Demand Forecaster: Weekly Demand for Every Part at Every Location
Weekly demand for every part at every location, scored against real orders.
The Demand Forecaster is the AI demand forecasting agent in the Ai1 platform by MyZone AI for distributors and parts businesses: it estimates weekly demand for each part at each location with a range and a plain reason, then scores last week's forecast against what customers actually ordered.
Each estimate comes with a range and a plain reason, so your team sees where a number is firm and where it is closer to a guess.
Honest scoring, week by week
Once a week has closed, the forecast is compared with actual demand, so you learn which parts it reads well and which it does not.
Seasonal parts read in context
Year-over-year patterns are built in, so a part that ramps up each spring is not mistaken for a sudden spike.
Demand hidden by empty shelves
Lost sales your team logged by hand are added back with a confidence weight, so out-of-stock weeks do not drag the estimate down.
Meet your agent
Get to know Teodor
Teodor tracks the rain in his vegetable garden as carefully as he tracks sales. He forecasts weekly demand for each part at each location and scores how accurate last week's forecast was, so your team plans with open eyes.
AI agent. Fictional persona; not a real person.
Teodor, the persona of the Demand Forecaster
Name
Teodor P.
Role
Demand Planner
Experience
27 years
Work history
Demand planner at a regional auto parts wholesaler Stock analyst at a building supplies distributor
Personality
Forecasts only, never orders, honest about misses
Based in
Plovdiv, Bulgaria
Why Teodor has a personality
AI does better work as a specific expert, so the Demand Forecaster works as Teodor (Demand Planner, Plovdiv, Bulgaria). You can rename Teodor or change the personality, experience and profile settings at any time.
An estimate per part and per location, built from your own sales and order history
Range and reason
Every estimate shows how sure it is, and a thin history is flagged
Method per part
Steady, seasonal and occasional sellers each get a method suited to how they move
Big one-off orders
Kept apart from everyday demand, so they do not inflate the forecast
Scored weekly
Last week's forecast set against actual demand, part by part
Your system
Read from a weekly export; it never logs in to or changes your inventory system
Capabilities
What the Demand Forecaster does for you
Forecasts each part at each location
A weekly estimate per part and per location, built from your own sales and order history.
Matches the method to the pattern
Steady sellers, seasonal lines and parts that sell only occasionally each get a forecasting method suited to how they move.
Shows the uncertainty
Each estimate carries a range, and a thin history is flagged rather than dressed up as precision.
Keeps one-off big orders apart
A large one-time order from a single customer is separated out, so it does not inflate the everyday forecast.
Scores the week that just closed
It compares last week's forecast with what customers actually ordered and keeps a running record of how it did.
Works from an export, not a login
It reads a weekly snapshot from your inventory system and does not connect to the system itself.
How the Demand Forecaster differs from a spreadsheet average
A spreadsheet average treats every part the same, so a seasonal line looks like a spike and a part that sells now and then looks like it never sells. The Demand Forecaster matches the method to each part's pattern, keeps one-off big orders apart, adds back logged lost sales and shows a range with every estimate. It then scores itself each week. It only forecasts; your people decide what to stock.
How this differs from the Inventory Lead: To turn forecasts into approved min/max changes, use the Inventory Lead; the Demand Forecaster estimates weekly demand for each part at each location and scores last week's forecast.
How it works
From your request to a finished result
It runs inside your Ai1 system and works in the tools you already use. You ask, it works, it reports, and it stops for your OK where it matters.
Tap the diagram to enlarge it
Step 1: Share a weekly export
Your team, with help from your Ai1 team, sets up a weekly snapshot of stock, sales and open orders.
Step 2: It runs the forecast
Each part at each location gets an estimate, a range and a short reason.
Step 3: It grades the previous week
Last week's numbers are set against what customers really ordered, part by part.
You approve
Step 4: People decide what changes
The forecast feeds the weekly review run by the Inventory Lead, where your people decide whether any stock level moves.
When to use it
Hand these situations to the Demand Forecaster
Illustrative photo. Which parts will each branch need more of next week?
If this is you…
…here's what it does
A seasonal part starts its yearly ramp
It reads the same weeks in past years, so the rise is expected rather than treated as a surprise.
