Predictika · Semiconductor inspection system quotes

An AI sales agent's inspection system quote, checked before it goes out

An AI sales agent prepares a quote for semiconductor inspection systems: a model, a substrate, options, the defect size and throughput the customer needs, load ports, a discount and a support plan. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the Onto Innovation inspection line and returns a verdict, every reason and the quote total it computed.

Scope: Onto Innovation Dragonfly G3 and G5, NSX 330, Firefly G5, F30, PrimaScan P and NovusEdge G2 inspection systems with 3D metrology, ClearFind, sub-surface inspection, machine-learning detection, the EB40 module, load ports, discount and support plan. See everything this validator covers

This validator models Onto Innovation inspection system quotes, not just the scenarios shown below.

  • Systems, substrates and options — which wafer sizes or panels each system handles, and which options (3D metrology, ClearFind, sub-surface inspection, machine-learning die-to-die detection, EB40 edge and backside module) each system accepts.
  • Customer requirements — the smallest defect size and the throughput each system can meet (3D metrology lowers throughput), and the load ports each system takes.
  • Commercial limits and total — a sales director's approval for discounts above 10 percent, support terms above 3 years and orders of more than 5 systems, hard limits at a 25 percent discount and a 5-year support term, and the quote total the quote states against the total computed.

Any Onto Innovation inspection quote that touches these rules is in scope — not only the scenarios pictured here. Not covered: metrology, lithography and software lines. The system names and the main options come from Onto Innovation's public material; the compatibility rules, detection and throughput limits, thresholds and all prices are modeled and illustrative (see the note at the foot of the page).

The AI sales agentBuilds the inspection system 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.

6 scenarios in 3 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
Options that exist only on some systemsOnly with care
Each option is available on a different set of systems, so each needs a rule keyed to the model, and the list grows with every new system.		The model and every option are in one request and checked together on every call.
Requirements the chosen system must meetOnly with care
Defect size and throughput limits differ by system, and 3D metrology lowers throughput, so the rules have to be kept in step.		The model, the options and the customer requirements are fields of one request and are checked together on every call.
Approval thresholds beside hard limitsOnly with care
A discount, a support term and an order size each have an approval band and a hard limit, so each needs two rules.		Approvals and blocks come back in one answer, each with its own reason.
A stated total that must match the computed totalOnly with workarounds
The total is built from system, options, load ports, discount and support, so a separate rule is needed to recompute it and compare it with the stated figure.		The stated total is a field of the request and is compared with the computed quote total on every call.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the quote is saved or a step is completed, so a rep often fixes one error and meets the next.		All findings come back together, whatever order the choices arrived in.

  

  

    
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.

    
1The selected substrate (wafer size or panel) is not handled by the selected system.
2The 3D metrology option is not available on the selected system.
3The ClearFind option is available only on Dragonfly systems.
4The machine-learning die-to-die detection option is available only on Dragonfly G5.
5The EB40 edge and backside module cannot be added to the selected system.
6The customer must detect defects of 250 nm; the selected system detects down to 300 nm.
7The customer needs 40 units per hour; the selected configuration delivers at most 36.
8The quote has 5 load ports; the selected system takes at most 4.
9The quote gives a 30 percent discount; the maximum that can be quoted is 25 percent.
10The quote gives a 15 percent discount; discounts above 10 percent need a sales director.
11The support plan runs 6 years; the longest term is 5 years.
12The support plan runs 5 years; terms above 3 years need a person to confirm.
13The quote is for 6 systems; orders of more than 5 systems need a person to confirm delivery and installation capacity.
14The quote states a total of 5,000,000 dollars; the quoted items add up to 5,890,000.

  

  

    
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. Detection, throughput and load ports: A two-system Dragonfly G3 quote with machine-learning detection, a 200 nm defect target, 70 units per hour with 3D metrology and five load ports

    
The agent's request. Please quote 2 Dragonfly G3 systems for 300 mm wafers, with 3D metrology, machine-learning detection, defects down to 200 nm and 70 units per hour, 5 load ports, a 5 percent discount, and the Standard support plan for 3 years, for a total of 5,937,120 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Machine-learning detection not available. The machine-learning die-to-die detection option is available only on Dragonfly G5.
    • 
      
  • Defect size below the detection limit. The customer must detect defects of 200 nm; the selected system detects down to 300 nm.
    • 
      
