Predictika · Cloud infrastructure orders

An AI sales agent's cloud instance order, checked before it goes out

An AI sales agent prepares an order for Public Cloud instances for a software company: instance family and size, number of instances, Windows licence, Savings Plan term and Local Zone deployment, with the total vCores it expects. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the published product information and returns a verdict, every reason and the figures it computed.

Scope: OVHcloud Public Cloud instances (General Purpose B3, Compute Optimised C3, Memory Optimised R3, Discovery D2): size, count, Windows licence, Savings Plan term, Local Zone, vCore total. See everything this validator covers

This validator models orders for four OVHcloud Public Cloud instance families, not just the scenarios shown below.

  • Families and sizes — B3 (8 to 640 GB), C3 (4 to 320 GB), R3 (16 to 1,024 GB) and Discovery D2 (2 to 8 GB); the instance name gives the memory, and the model knows the vCores of each size (for example b3-16 has 4 vCores, c3-16 has 8 and r3-16 has 2).
  • Savings Plans — 12 or 36 month plans, available on B3, C3 and R3 instances; Local Zone resources are not eligible.
  • Windows and Local Zones — the Windows licence is billed per vCore and hour; the model offers Windows and Local Zones on B3, C3 and R3, and sends Local Zone orders to a person.
  • Totals and approvals — the stated vCore total compared with instances times vCores, and approvals for the largest sizes, large vCore totals and many Discovery instances.

Any order for these families that touches these rules is in scope — not only the scenarios pictured here. Not covered: prices (this validator produces no price build-up or itemized list; OVHcloud publishes prices but they were not modeled), the older B2, C2 and R2 generations, storage-optimised I1 instances, GPU, Metal and other OVHcloud services, regions, and block storage. Several limits are modeled by Predictika from comparable cloud offers rather than published by OVHcloud; the footer says which.

The AI sales agentBuilds the instance order 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
An instance size that the chosen family does not offerOnly with care
A product catalog can list sizes per family, but the name, memory and vCores must stay in step across more than 20 sizes.		Family and size are checked together, and the vCores of each size are computed.
A Savings Plan on a Discovery instance or a Local ZoneOnly with care
Eligibility rules depend on family, location and term together and are easy to miss in combination.		Plan term, family and location are checked together on every call.
A stated vCore total that must match instances times vCoresOnly with care
A stated figure is a free-entry field; a rule can compare it with a calculation, but someone has to write and maintain that comparison.		The stated total travels with the order and is compared with the computed one on every call.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the order 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 order has 0 instances; at least 1 is needed.
2Instances of 160 vCores are the largest in the family; a person should confirm capacity in the region (modeled).
3An instance with 4 GB of memory is not offered in the selected family.
4An instance with 1,024 GB of memory is not offered in the selected family.
5An instance with 12 GB of memory is not offered in the selected family.
6An instance with 640 GB of memory is not offered in the selected family.
7An instance with 2 GB of memory is not offered in the selected family.
8Instances of 128 vCores are the largest in the family; a person should confirm capacity in the region (modeled).
9An instance with 8 GB of memory is not offered in the selected family.
10An instance with 16 GB of memory is not offered in the selected family.
11A Savings Plan of 24 months is not offered; the plans are 12 or 36 months, or no commitment.
12A Savings Plan of 6 months is not offered; the plans are 12 or 36 months, or no commitment.
13Savings Plans apply to B3, C3 and R3 instances; Discovery instances are not eligible.
14Local Zone resources are not eligible for Savings Plans.
15A Local Zone deployment needs a person to confirm availability and additional IP address costs.
16A Windows licence is requested on a Discovery instance; the model offers Windows on B3, C3 and R3.
17A Windows licence on a Savings Plan; a person should confirm how the licence is billed (modeled).
18A Local Zone deployment is requested on a Discovery instance; the model offers Local Zones on B3, C3 and R3.
19The order states 20 vCores in total; 4 instances of 4 vCores make 16.
20The order states 8 vCores in total; 4 instances of 4 vCores make 16.
21The order totals 512 vCores; a person should confirm the project quota (modeled threshold of 512 vCores).
2210 Discovery instances share resources and suit test and development; a person should confirm the use (modeled).

