Predictika · Hard-drive supply commitments

An AI supply-planning agent's drive commitment, checked before it goes to the customer

An AI supply-planning agent drafts a commitment of hard-drive supply for a hyperscale or channel customer, either a new order or a pull-in to an earlier quarter. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the ramp schedule, customer qualification, export control, agreement, price floor, capacity and channel reserve, and returns a verdict, every reason and the figures it computed.

Scope: Cobalt Ridge Storage (a fictional hard-drive maker) supply commitments for UltraSMR 32TB and HAMR 44TB drives, as new orders or pull-ins, to hyperscale and channel customers. See everything this validator covers

This validator models Cobalt Ridge supply commitments, not just the scenarios shown below.

  • Platform and customer — the HAMR ramp quarter, whether the customer has qualified the drive for the quarter, export control, and restricted destinations (which need trade compliance review).
  • Agreements and price — the agreement maximum (base plus upside, less what is booked), the price floor blended from the base and upside parts of the order, spot deals outside any agreement, and a 5 percent price exception band.
  • Capacity — hyperscale capacity per quarter after the channel reserve, manufacturing flex beyond the 52-week build lead time, the channel reserve for channel orders, and the direction of a pull-in.
  • Stated figures — the order value and the capacity committed that the agent states against the figures computed from the model.

Any commitment that touches these rules is in scope, not only the scenarios pictured here. All customers, capacity, agreements, floors and prices are invented for illustration and are not Cobalt Ridge policy; the model does not cover other drive platforms, quarters beyond 2028-Q2, currency, payment terms or cancellation.

The AI supply-planning agentBuilds the commitment from the conversation and says it is ready.
Predictika validatorChecks every rule at once and returns the verdict, each finding, and the computed figures.

8 scenarios in 4 groups · each group runs from the agent’s first attempt to the outcome · click one to switch

What the agent proposes

The AI supply-planning agent says

Illustrative agent statement. The validator receives the structured request below.

Proposing agent: AI supply-planning agent (scripted)

Before sending, the agent must call Recorded run

Predictika validator


      

        
        
        
      

      
Waiting for the call.

      
    

  

  
  

    
Where rules-based CPQ gets it wrong
Click to expand
    
Our assessment of what a typical rules-based CPQ (product rules, price rules and validation rules, as in Salesforce CPQ) does with the same cases. It is a judgment about effort and reliability, not a claim that nothing can be built.

    
What the validator caughtTypical rules-based CPQPredictika
A price floor that depends on how the order splits across the agreementOnly with workarounds
The floor blends a base price and an upside price according to how much of the order is still inside the base volume, so each split needs its own price rule and the blend is recomputed by hand.		The booked volume, the order and the agreement terms are in one request and the blended floor is computed on every call.
A capacity check that depends on the quarter and the order typeOnly with care
Capacity, bookings and the channel reserve differ for every quarter and platform, so the rule list holds a check per combination and grows with each new quarter.		Capacity per quarter is data; one rule compares the need with it for any quarter, platform and order.
Approvals beside outright holdsOnly with care
A spot deal, a restricted destination, a price exception and manufacturing flex each route to a different approver, so they are separate approval rules that must not be mistaken for blocks.		Approvals and blocks are different findings in the same answer, and the verdict is block if any hold exists, otherwise needs_approval.
Fixing one issue changes the inputs of the nextOnly with workarounds
Moving a quarter changes the ramp, the qualification, the capacity and the agreement coverage together; rules fire in a fixed order, so a rep fixes one error and meets the next.		Every rule is checked on every call, in whatever order the choices arrive, so the full list of holds returns at once.
Figures the agent states itselfOnly with workarounds
The order value and capacity committed are computed from several lines, so a separate rule must recompute them and compare them with the stated figures.		The stated figures are fields of the request and are compared with the computed ones on every call.

  

  

    
What this validator catches: findings from its test cases
Click to expand
    
Every rule is checked on every call, in any order the choices arrive; one problem never hides another. Blocking findings stop the proposal; approval findings need a person to sign off; everything else is computed.