  • Throughput below the requirement. The customer needs 70 units per hour; the selected configuration delivers at most 36.
    • 
      
  • Too many load ports. The quote has 5 load ports; the selected system takes at most 4.
    • 
    

    
2. Detection, throughput and load ports: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 Dragonfly G5 systems for 300 mm wafers, with 3D metrology, machine-learning detection, defects down to 200 nm and 150 units per hour, 4 load ports, a 5 percent discount, and the Standard support plan for 3 years, for a total of 7,727,680 dollars. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. 2 system(s) at 3,280,000 dollars each before discount; equipment 6,232,000 dollars after discount; support 1,495,680 dollars; quote total 7,727,680 dollars (illustrative prices).
    • 
    

    
3. Substrate, options and load ports: A two-system Firefly G5 quote on 300 mm wafers with ClearFind and two load ports

    
The agent's request. Please quote 2 Firefly G5 systems for 300 mm wafers, with ClearFind, defects down to 600 nm and 20 units per hour, 2 load ports, a 5 percent discount, and the Standard support plan for 3 years, for a total of 5,772,200 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Substrate not supported by the model. The selected substrate (wafer size or panel) is not handled by the selected system.
    • 
      
  • ClearFind not available. The ClearFind option is available only on Dragonfly systems.
    • 
      
  • Too many load ports. The quote has 2 load ports; the selected system takes at most 1.
    • 
    

    
4. Substrate, options and load ports: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 Firefly G5 systems for panels, defects down to 600 nm and 20 units per hour, 1 load port, a 5 percent discount, and the Standard support plan for 3 years, for a total of 5,183,200 dollars. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. 2 system(s) at 2,200,000 dollars each before discount; equipment 4,180,000 dollars after discount; support 1,003,200 dollars; quote total 5,183,200 dollars (illustrative prices).
    • 
    

    
5. Discount, support term and stated total: A six-system NSX 330 quote with a 30 percent discount, a six-year support plan and a quote total

    
The agent's request. Please quote 6 NSX 330 systems for 200 mm wafers, defects down to 1,500 nm and 20 units per hour, 2 load ports, a 30 percent discount, and the Premium support plan for 6 years, for a total of 15,000,000 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Discount above the maximum. The quote gives a 30 percent discount; the maximum that can be quoted is 25 percent.
    • 
      
  • Support term too long. The support plan runs 6 years; the longest term is 5 years.
    • 
      
  • Stated total differs from the quote. The quote states a total of 15,000,000 dollars; the quoted items add up to 11,592,000.
    • 
      
  • Volume order needs approval. The quote is for 6 systems; orders of more than 5 systems need a person to confirm delivery and installation capacity.
    • 
    

    
6. Discount, support term and stated total: The agent resubmits after fixing the errors

    
The agent's request. Please quote 5 NSX 330 systems for 200 mm wafers, defects down to 1,500 nm and 20 units per hour, 2 load ports, an 8 percent discount, and the Premium support plan for 3 years, for a total of 9,798,000 dollars. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. 5 system(s) at 1,500,000 dollars each before discount; equipment 6,900,000 dollars after discount; support 2,898,000 dollars; quote total 9,798,000 dollars (illustrative prices).
    • 
    

  

  

    
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 sales agent prepares a quote for semiconductor inspection systems: a model, a substrate, options, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the Onto Innovation inspection line and returns a verdict, every reason and the quote total it computed.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote 2 Dragonfly G3 systems for 300 mm wafers, with 3D metrology, machine-learning detection, defects down to 200 nm and 70 units per hour. It is ready to send. The rest of the request is on screen.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with four findings, all in one answer. The machine-learning die-to-die detection option is available only on Dragonfly G5. The customer must detect defects of 200 nm; the selected system detects down to 300 nm. The customer needs 70 units per hour; the selected configuration delivers at most 36. The other one is on screen.
    6. 
      
  6. The validator returned four 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 quote are checked against each other.
    7. 
      
  7. What this shows: for requirements the chosen system must meet, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the system to Dragonfly G5, the throughput needed to 150 units per hour, the load ports to 4 and the quote total stated on the quote to 7,727,680 dollars.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All four earlier findings are cleared.
    13. 
      
  13. Four 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. Now you can try the other scenarios yourself.
    14.