  

  

    
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. General Purpose: A General Purpose order with an unavailable size, a 24 month plan and a Local Zone

    
The agent's request. Please order 6 General Purpose B3 instances with 100 GB of memory each (b3-100), on a 24 month Savings Plan, with a Windows licence, in a Local Zone. The order states 200 vCores in total. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Size not offered for this family. An instance with 100 GB of memory is not offered in the selected family.
    • 
      
  • Savings Plan term not offered. A Savings Plan of 24 months is not offered; the plans are 12 or 36 months, or no commitment.
    • 
      
  • Local Zone not eligible for Savings Plans. Local Zone resources are not eligible for Savings Plans.
    • 
      
  • Stated vCore total does not match. The order states 200 vCores in total; 6 instances of 0 vCores make 0.
    • 
      
  • Local Zone deployment. A Local Zone deployment needs a person to confirm availability and additional IP address costs.
    • 
      
  • Windows licence on a Savings Plan. A Windows licence on a Savings Plan; a person should confirm how the licence is billed (modeled).
    • 
    

    
2. General Purpose: The agent resubmits after fixing the errors

    
The agent's request. Please order 6 General Purpose B3 instances with 128 GB of memory each (b3-128), on a 12 month Savings Plan, without a Windows licence, in a standard region. The order states 192 vCores in total. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 6 instance(s) of 32 vCores, 192 vCores in total.
    • 
    

    
3. Discovery and compute: A Discovery order with a Savings Plan and Windows

    
The agent's request. Please order 12 Discovery D2 instances with 4 GB of memory each (d2-4), on a 12 month Savings Plan, with a Windows licence, in a standard region. The order states 24 vCores in total. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Savings Plans only on B3, C3 and R3. Savings Plans apply to B3, C3 and R3 instances; Discovery instances are not eligible.
    • 
      
  • Windows licence not on Discovery instances. A Windows licence is requested on a Discovery instance; the model offers Windows on B3, C3 and R3.
    • 
      
  • Many Discovery instances. 12 Discovery instances share resources and suit test and development; a person should confirm the use (modeled).
    • 
      
  • Windows licence on a Savings Plan. A Windows licence on a Savings Plan; a person should confirm how the licence is billed (modeled).
    • 
    

    
4. Discovery and compute: The agent resubmits after fixing the errors

    
The agent's request. Please order 12 Compute Optimised C3 instances with 32 GB of memory each (c3-32), on a 12 month Savings Plan, without a Windows licence, in a standard region. The order states 192 vCores in total. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 12 instance(s) of 16 vCores, 192 vCores in total.
    • 
    

    
5. Memory Optimised: A very large Memory Optimised order with a wrong stated total

    
The agent's request. Please order 5 Memory Optimised R3 instances with 1024 GB of memory each (r3-1024), on a 36 month Savings Plan, without a Windows licence, in a standard region. The order states 600 vCores in total. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Stated vCore total does not match. The order states 600 vCores in total; 5 instances of 128 vCores make 640.
    • 
      
  • Largest instance sizes. Instances of 128 vCores are the largest in the family; a person should confirm capacity in the region (modeled).
    • 
      
  • Large vCore total. The order totals 640 vCores; a person should confirm the project quota (modeled threshold of 512 vCores).
    • 
    

    
6. Memory Optimised: The agent resubmits after fixing the errors

    
The agent's request. Please order 4 Memory Optimised R3 instances with 256 GB of memory each (r3-256), on a 36 month Savings Plan, without a Windows licence, in a standard region. The order states 128 vCores in total. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 4 instance(s) of 32 vCores, 128 vCores in total.
    • 
    

  

  

    
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 an order for Public Cloud instances for a software company: instance family and size, number of instances, Windows licence, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the published product information 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 order 6 General Purpose B3 instances with 100 GB of memory each (b3-100), on a 24 month Savings Plan, with a Windows licence, in a Local Zone. The order states 200 vCores in total. 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 and two approvals, all in one answer. An instance with 100 GB of memory is not offered in the selected family. A Savings Plan of 24 months is not offered; the plans are 12 or 36 months, or no commitment. Local Zone resources are not eligible for Savings Plans. The other one is on screen.
    6. 
      
  6. The validator returned four blocking findings and two approvals 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 and against the stated total.
    7. 
      
  7. What this shows: for a Savings Plan on a Discovery instance or a Local Zone, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the memory per instance to 128 GB, the Savings Plan to 12 months, the windows licence to no, the Local Zone to no and the stated total vCores to 192.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All six earlier findings are cleared, and the validator's figure matches the quote.
    13. 
      
  13. Six 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.