand manager
Agree the order with the brand
The plan is the starting point. The negotiation stays with your team.
05Supplier
Brand sends the proforma
The PDF is uploaded to the agent as it arrives.
06Agent
Check every proforma line
Against the plan, the stock, the 3-month limit and the last price paid.
07Buyer
Accept, change or reject
Each line gets a human decision before the proforma is signed.
08Your team
Everything else stays the same
Signing, the ERP purchase order, shipping and receipt carry on as today.
One answer per SKU
For every product, the agent tells you whether to buy, how many units and on what date
The plan uses the eight columns purchase teams ask for, plus the reason in plain English, so any buyer can check the numbers with a calculator.
| SKU | Current stock | Forecast (3 mo) | Open PO | In transit | Safety stock | Recommended order | Order date | Reason |
|---|---|---|---|---|---|---|---|---|
| KRF632FIG46Kelvo · Refrigerator | 12 | 95 | 165 | 0 | 32 | 59 | 12 Dec 2026Lead 60 d | Only 12 on hand, but 165 already ordered. Enough until mid-December, then order 59. |
| AV950F24CBAurex · Vacuum cleaner | 129 | 3,962 | 0 | 0 | 1,321 | 6,475 | Order nowLead 60 d | Below the reorder level. Being replaced by a newer model, so check before ordering. |
| KFS4019Kelvo · Fan | 1,338 | 3,594 | 0 | 18 | 1,198 | 4,635 | Order nowLead 60 d | Peak season starts in April. Order now so stock lands before demand triples. |
Forecast built from each SKU's own history and season
Actual salesForecastLikely range
How the 59 is worked out
Reorder level (sales over lead time + safety)95
On hand + open PO + in transit177
Stock falls to reorder level on12 Dec
Order to reach the maximum59
Without the 165 on order, the agent would say order now. That is why open orders matter so much.
Features
The agent answers the seven questions a purchase team asks every month
Each answer is worked out from your own data and your own rules, and updates whenever the data or a setting changes.
Stock-outUnderstockHealthyOverstockSlow-movingNon-movingNew launch
Every SKU labelled automatically
Overstock, understock, slow-moving and non-moving, using your own definitions.
Maximum285
Reorder177
Safety32
Optimum, minimum, maximum and safety stock
Worked out per SKU from its sales and its supplier's lead time. Never above three months of sales.
Sales falling 4 monthsHigh
No sale in 5 monthsMedium
Newer model in rangeWatch
Obsolescence caught early
A risk level and the reason, before stock becomes dead stock.
Buy 59 units on 12 December. 165 are already on order, enough until then.
Should we buy, how much and when?
One plain sentence per SKU, with the numbers behind it.
ForecastStockOpen POIn transitExpected sales= PO quantity
PO quantity from five inputs
So the agent never tells you to order what is already coming.
3 months6 months12 monthsUnitsUSD
3, 6 and 12-month forecast, in units and value
Month by month for every SKU, so finance sees the purchase spend coming for each brand.
23Stock-out SKUs
72Proforma lines waiting
10Codes not in the ERP
54Orders with no price
Alerts that land on exactly the items to fix
Business exceptions grouped by what someone needs to do next, each opening the affected SKUs.
Stock health
Every product is labelled by stock health, so you can see which ones need attention
Each bar is a brand's full range, split by label. Every count opens those SKUs, so the purchase team starts with what matters.
By brand
AUREX350 SKUs
78 order now
Stock-out 16Understock 76Healthy 1Overstock 42Slow-moving 31Non-moving 94New launch 4Inactive 86
KELVO348 SKUs
85 order now
Stock-out 4Understock 85Healthy 9Overstock 49Slow-moving 66Non-moving 66New launch 5Inactive 58
NORDA289 SKUs
55 order now
Stock-out 3Understock 55Healthy 3Overstock 48Slow-moving 37Non-moving 61New launch 22Inactive 60
Overstock
Stock above three months of average sales.
Slow-moving
Above the three-month level and sales coming down.
Non-moving
Stock on hand and no sale in six months.
Stock-out
Nothing on hand or coming, and still selling.
Proforma check
The agent checks each supplier proforma line by line before you sign it
Upload the brand's PDF. The agent reads each line, matches it to your item codes, and asks one question: when these goods arrive, how many months of stock will you have?
