Predictika · Cloud server quotes

An AI sales agent's virtual server quote, checked before it goes out

An AI sales agent prepares a monthly quote for Samsung Cloud Platform virtual servers: the server family, vCores and memory, a GPU model and card count where it applies, the operating system, Block Storage, a Load Balancer, backup and NAT IP, the commitment term, a negotiated discount, and the monthly total it states. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the offering and returns a verdict, every reason and the total it computed.

Scope: Samsung Cloud Platform Virtual Server (Standard, High Capacity, GPU Server) with Block Storage, Load Balancer, Backup, NAT IP, commitment terms and a negotiated discount. See everything this validator covers

This validator models Virtual Server quotes, not just the scenarios shown below.

  • Server sizes — the vCore and memory range of the Standard and High Capacity families, the memory per vCore, and the GPU models (A100, H100, B300) and card counts a GPU Server takes.
  • Images, storage and services — the operating system images and where Windows is allowed, Block Storage size and type, the Load Balancer needing two servers, Backup and NAT IP.
  • Term, discount and total — the on-demand, one-year and three-year terms, a negotiated discount that needs a person above 10 percent and is blocked above 30, large orders and the B300 for a person to confirm, and the monthly total the agent states against the validator's own.

Any Virtual Server quote that touches these rules is in scope — not only the scenarios pictured here. Not covered: Kubernetes Engine, databases, Object and File Storage, networking products and bare metal. Public versus modeled: the product names, the Standard and High Capacity tiers, the GPU models and the standard operating systems of the GPU Server follow Samsung SDS public material; every price (in cents a month), the size limits, the GPU card counts, the storage range, the discount limits, the quotas and the term factors are modeled by Predictika and are not Samsung SDS prices or terms.

The AI sales agentBuilds the virtual server 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 depend on the server familyOnly with care
GPU models, card counts, operating systems and disk types each depend on the family chosen in another field, so each needs a rule tied to that choice.		The family and every option are fields of one quote and are checked together on every call.
Size limits that depend on the family and on each otherOnly with care
The vCore and memory ceilings differ by family and memory must stay within a range per vCore, so the limits interact.		The family, vCores and memory are in one request and are compared on every call.
A stated total that must equal the computed totalOnly with workarounds
The total is built from servers, storage, services, the term and the discount; checking a stated figure against it takes extra formulas that interact with the other rules.		The stated total is a field of the request and is compared with the computed total.
Quotes that should go to a person rather than be blockedOnly with care
Large orders, the B300 and mid-size discounts need a person, and sit beside separate hard limits that must not overlap them.		Blocks and approvals come back together; approval-only quotes are not blocked.
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.

    
1A Standard virtual server is up to 16 vCores and 256 GB; the quote asks for 17 vCores and 136 GB. Use High Capacity.
2A Standard virtual server is up to 16 vCores and 256 GB; the quote asks for 16 vCores and 257 GB. Use High Capacity.
3257 GB of memory with 16 vCores is outside the 2 to 16 GB per vCore range.
4A High Capacity virtual server is modeled up to 128 vCores and 2,048 GB; the quote asks for 129 vCores and 1,500 GB.
5An order of 258 vCores in total is above the modeled 256 vCore account quota; a person should confirm capacity.
6A High Capacity virtual server is modeled up to 128 vCores and 2,048 GB; the quote asks for 128 vCores and 2,049 GB.
72,049 GB of memory with 128 vCores is outside the 2 to 16 GB per vCore range.
815 GB of memory with 8 vCores is outside the 2 to 16 GB per vCore range.
9129 GB of memory with 8 vCores is outside the 2 to 16 GB per vCore range.
10The quote has 0 servers; at least 1 is needed.
11A GPU Server needs a GPU model (A100, H100 or B300).
12A GPU model is chosen but the server family is not GPU Server.
13A GPU Server has 1, 2, 4 or 8 GPU cards; the quote asks for 3.
14B300 servers are a newer GPU option; a person should confirm availability and lead time.
15The B300 is modeled as an 8-card server only; the quote asks for 4 cards.
16GPU Server standard OS images are RHEL and Ubuntu; Windows is not available.
17A GPU Server needs SSD Block Storage; HDD is not allowed.
18Block Storage of 9 GB per server is outside the modeled 10 to 16,384 GB range.
19Block Storage of 16,385 GB per server is outside the modeled 10 to 16,384 GB range.
20A Load Balancer needs at least 2 servers behind it; the quote has 1.
21An order of 264 vCores in total is above the modeled 256 vCore account quota; a person should confirm capacity.
22An order of 24 GPU cards in total is above the modeled 16 card account quota; a person should confirm capacity.
23A negotiated discount of 11 percent is above the modeled 10 percent sales limit; a person should approve it.
24A negotiated discount of 30 percent is above the modeled 10 percent sales limit; a person should approve it.
25A negotiated discount of 31 percent is outside the 0 to 30 percent range.
26The quote states a monthly total of 120,800 cents; the modeled prices give 120,900 cents.
27The quote states a monthly total of 120,901 cents; the modeled prices give 120,900 cents.
28The quote states a monthly total of 1 cents; the modeled prices give 1,023,300 cents.
29An order of 320 vCores in total is above the modeled 256 vCore account quota; a person should confirm capacity.
30A negotiated discount of 15 percent is above the modeled 10 percent sales limit; a person should approve it.

