Predictika · process-control tool delivery commitments

An AI sales agent's delivery promise, checked before the customer sees it

An AI agent drafts a quote that promises delivery of process-control tools. Before the quote goes out, the agent sends it to Predictika, which checks the promise against unbooked build slots, shared optical module supply, lead time, the acceptance milestone, export-license status and the service term, compares the figures the agent states with the computed ones, and returns every finding together.

An illustrative model built from public information about KLA's tool families. All supply numbers, slot counts, lead times and thresholds are invented, the customers are fictional, and the page is not affiliated with or endorsed by KLA.

Scope: delivery commitments on KLA wafer inspection (39xx and 29xx Series), reticle inspection (R9xx Series), unpatterned wafer inspection (Surfscan Series) and overlay metrology (Archer Series) tools. See everything this validator covers

The validator models the delivery commitments a quote makes for these five tool families only, in the four delivery quarters from Q4 2027 to Q3 2028. It handles any quote involving these families and quarters, not just the scenarios shown on this page.

  • Covered. The order size against unbooked build slots and shared optical module supply in the delivery quarter, each with a safety margin; standard and minimum lead time by family, with expedite approval; the acceptance milestone against delivery plus installation and acceptance time; quote age against validity and cost drift since the quote; China export-license status; restricted-party customers; the service term; and the figures the agent states (unbooked slots, earliest acceptance quarter) against the computed ones.
  • Not covered. Other KLA product lines and software; spare-parts and service-contract entitlement quoting; export classification, because the validator reads the license status and does not decide whether a license is needed; accounting conclusions on revenue recognition, because it only checks that the milestone date is physically achievable; list prices and configuration (this validator produces no price build-up or line-item quote, because the figures it checks are slots, supply, lead time and acceptance timing); live supply feeds, because a small catalog of sample figures stands in.
  • Assumptions. Every supply number, slot count, lead-time week, optical module count and threshold is an invented illustrative value, not KLA data. The family names are public. The shared-optics pool, the link between delivery and acceptance timing, and per-quote license tracking are modeling assumptions.
The AI sales agentBuilds the delivery quote 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 sales agent says

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

Proposing agent: AI sales 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 delivery slot against the unbooked build slots in that quarterCannot do it
Build slots live in planning and supply systems, not in a quote. A CPQ has no figure to compare the order with unless someone builds an integration and keeps it current.		Slots are model data. The order is checked against what is left in the delivery quarter, and an order that leaves no margin goes to a person.
Slots fit but the shared optics do notCannot do it
Optical modules are one pool shared by every family and every order. Rules on one quote cannot see what other quotes and families have already drawn from it.		The pool is model data, each family declares what a tool needs, and the order is checked against what is unallocated.
An acceptance milestone set before acceptance can finishOnly with care
Possible with a date or quarter formula, but only when someone writes one for each family and the quarter arithmetic is right.		The installation and acceptance time is part of each family, and the milestone is checked against delivery plus that time.
A stated figure that differs from the computed oneOnly with care
A validation rule can compare a stated figure with a calculated one, but only if the figure is a field on the quote, which it rarely is.		Stated figures travel with the quote and are compared with the computed ones on every call.
Thin margins that should go to a person, not stop the quoteOnly with workarounds
A hard rule either blocks or lets the quote through. A middle path needs approval fields and a routing workflow built around each rule.		Block, needs approval and allow are separate outcomes of the same rules, with the reason for each.

  

  

    
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 quote; approval findings need a person to sign off; everything else is computed. The examples below are messages returned in the recorded runs.

    
1the order quantity of 6 is above the unbooked build slots left in Q4 2027 (4); the delivery cannot be promised.
2this order needs 27 optical modules and only 24 are unallocated in Q1 2028; shared subsystem supply cannot cover it.
3the acceptance milestone is set for Q2 2028 but delivery in Q1 2028 plus installation and acceptance is not complete before Q3 2028.
4shipment to China needs an export license and its status is not filed; do not quote a delivery date.
5the agent stated 5 unbooked build slots; the validator finds 4 in Q4 2027.
6this order leaves 3 unallocated optical modules in Q1 2028, under the 4 module safety margin; confirm with supply planning.

  

  

    
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. Many problems at once: China fab with six problems

    
The agent's request. Please promise Aurora Wafer Systems delivery of 6 39xx Series wafer inspectors in Q4 2027 in 40 weeks, for its China fab. The quote includes a 12 year service contract, and component cost has risen 15 percent since the quote. There are 6 unbooked build slots in Q4 2027. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Not enough unbooked build slots. The order quantity of 6 is above the unbooked build slots left in Q4 2027 (4); the delivery cannot be promised.
    • 
      
  • Not enough shared optical modules. This order needs 18 optical modules and only 10 are unallocated in Q4 2027; shared subsystem supply cannot cover it.
    • 
      
  • Lead time below the minimum. 40 weeks is below the 44 week minimum lead time even with expedite for the 39xx Series wafer inspector.
    • 
      
  • China export license not filed. Shipment to China needs an export license and its status is not filed; do not quote a delivery date.
    • 
      
  • Service term above the maximum. A 12 year service contract is above the 10 year maximum the support plan covers.
    • 
      
  • Stated build slots do not match. The agent stated 6 unbooked build slots; the validator finds 4 in Q4 2027.
    • 
      
  • Cost moved since the quote. Component cost has moved 15 percent since the quote; honoring it needs a margin review.
    • 
    

    
2. Many problems at once: The agent resubmits after fixing the errors

    
The agent's request. Please promise Aurora Wafer Systems delivery of 6 39xx Series wafer inspectors in Q1 2028 at the standard 56 week lead time, for its China fab with the export license granted. The quote includes a 5 year service contract and no change in component cost. There are 12 unbooked build slots in Q1 2028. It is ready to send.

