Predictika · Microsoft Azure cloud quotes

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

An AI sales agent prepares an Azure quote: VM, Blob storage, Fabric and Azure SQL lines, commitments, agreement term, data transfer, any MACC deal discount and a support plan, in US dollars a month. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the pricing and quoting rules, compares the monthly total the quote states with the computed one, and returns every finding together with the line-item quote. This is an illustrative model built from public product information with illustrative prices; it is not Microsoft's internal rule set.

Scope: Azure cloud quotes (VM, Blob, Fabric and Azure SQL lines, commitments, egress, MACC discount and support plan). See everything this validator covers

This validator models one slice of Azure pricing: quotes built from VM, Blob storage, Microsoft Fabric and Azure SQL lines, with the agreement term, data transfer, MACC deal discount and support plan around them. It handles any quote built from those parts, not only the scenarios shown on this page. It does not cover other Azure services (AKS, Functions, networking beyond egress) or other vendors.

  • Covered: VMs and storage — pay-as-you-go, Spot, reservations and savings plans (including combinations Azure does not sell, such as retired series and ND H100 v5 savings plans), hours planned against full-month commitment billing, and Blob storage classes with minimum durations.
  • Covered: Fabric and Azure SQL — Fabric capacity SKUs, reservation terms and report-viewer licensing, and Azure SQL tiers (serverless against provisioned reserved capacity, the 128-vCore limit, Hybrid Benefit and zone redundancy).
  • Covered: commercial terms — commitment length against the agreement, data transfer with the free allowance, the MACC deal-discount band, support plans, the empty-quote check, and the monthly total the agent states.
  • Not covered — Azure services other than those four, taxes, and real Microsoft contract terms. This model has no approval step: every finding is a block or a warning, so none of the scenarios needs an approval.
  • Labeled as assumptions — prices, discount bands, SQL license adders and the serverless rate are illustrative values built from public product information, not a Microsoft price list.
The AI sales agentBuilds the Azure 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
Monthly total built from compute, storage, analytics, database, transfer, discount and supportOnly with care
The total chains VM prices by commitment type, Blob rates, Fabric capacity rates, SQL vCore and license rates, tiered egress after a free allowance, a spend-based discount and a support fee. A rule set can price each step, but every step has to be kept in step with the others.		The whole chain is computed on every call and returned as line items that add up to the monthly total, the first-year spend and the whole-agreement total.
Commitment length matched to the agreementOnly with care
VMs, Fabric and Azure SQL each have their own commitment options, and each has to fit inside the agreement. Each check works alone; together they are easy to leave out of step.		Every commitment is compared with the agreement term on every call, and a longer one is blocked with the line named.
Reservations and savings plans that Azure does not sell for a VM sizeOnly with workarounds
Which commitments exist depends on the series (retired series, GPU series with no savings-plan rate), and those exceptions are facts about the catalog, not rules about the quote.		The catalog of what is sold is part of the model, so an option that is not sold is blocked and the sold alternatives stay available.
MACC discount limited by annual spendOnly with care
The permitted discount depends on the annual spend after commitments, which is itself the end of a long calculation.		The permitted discount band is computed from the spend on every call and compared with the one requested.
Monthly total that the agent states must match the computed oneOnly with workarounds
A validation rule can compare a stated figure with a calculated one only if the stated figure is a field on the quote, which it rarely is.		The stated monthly total travels with the quote and is compared with the computed total on every call; a mismatch is its own block.

  

  

    
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 reservation or savings-plan commitment that Azure does not sell for a VM size is blocked, for example any reservation on retired series such as B v1 and Fsv2, or a savings plan on ND H100 v5.
2Spot pricing for a VM size that is not offered as Spot is blocked.
3A VM, Fabric or Azure SQL commitment cannot be longer than the agreement; a 3-year commitment on a 1-year agreement is blocked.
4Azure SQL serverless cannot be covered by reserved capacity, and a provisioned tier allows at most its vCore limit per database (128 for General Purpose and Business Critical).
5The MACC deal discount requested is limited to the band for the annual spend; above it the quote is blocked.
6The monthly total the agent states must match the computed total to within one dollar; otherwise the quote is blocked.
7A quote with no service lines is blocked.
8A savings plan where a reservation of the same term saves much more is flagged.
9Spot GPU VMs are flagged because they can be evicted with 30 seconds notice.
10A commitment that costs more than pay-as-you-go at the planned hours is flagged, and so is a commitment billed for every hour of the month.
11A 3-year Fabric reservation is flagged because it costs the same per hour as a 1-year one.
12Fabric capacity below F64 with report viewers is flagged because each viewer needs a Power BI Pro license.
13Azure SQL reserved capacity discounts compute only; the license stays at pay-as-you-go unless Azure Hybrid Benefit applies, and that is flagged.
14Zone redundancy on a tier that does not include it is flagged because the surcharge is not priced.
15Blob storage kept for less than the class minimum duration (Archive, 180 days) is flagged.
16Production-size spend on Developer or Basic support is flagged.
17Egress is priced in tiers after a 100 GB free allowance, the MACC discount applies after commitments but not to support, and support is the plan monthly minimum.