  

  

    
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. A GPU Server asked for the wrong way: A B300 GPU Server with 3 cards, Windows and HDD Block Storage

    
The agent's request. Please quote 2 GPU Server virtual servers with 3 B300 GPU cards, with 8 vCores and 32 GB of memory each, Windows, and HDD Block Storage of 1000 GB each, on an on-demand term, for a monthly total of 26,508.00 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • GPU count. A GPU Server has 1, 2, 4 or 8 GPU cards; the quote asks for 3.
    • 
      
  • B300 only as 8 GPU. The B300 is modeled as an 8-card server only; the quote asks for 3 cards.
    • 
      
  • GPU server OS image. GPU Server standard OS images are RHEL and Ubuntu; Windows is not available.
    • 
      
  • GPU server disk. A GPU Server needs SSD Block Storage; HDD is not allowed.
    • 
      
  • B300 availability. B300 servers are a newer GPU option; a person should confirm availability and lead time.
    • 
    

    
2. A GPU Server asked for the wrong way: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 GPU Server virtual servers with 4 A100 GPU cards, with 8 vCores and 32 GB of memory each, Ubuntu, and SSD Block Storage of 1000 GB each, on an on-demand term, for a monthly total of 14,840.00 dollars. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 2 server(s), 16 vCores or 8 GPU cards in total, estimated monthly total of 1,484,000 cents at illustrative prices.
    • 
    

    
3. An oversized Standard server behind a Load Balancer: One Standard server with 24 vCores and 400 GB of memory behind a Load Balancer

    
The agent's request. Please quote 1 Standard virtual server with 24 vCores and 400 GB of memory each, Ubuntu, and SSD Block Storage of 200 GB each, with a Load Balancer, on an on-demand term, for a monthly total of 1,777.00 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Standard server too large. A Standard virtual server is up to 16 vCores and 256 GB; the quote asks for 24 vCores and 400 GB. Use High Capacity.
    • 
      
  • Memory to vCore ratio. 400 GB of memory with 24 vCores is outside the 2 to 16 GB per vCore range.
    • 
      
  • Load Balancer needs two servers. A Load Balancer needs at least 2 servers behind it; the quote has 1.
    • 
    

    
4. An oversized Standard server behind a Load Balancer: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 High Capacity virtual servers with 24 vCores and 192 GB of memory each, Ubuntu, and SSD Block Storage of 200 GB each, with a Load Balancer, on an on-demand term, for a monthly total of 3,049.00 dollars. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. 2 server(s), 48 vCores or 0 GPU cards in total, estimated monthly total of 304,900 cents at illustrative prices.
    • 
    

    
5. A large order, a discount and a wrong total: An order of 33 Standard servers with a 15 percent negotiated discount

    
The agent's request. Please quote 33 Standard virtual servers with 8 vCores and 32 GB of memory each, Ubuntu, and SSD Block Storage of 200 GB each, with a Load Balancer, on an on-demand term, with a negotiated discount of 15 percent, for a monthly total of 0.01 dollars. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Stated total does not match. The quote states a monthly total of 1 cents; the modeled prices give 844,660 cents.
    • 
      
  • Large vCore order. An order of 264 vCores in total is above the modeled 256 vCore account quota; a person should confirm capacity.
    • 
      
  • Discount above 10 percent. A negotiated discount of 15 percent is above the modeled 10 percent sales limit; a person should approve it.
    • 
    

    
6. A large order, a discount and a wrong total: The agent resubmits after fixing the errors

    
The agent's request. Please quote 33 Standard virtual servers with 8 vCores and 32 GB of memory each, Ubuntu, and SSD Block Storage of 200 GB each, with a Load Balancer, on an on-demand term, with a negotiated discount of 15 percent, for a monthly total of 8,446.60 dollars. It is ready to send.

    
Verdict: Needs approval.

    
    
      
  • Large vCore order. An order of 264 vCores in total is above the modeled 256 vCore account quota; a person should confirm capacity.
    • 
      
  • Discount above 10 percent. A negotiated discount of 15 percent is above the modeled 10 percent sales limit; a person should approve it.
    • 
      
  • Order summary. 33 server(s), 264 vCores or 0 GPU cards in total, estimated monthly total of 844,660 cents at 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 monthly quote for Samsung Cloud Platform virtual servers: the server family, vCores and memory, a GPU model and card count where it applies, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the offering and returns a verdict, every reason and the total it computed.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote 2 GPU Server virtual servers with 3 B300 GPU cards, with 8 vCores and 32 GB of memory each. 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 and one approval, all in one answer. A GPU Server has 1, 2, 4 or 8 GPU cards; the quote asks for 3. The B300 is modeled as an 8-card server only; the quote asks for 3 cards. GPU Server standard OS images are RHEL and Ubuntu; Windows is not available. The other one is on screen.
    6. 
      
  6. The validator returned four 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 options that depend on the server family, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the GPU model to A100, the GPU cards per server to 4, the operating system to ubuntu, the Block Storage type to SSD and the stated monthly total to 1,484,000 cents.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All five earlier findings are cleared, and the validator's total, 14,840.00 dollars a month, matches the quote.
    13. 
      
  13. Five 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.