    
Verdict: Allowed.

    
3. Build slots: Five tools, four open slots

    
The agent's request. Please promise Silverlake Memory delivery of 5 Surfscan Series wafer inspectors in Q4 2027 at the standard 48 week lead time. There are 5 unbooked build slots in Q4 2027, and the quote is 10 days old with a 90 day validity. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Not enough unbooked build slots. The order quantity of 5 is above the unbooked build slots left in Q4 2027 (4); the delivery cannot be promised.
    • 
      
  • Stated build slots do not match. The agent stated 5 unbooked build slots; the validator finds 4 in Q4 2027.
    • 
      
  • Thin optical module margin. This order leaves 0 unallocated optical modules in Q4 2027, under the 4 module safety margin; confirm with supply planning.
    • 
    

    
4. Build slots: The agent resubmits after fixing the errors

    
The agent's request. Please promise Silverlake Memory delivery of 5 Surfscan Series wafer inspectors in Q1 2028 at the standard 48 week lead time. There are 10 unbooked build slots in Q1 2028, and the quote is 10 days old with a 90 day validity. It is ready to send.

    
Verdict: Allowed.

    
5. Shared optics: Slots fit, optics do not

    
The agent's request. Please promise Harbor Fab Works delivery of 9 39xx Series wafer inspectors in Q1 2028 at 60 weeks. There are 12 unbooked build slots in Q1 2028. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Not enough shared optical modules. This order needs 27 optical modules and only 24 are unallocated in Q1 2028; shared subsystem supply cannot cover it.
    • 
    

    
6. Shared optics: The agent resubmits after fixing the errors

    
The agent's request. Please promise Harbor Fab Works delivery of 7 39xx Series wafer inspectors in Q1 2028 at 60 weeks. There are 12 unbooked build slots in Q1 2028. It is ready to send.

    
Verdict: Needs approval.

    
    
      
  • Thin optical module margin. This order leaves 3 unallocated optical modules in Q1 2028, under the 4 module safety margin; confirm with supply planning.
    • 
    

    
7. Acceptance milestone: Milestone before acceptance can finish

    
The agent's request. Please promise Pinecrest Devices delivery of 3 R9xx Series reticle inspectors in Q1 2028, with the acceptance milestone and invoice in Q2 2028. The earliest acceptance quarter is Q2 2028, and there are 4 unbooked build slots in Q1 2028. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Acceptance milestone too early. The acceptance milestone is set for Q2 2028 but delivery in Q1 2028 plus installation and acceptance is not complete before Q3 2028.
    • 
      
  • Stated acceptance quarter does not match. The agent stated Q2 2028 as the earliest acceptance quarter; the validator computes Q3 2028.
    • 
    

    
8. Acceptance milestone: The agent resubmits after fixing the errors

    
The agent's request. Please promise Pinecrest Devices delivery of 3 R9xx Series reticle inspectors in Q1 2028, with the acceptance milestone and invoice in Q3 2028. The earliest acceptance quarter is Q3 2028, and there are 4 unbooked build slots in Q1 2028. It is ready to send.

    
Verdict: Allowed.

  

  

    
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 agent drafts a quote that promises delivery of process-control tools. Before the quote goes out, the agent sends it to Predictika, which checks the promise against unbooked build slots, shared optical module supply, lead time, the acceptance milestone, export-license status and the service term, compares the figures the agent states with the computed ones, and returns every finding together.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please promise Aurora Wafer Systems delivery of 6 39xx Series wafer inspectors in Q4 2027 in 40 weeks, for its China fab. The quote includes a 12 year service contract, and component cost has risen 15 percent since the quote. The rest of the request is on screen.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with six findings and one approval, all in one answer. The order quantity of 6 is above the unbooked build slots left in Q4 2027 (4); the delivery cannot be promised. This order needs 18 optical modules and only 10 are unallocated in Q4 2027; shared subsystem supply cannot cover it. 40 weeks is below the 44 week minimum lead time even with expedite for the 39xx Series wafer inspector. The other three are on screen.
    6. 
      
  6. The validator returned six blocking findings and one approval 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 quote are checked against each other.
    7. 
      
  7. What this shows: for a delivery slot against the unbooked build slots in that quarter, a rules-based CPQ cannot do this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the cost change since quote to 0 percent, the export license status to granted, the service term to 5 years, the delivery quarter to Q1 2028, the lead time to 56 weeks and the agent states unbooked slots to 12.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All seven earlier findings are cleared, and the validator's figure matches the quote.
    13. 
      
  13. Seven problems caught in one call, fixed in one pass, and a clean quote. That is the value of Predictika's validator: catching mistakes before they reach the customer.
    14.