  

  

    
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. GPU commitment: 3-year savings plan on a GPU VM

    
The agent's request. Quote for Halcyon Imaging on Azure: 2 Standard_ND96isr_H100_v5 VMs on a 3-year savings plan. Agreement term: 3 years. Support: Professional Direct. No MACC deal discount. The quote states a monthly total of $144,547.20. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Pricing model not sold for this VM. Azure does not sell 3-year savings plan for Standard_ND96isr_H100_v5: reservations for older series such as B v1 and Fsv2 cannot be bought or renewed since July 1, 2026, and the price list shows no savings-plan rate for ND H100 v5. Use a savings plan or a current series.
    • 
    

    
2. GPU commitment: 1-year reservation instead

    
The agent's request. Quote for Halcyon Imaging on Azure: 2 Standard_ND96isr_H100_v5 VMs on a 1-year reservation. Agreement term: 1 year. Support: Professional Direct. No MACC deal discount. The quote states a monthly total of $92,870.21. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. Monthly total $92,870.21: compute $91,870.21; support $1,000.00.
    • 
    

    
3. MACC discount: 8 percent MACC discount

    
The agent's request. Quote for Marlowe Freight Systems on Azure: 20 Standard_M64s VMs on a 1-year reservation. Agreement term: 1 year. Support: Professional Direct. A 8 percent MACC deal discount is requested. The quote states a monthly total of $52,507.35. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Discount above what is allowed. The requested 8 percent MACC discount exceeds the 5 percent band for about $671,835.00 of annual spend.
    • 
    

    
4. MACC discount: 5 percent discount, inside the limit

    
The agent's request. Quote for Marlowe Freight Systems on Azure: 20 Standard_M64s VMs on a 1-year reservation. Agreement term: 1 year. Support: Professional Direct. A 5 percent MACC deal discount is requested. The quote states a monthly total of $54,186.94. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. Monthly total $54,186.94: compute $55,986.25; discount less $2,799.31; support $1,000.00.
    • 
    

    
5. Stated total: Stated total is low

    
The agent's request. Quote for Brightwater Retail on Azure: 6 Standard_D8s_v5 VMs on a 3-year reservation; 20 TB of Blob Hot storage. Agreement term: 3 years. Support: Standard. Internet data transfer out: 5 TB a month. No MACC deal discount. The quote states a monthly total of $1,496.91. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Stated total does not match. Stated monthly total does not match: the agent stated $1,496.91 a month; the validator computes $1,627.08 a month.
    • 
    

    
6. Stated total: Correct total stated

    
The agent's request. Quote for Brightwater Retail on Azure: 6 Standard_D8s_v5 VMs on a 3-year reservation; 20 TB of Blob Hot storage. Agreement term: 3 years. Support: Standard. Internet data transfer out: 5 TB a month. No MACC deal discount. The quote states a monthly total of $1,627.08. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. Monthly total $1,627.08: compute $664.36; storage $425.98; data transfer $436.74; support $100.00.
    • 
    

    
7. Fabric commitment: 3-year Fabric reservation, 1-year agreement

    
The agent's request. Quote for Larkspur Analytics on Azure: Fabric F128 capacity on a 3-year reservation. Agreement term: 1 year. Support: Standard. No MACC deal discount. The quote states a monthly total of $10,105.37. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Commitment longer than the agreement. A 3-year Fabric reservation is on a 1-year agreement. Deal-desk policy: commitments may not outlast the agreement.
    • 
      
  • No extra discount for 3 years. A 3-year Fabric reservation costs the same per hour as a 1-year one ($938.00 per CU per year); it adds lock-in without extra discount.
    • 
    

    
8. Fabric commitment: 1-year reservation instead

    
The agent's request. Quote for Larkspur Analytics on Azure: Fabric F128 capacity on a 1-year reservation. Agreement term: 1 year. Support: Standard. No MACC deal discount. The quote states a monthly total of $10,105.37. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Quote summary. Monthly total $10,105.37: analytics $10,005.37; support $100.00.
    • 
    

  

  

    
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 Azure quote: VM, Blob storage, Fabric and Azure SQL lines, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the pricing and quoting rules, compares the monthly total the quote states with the computed one, and returns every finding together with the line-item quote. This is an illustrative model built from public product information with illustrative prices; it is not Microsoft's internal rule set.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Quote for Halcyon Imaging on Azure: 2 Standard_ND96isr_H100_v5 VMs on a 3-year savings plan. Agreement term: 3 years. The rest of the request is on screen. The quote states a monthly total of $144,547.20.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with one finding, all in one answer. Azure does not sell 3-year savings plan for Standard_ND96isr_H100_v5: reservations for older series such as B v1 and Fsv2 cannot be bought or renewed since July 1, 2026, and the price list shows no savings-plan rate for ND H100 v5. Use a savings plan or a current series.
    6. 
      
  6. The validator returned one blocking finding 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 reservations and savings plans that Azure does not sell for a VM size, a rules-based CPQ can do this only with workarounds.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the agreement term to 1 year, the service 1 to 2 Standard_ND96isr_H100_v5 VMs on a 1-year reservation and the stated monthly total to $92,870.21.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. The validator also returns the line-item quote, on screen.
    13. 
      
  13. The earlier finding is cleared, and the validator's total, $92,870.21, matches the quote.
    14. 
      
  14. One problem caught in one call, fixed in one pass, and a clean $92,870.21 quote. That is the value of Predictika's validator: catching mistakes before they reach the customer.
    15.