1
Read
Model, quantity, price and shipment month, from scans or any layout.
2
Match codes
"R-BG410PUK6X-4 GBK" becomes item RBG410PUK6X4GBK. Unknown models are flagged.
3
Judge each line
In line, stop, still short, or new launch.
4
Check the price
Any change from the last price paid is shown.
5
Buyer decides
Accept, change or reject each line. A "stop" is a warning, never a block.
PDF
PI-2026-1184 · KelvoShipment November · 4 lines · USD 412,800
Reading…ModelQtyPriceVerdict
KRF632FIG46✓ Matched
60$412In line2.4 months on arrival
KAC18-INV-W✓ Matched
900$268+4% vs lastStop4.6 months on arrival, above the 3-month limit
KFS4019✓ Matched
1,200$38Still shortLikely to run out in May even with this order
KDW-60X PRONot in ERP
120$355New launchNo sales history, buyer decides
Totals read match the document: 2,280 units
Value forecast
The forecast shows how much money you will spend on each brand, month by month
Quantities are turned into value with the proforma price, or the last FOB price paid, month by month for the next twelve months.
Planned purchases, next 12 months
USD 48.6M
October, including today's order-now list: USD 11.1M
AurexUSD 23.4M
KelvoUSD 14.8M
NordaUSD 9.9M
SolisUSD 0.6M
By month, USD
AurexKelvoNordaSolis
11.1MOct
4.7MNov
1.4MDec
5.0MJan
1.3MFeb
4.6MMar
2.2MApr
4.6MMay
1.8MJun
4.9MJul
2.2MAug
4.9MSep
Your rules, your call
You set the rules the agent follows, and your buyers make every final decision
Nothing about how the agent decides is hidden, and nothing it suggests happens without a buyer.
Rules in plain text
A versioned knowledge base anyone can open to see why the agent says what it says.
Order planning rules v1.1
Stock must not go above 3 months of average sales.
Average sales = running average of the last 3 months.
A "stop" is a warning. The buyer may override it.
Settings, not development
Change the limit and lead times on screen. The plan recalculates.
The buyer stays in control
Brands sometimes ask you to take extra stock. The buyer can always say yes.
KAC18-INV-W · 900 units4.6 mo
Stop · above limitResults
On real distributor data, the agent cut stock-outs by 42%
1,001 SKUs across four brands, with no mock data. These are the results from back-testing and replaying the agent against what really happened.
Fewer stock-outs
−42%Stock-out product-months in a 9-month replay: 176 following the agent, against 305 in reality.
Forecast accuracy
90–96%At brand level, forecasting months the agent was not shown.
Proformas read correctly
17/17Lines read added up exactly to the printed totals: 19,944 units, USD 4.55M.
Codes matched
84/87Models matched to ERP item codes automatically. The other 3 were new models.
Estimate your working-capital release
A simple model: stock above your limit, times what it costs to hold.
Cash tied up above target
USD 5.0M
Holding cost avoided each yearUSD 1.0M
Stock above target25%
An estimate to start a conversation, not a guarantee. A proof of concept measures it on your own data.
Before and after
How monthly purchase planning changes when the agent does the analysis
BeforeWith Purchase Planner AI
Building the monthly planDays in Excel→Ready at month end
Orders already on the wayOrdered again→Counted automatically
The 3-month limitChecked afterwards→"Stop" on every line
Seasonal productsOrdered late→Ordered before peak
Dead stockFound too late→Risk flagged early
Proforma reviewBy eye→Line by line, with price
Purchase spendUnknown until invoiced→12-month value forecast
FAQ
Common questions from purchase and supply chain teams
What is Purchase Planner AI?
An agentic AI for distributors and importers. For every SKU it answers should we buy, how much and when, using sales history, current stock, stock in transit and open purchase orders. It also checks supplier proformas before they are signed.
Does the agent place orders in our ERP?
No. The agent recommends only. Buyers accept, change or reject every recommendation, and signing, purchase orders and the ERP carry on as they do today.
Do we need an ERP integration to start?
No. A proof of concept runs on month-end exports such as stock, sales and goods receipt reports, uploaded as files or dropped in a folder. A direct connection can come later.