    
1$16.0 per TB is below the $17.25 per TB price floor for this order of UltraSMR 32TB.
2Helios Compute already has 9 EB of UltraSMR 32TB booked under its agreement. Adding 8 EB exceeds the agreement maximum of 16 EB (10 base plus 6 upside). This needs an amended agreement, not an order.
3UltraSMR 32TB hyperscale capacity in 2028-Q1 is 54 EB after the channel reserve and 52 EB is already committed; this commitment would need 60 EB, which cannot be met in 2028-Q1.
4UltraSMR 32TB hyperscale capacity in 2028-Q1 is 54 EB and this commitment takes the need to 57 EB. 2028-Q1 is beyond the 52-week build lead time, so manufacturing can add up to 4 EB of build starts. Needs manufacturing approval.
5HAMR 44TB first ships in 2027-Q2; 2027-Q1 is before the ramp.
6Northwind Cloud has not qualified HAMR 44TB for shipments in 2027-Q1. Volume can only be promised where the customer has qualified the drive.
7HAMR 44TB is an export-controlled platform and Lantern Cloud is a restricted destination.
8No agreement covers Lantern Cloud for HAMR 44TB in 2027-Q4. This is a spot deal with a price floor of $21.0 per TB and needs sales VP approval.
9The request states an order value of $50.0 million, but the computed value is $52.2 million.
10The request states 61 EB committed in 2027-Q4 after this commitment, but the computed figure is 59 EB.
11$16.5 per TB is below the $17.25 per TB floor but within the 5 percent exception band. Needs a pricing exception approval.
12A pull-in moves booked volume to an earlier quarter, but 2027-Q3 is not earlier than 2027-Q2.
13The UltraSMR 32TB channel reserve for 2027-Q3 has 1 EB left and this order asks for 2 EB. Channel orders cannot draw on hyperscale capacity.
14Lantern Cloud is a restricted destination. Trade compliance must review this shipment before it is confirmed.

  

  

    
Scenarios and results
Click to expand
    
The recorded result of the validator for each scenario on this page: the request the AI agent sent, the verdict, and every finding, in the order they appear above.

    
1. Hyperscale HAMR order: A Northwind Cloud HAMR 44TB order of 6 EB for 2027-Q1 at $19.00 with an order value and capacity figure of its own

    
The agent's request. Please commit 6 EB of HAMR 44TB to Northwind Cloud for 2027-Q1 at $19.00 per TB, stating an order value of $120.0 million and 4 EB committed in 2027-Q1 after this commitment. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Before the ramp. HAMR 44TB first ships in 2027-Q2; 2027-Q1 is before the ramp.
    • 
      
  • Customer has not qualified the drive. Northwind Cloud has not qualified HAMR 44TB for shipments in 2027-Q1. Volume can only be promised where the customer has qualified the drive.
    • 
      
  • Agreement maximum exceeded. Northwind Cloud already has 3 EB of HAMR 44TB booked under its agreement. Adding 6 EB exceeds the agreement maximum of 8 EB (6 base plus 2 upside). This needs an amended agreement, not an order.
    • 
      
  • Price below the floor. $19.0 per TB is below the $20.2 per TB price floor for this order of HAMR 44TB.
    • 
      
  • Hyperscale capacity exceeded. HAMR 44TB hyperscale capacity in 2027-Q1 is 0 EB after the channel reserve and 0 EB is already committed; this commitment would need 6 EB, which cannot be met in 2027-Q1.
    • 
      
  • Stated order value does not match. The request states an order value of $120.0 million, but the computed value is $114.0 million.
    • 
      
  • Stated capacity committed does not match. The request states 4 EB committed in 2027-Q1 after this commitment, but the computed figure is 6 EB.
    • 
    

    
2. Hyperscale HAMR order: The agent resubmits after fixing the errors

    
The agent's request. Please commit 3 EB of HAMR 44TB to Northwind Cloud for 2027-Q4 at $20.00 per TB, stating an order value of $60.0 million and 8 EB committed in 2027-Q4 after this commitment. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Computed figures. Hyperscale capacity 10 EB; 8 EB committed after this commitment; 2 EB left; price floor $19.8 per TB; order value $60.0 million.
    • 
    

    
3. Agreement maximum and capacity: A Helios Compute UltraSMR 32TB order of 8 EB for 2028-Q1

    
The agent's request. Please commit 8 EB of UltraSMR 32TB to Helios Compute for 2028-Q1 at $17.80 per TB, stating an order value of $142.4 million and 60 EB committed in 2028-Q1 after this commitment. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Agreement maximum exceeded. Helios Compute already has 9 EB of UltraSMR 32TB booked under its agreement. Adding 8 EB exceeds the agreement maximum of 16 EB (10 base plus 6 upside). This needs an amended agreement, not an order.
    • 
      
  • Hyperscale capacity exceeded. UltraSMR 32TB hyperscale capacity in 2028-Q1 is 54 EB after the channel reserve and 52 EB is already committed; this commitment would need 60 EB, which cannot be met in 2028-Q1.
    • 
    