A part sells only a few times a year
Its forecast uses a method made for patchy demand and says plainly how uncertain it is.
One customer places a huge one-off order
That order is kept apart from everyday demand so next week's estimate is not thrown off.
A part was out of stock for weeks
Lost sales your team recorded are added back, so the empty weeks do not read as zero demand.
You want to know if the forecasts hold up
The weekly score shows how close recent forecasts came, part by part and location by location.
What you get
A weekly demand forecast per part and per location
A range and a plain reason for each estimate
A weekly accuracy score against actual demand
Flags for thin histories and unusual orders
A short written summary for your team
What it won't do
✕Change stock levels in your systemHandled by: The Inventory Lead, with your team's approval
Example: weekly parts demand forecast with last week scored
Example with a fictional company. Names, people and figures are invented to show the agent's output. Any resemblance to a real company or person is unintended.
What it was asked: How many of each part will each branch need next week, and how close was last week's forecast?
Example report
Next week at a glanceHow each forecast was built: three parts up closeLast week's forecast, scored: 12 to 18 OctNotes for your weekly stock review
What it found: 3 findings, with the numbers
Demand Forecaster expects 18,940 units across 4,212 part and branch lines for the week of 26 Oct, 3% above the four-week average, led by heater parts at North branch.
Quarvell's own forecast for 12 to 18 Oct ran 1.5% high overall, with a 14.2% weighted error and 81% of lines landing inside their likely range.
Seven one-customer bulk orders were kept out of the base, 34 metres of logged lost hose demand were added back, and three data questions wait for your team.
This is an illustrative example of the weekly forecast and scorecard Demand Forecaster is being built to produce. It was written by hand from the agent's design for an invented company with invented parts, branches and figures; no real systems or people were involved, and nothing was sent, posted, paid or changed. Run date: .
Want a weekly demand forecast like this for your own sales history? Book a Demand Forecast walkthrough.
Dead Stock & Returns reads the forecast before calling a part slow or dead.
Common questions
Frequently asked questions about the Demand Forecaster
How do you forecast demand for parts that sell only now and then?
Parts that sell in irregular bursts break ordinary averages, because most weeks show nothing and then a large order lands. Methods built for intermittent demand look separately at how often orders come and how big they are. The Demand Forecaster picks a method that fits each part's pattern and shows how uncertain the estimate is.
Why should a demand forecast be scored every week?
A forecast nobody checks can drift for a long time before anyone notices. Comparing each closed week with actual demand shows which parts the method reads well and which it does not, so your team knows how much weight to give each estimate. The Demand Forecaster keeps that score part by part and location by location, and promises no accuracy figure up front.
Why do out-of-stock weeks distort a demand forecast?
When a part sits at zero on the shelf, sales history records no demand, even if customers kept asking for it. A forecast built on those weeks reads the part as less popular than it is. The Demand Forecaster adds back the lost sales your team logged, with a confidence weight, while back orders stay counted as real demand.
What is AI demand forecasting for parts?
It estimates how many of each part customers will order, week by week and location by location, from your own history. The Demand Forecaster adds a range and a reason to each estimate and scores itself against what actually sold.
How accurate are its parts forecasts?
That depends on your parts and your history, so no figure is promised up front. Each closed week is scored against actual demand, so you can see how it performs on your own data.
Does the Demand Forecaster change our stock levels?
No. It only forecasts. The Inventory Lead turns forecasts into proposed changes, and your people approve each one before it reaches your inventory system.
Does it connect to our inventory system?
It works from a weekly export of stock, sales and orders. It does not log in to your system or write to it.
Is the Demand Forecaster available now?
Not yet. Your Ai1 team will tell you when the Demand Forecaster is ready for your account. You can talk to us today about forecasting demand for the parts you carry.
About Ai1
How much does Ai1 cost?
Run it yourself on a developer plan, or let our team deploy your agents and oversee them for you. Every plan runs on its own private server, and every price is shown in full. See Ai1 pricing →
Do I get just the Demand Forecaster, or the whole platform?
The whole platform. The Demand Forecaster is not sold on its own: every Ai1 agent, including this one, is included on Developer Pro and every Fully Managed option, with no per-agent charge. Developer Core includes the development agents. Compare plans →
Operations
Operations
Operations
Operations
Operations