How does it handle seasonal products?
Each SKU is forecast from its own history, including its season. A fan that sells three times its average in April and May is ordered early enough to land before the peak.
What if a brand asks us to take extra stock?
The agent marks the line "stop" because it would take stock above your limit, and explains why. The buyer can still accept it. A stop is a warning, never a block.
Can we change the rules?
Yes. The inventory limit, lead times by supplier and stock labels are settings. The rules themselves are kept as readable, versioned text, so changes need no development.
How accurate is the forecast?
On a real distributor's data it was 90% to 96% accurate at brand level. Single SKUs are less predictable because of bulk orders, which is why the agent keeps a safety buffer and a buyer reviews every order.
What is agentic AI in purchase planning?
Agentic AI is software that works through a task on its own, step by step, instead of only answering a single question. In purchase planning, the agent reads your sales, stock and open orders, forecasts demand, works out what to order for every SKU and checks supplier proformas. A buyer then reviews and approves the result.
Who is Purchase Planner AI for?
Distributors, importers and retailers that buy stock from brands or suppliers in advance, usually with lead times of one to four months. It suits purchase, supply chain and finance teams that plan hundreds or thousands of SKUs.
How does the agent decide how much to order?
For each SKU it adds expected sales over the supplier lead time to a safety stock to get the reorder level. It subtracts stock on hand, stock in transit and open purchase orders. If stock will fall below the reorder level, it recommends the quantity needed to reach the maximum level, which is never more than your inventory limit.
How is the order date worked out?
The order date is the day stock, including what is already on order, is expected to fall to the reorder level. If stock is already below that level, the order date is today and the SKU is marked "Order now".
How does it identify overstock and slow-moving stock?
Overstock is stock above your limit, usually three months of average sales. Slow-moving stock is above that limit with sales falling. Non-moving stock has had no sale in six months. You can change each definition in the settings.
How does it predict obsolete inventory?
The agent watches for warning signs: falling sales, months since the last sale, too many months of stock, a brand that has stopped supplying the model, and a newer model in the range. Each SKU gets a risk level and the reason.
What is a proforma check?
A proforma invoice is the supplier's quote for an order before it is confirmed. The agent reads each line, matches the supplier's model code to your ERP item code, and tells you whether the order is in line, would take stock above your limit, still leaves you short, or is a new launch. It also flags any price that differs from the last price paid.
Can it read scanned PDF proformas?
Yes. Proformas usually arrive in each supplier's own layout and are often scanned. The agent reads model, quantity, price and shipment month from them, and checks that the lines add up to the totals printed on the document.
Which ERP systems does it work with?
It works with any ERP that can export stock, sales and goods receipt reports to Excel or CSV, including SAP, Oracle, Microsoft Dynamics and Odoo. No ERP access is needed for a proof of concept.
What data does it need?
Monthly sales and stock by SKU, a goods receipt report with prices and dates, a list of open purchase orders, and sample supplier proformas. Around 18 to 24 months of sales history is enough to capture seasonality.
Does it forecast in value as well as quantity?
Yes. Forecasts and planned purchases are shown in units and in money, using the proforma price where there is one or the last purchase price from the goods receipt report. Finance can see the purchase spend coming for each brand each month.
How is this different from ERP reorder points or a spreadsheet?
ERP reorder points and spreadsheets use fixed numbers that someone has to update by hand. The agent recalculates minimum, maximum and safety stock for every SKU from its own sales, season and lead time every month, counts orders already on the way, and explains each recommendation in plain English.
What return on investment can we expect?
The main returns are less cash tied up in excess stock, fewer lost sales from stock-outs, and fewer days spent building the plan. On a real distributor's data, the agent reduced stock-out product-months by 42% for about 13% more average stock. A proof of concept measures the result on your own data.
How long does a proof of concept take?
A proof of concept runs on your month-end exports and a few sample proformas, with no ERP integration, so it can be set up in a few weeks.
Is our data secure?
The agent works on exported files and does not need write access to your ERP. It never creates or changes purchase orders. Data access can be limited to the purchase team.
Proof of concept
Try the agent on your own data in a proof of concept
Send your month-end exports and a few proformas. We run the agent on your data and show every recommendation, so your team can check it themselves.