    
4. Agreement maximum and capacity: The agent resubmits after fixing the errors

    
The agent's request. Please commit 2 EB of UltraSMR 32TB to Helios Compute for 2028-Q1 at $17.80 per TB, stating an order value of $35.6 million and 54 EB committed in 2028-Q1 after this commitment. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Computed figures. Hyperscale capacity 54 EB; 54 EB committed after this commitment; 0 EB left; price floor $16.65 per TB; order value $35.6 million.
    • 
    

    
5. Channel order: An Arcline Distribution channel order of 3 EB of UltraSMR 32TB for 2027-Q3 at $18.00

    
The agent's request. Please commit 3 EB of UltraSMR 32TB to Arcline Distribution for 2027-Q3 at $18.00 per TB, stating an order value of $60.0 million and 60 EB committed in 2027-Q3 after this commitment. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Price below the floor. $18.0 per TB is below the $19.0 per TB price floor for this order of UltraSMR 32TB.
    • 
      
  • Channel reserve exceeded. The UltraSMR 32TB channel reserve for 2027-Q3 has 1 EB left and this order asks for 3 EB. Channel orders cannot draw on hyperscale capacity.
    • 
      
  • Stated order value does not match. The request states an order value of $60.0 million, but the computed value is $54.0 million.
    • 
    

    
6. Channel order: The agent resubmits after fixing the errors

    
The agent's request. Please commit 1 EB of UltraSMR 32TB to Arcline Distribution for 2027-Q3 at $19.00 per TB, stating an order value of $19.0 million and 58 EB committed in 2027-Q3 after this commitment. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Computed figures. Channel reserve left before this order 1 EB; price floor $19.0 per TB; order value $19.0 million.
    • 
    

    
7. Pull-in: A Helios Compute pull-in of 7 EB of UltraSMR 32TB from 2027-Q2 into 2027-Q3

    
The agent's request. Please pull 7 EB of UltraSMR 32TB for Helios Compute in from 2027-Q2 into 2027-Q3, stating 64 EB committed in 2027-Q3 after this commitment. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Hyperscale capacity exceeded. UltraSMR 32TB hyperscale capacity in 2027-Q3 is 60 EB after the channel reserve and 57 EB is already committed; this commitment would need 64 EB, which cannot be met in 2027-Q3.
    • 
      
  • Pull-in points the wrong way. A pull-in moves booked volume to an earlier quarter, but 2027-Q3 is not earlier than 2027-Q2.
    • 
    

    
8. Pull-in: The agent resubmits after fixing the errors

    
The agent's request. Please pull 1 EB of UltraSMR 32TB for Helios Compute in from 2027-Q3 into 2027-Q2, stating 53 EB committed in 2027-Q2 after this commitment. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Computed figures. Hyperscale capacity 58 EB; 53 EB committed after this pull-in; 5 EB left.
    • 
    

  

  

    
Narration transcript
Click to expand
    
The text of the narrated walkthrough, step by step.

    
    
      
  1. Welcome to a demo of Predictika's Validator. An AI supply-planning agent drafts a commitment of hard-drive supply for a hyperscale or channel customer, either a new order or a pull-in to an earlier quarter. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the ramp schedule, customer qualification, export control, agreement, price floor, capacity and channel reserve, and returns a verdict, every reason and the figures it computed.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please commit 6 EB of HAMR 44TB to Northwind Cloud for 2027-Q1 at $19.00 per TB, stating an order value of $120.0 million and 4 EB committed in 2027-Q1 after this commitment. It is ready to send.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with seven findings, all in one answer. HAMR 44TB first ships in 2027-Q2; 2027-Q1 is before the ramp. Northwind Cloud has not qualified HAMR 44TB for shipments in 2027-Q1. Volume can only be promised where the customer has qualified the drive. Northwind Cloud already has 3 EB of HAMR 44TB booked under its agreement. Adding 6 EB exceeds the agreement maximum of 8 EB (6 base plus 2 upside). This needs an amended agreement, not an order. The other four are on screen.
    6. 
      
  6. The validator returned seven blocking findings in one answer. A check that reads one field at a time would pass most of this; it shows only when the choices in the proposal are checked against each other.
    7. 
      
  7. What this shows: for a price floor that depends on how the order splits across the agreement, a rules-based CPQ can do this only with workarounds.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the quarter to 2027-Q4, the volume to 3 EB, the price per TB to $20, the stated order value to 60 million dollars and the stated capacity committed to 8 EB.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All seven earlier findings are cleared.
    13. 
      
  13. Seven problems caught in one call, fixed in one pass, and a clean proposal. That is the value of Predictika's validator: catching mistakes before they reach the merchant.
    14.