Predictika · Power generation

An AI sales agent's gas turbine plant quote, checked before it goes out

An AI sales agent prepares a quote for a gas turbine power plant: a turbine from the Ansaldo Energia heavy-duty range (AE64.3A, AE94.2, AE94.3A, GT26 or GT36-S5) in simple cycle or combined cycle, the number of plant blocks, grid frequency, fuel and hydrogen blend, the site conditions, a service agreement and a discount. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model and returns a verdict, every reason and the figures it computed. The turbine names and ratings are from the public Ansaldo Energia brochure; the prices and several of the rules are modeled for this demonstration and are not Ansaldo Energia's.

Scope: Ansaldo Energia heavy-duty gas turbine plants (AE64.3A, AE94.2, AE94.3A, GT26, GT36-S5), simple or combined cycle: ratings, grid frequency, hydrogen blend, fuel, site derating, service and an illustrative price. See everything this validator covers

This validator models a quote for one turbine model at a time, in simple cycle, 1x1 combined cycle or 2x1 combined cycle blocks, not just the scenarios shown below. It separates what is public from what is modeled.

  • Public, from the Ansaldo Energia heavy-duty gas turbine brochure (April 2025) — the five models, their ISO simple-cycle output (GT36-S5 563 MW, GT26 370, AE94.3A 340, AE94.2 190, AE64.3A 78), the combined-cycle output ranges, the grid frequency (50 Hz, with the AE64.3A also listed for 60 Hz) and the hydrogen capability of each (GT36 up to 70 percent by volume, GT26 45, AE94.3A, AE94.2 and AE64.3A 40).
  • Modeled, not published by the company — every price (turbines, steam turbine and heat recovery, balance of plant, service) is an illustrative figure typical for the industry, not an Ansaldo Energia price. So are the output loss with site temperature and altitude, the fuel restrictions (heavy oil only for the E-class models), the service tiers and years, the discount limits and the approval thresholds (hydrogen above 30 percent, altitude above 1,000 metres, heavy oil, discount above 5 percent).
  • An assumption about the combined-cycle ratings — the brochure gives each model a combined-cycle range "depending on configuration". The lower and upper values are taken here as a 1x1 block (one gas turbine and one steam turbine) and a 2x1 block (two gas turbines and one steam turbine).
  • What is checked — frequency, hydrogen blend against the turbine's rating, fuel, site output against the customer's requirement, service tier and years, discount, and the output and net price the agent states, which must equal the computed ones.

Any gas turbine plant quote that touches these rules is in scope — not only the scenarios pictured here. Not covered: the AE-T100 microturbine, steam turbine and generator choices beyond the block types above, emissions permits, grid connection, financing, real performance guarantees, and any real Ansaldo Energia price or term. The customer, the site and the prices on this page are fictional or illustrative.

The AI sales agentBuilds the plant 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.

7 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
Only the grid frequencies, hydrogen blends and fuels each turbine is listed forOnly with workarounds
Each of the five models has its own frequency and hydrogen rating, and the fuel limits differ by class. A rules-based CPQ needs a table or one rule per model, and every brochure update means editing them.		Model, frequency, hydrogen blend and fuel arrive in one request and are compared with each model's row on every call.
Plant output that depends on blocks, configuration and the siteOnly with care
Output is a per-block rating times the number of blocks, reduced by temperature and altitude, then compared with the customer's requirement. It works, but the derating has to be kept in step with the ratings.		The rated output, the site output and the requirement are computed and compared together.
A price that adds turbines, steam cycle, balance of plant and serviceOnly with care
Several price rules that depend on the model, the configuration and the service years, and a stated price has to be recomputed and compared.		The price build-up is computed on every call and compared with the net price the agent states.
Discount thresholds that trigger approval or a blockOnly with care
Two thresholds as percentages of a gross price that other rules change; a discount rule has to run after every price rule.		The gross price, both thresholds and the discount are evaluated together, in any order.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the quote is saved or a step is completed, so a person 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 gas turbine is listed for 50 Hz only; the AE64.3A is the model listed for 50 and 60 Hz.
2A hydrogen blend above 30 percent by volume needs a combustion engineering review before it is promised.
3A hydrogen blend of 41 percent by volume is above the 40 percent the selected turbine is listed for.
4A hydrogen blend of 46 percent by volume is above the 45 percent the selected turbine is listed for.
5A hydrogen blend of 71 percent by volume is above the 70 percent the selected turbine is listed for.
6A hydrogen blend is quoted only with natural gas as the base fuel.
7Heavy oil firing needs an emissions and maintenance review.
8Heavy oil is quoted only for the E-class turbines (AE64.3A and AE94.2).
9The customer needs 78 MW at the site; the plant gives 77.454 MW after the site derating.
10The site temperature of 46 degrees Celsius is above the 45 degree limit of the output model.
11A site above 1,000 metres needs an engineering check of the output and cooling.
12The quote has 0 plant blocks; at least 1 is needed.
13The service tier and the number of service years do not agree; a tier needs 1 to 20 years and no tier takes none. The quote has 5 years.
14The service tier and the number of service years do not agree; a tier needs 1 to 20 years and no tier takes none. The quote has 0 years.
15The service tier and the number of service years do not agree; a tier needs 1 to 20 years and no tier takes none. The quote has 21 years.
16The discount is more than 5 percent of the gross price; a sales director should approve it.
17The discount of $63.301 million is more than 20 percent of the $316.5 million gross price.
18The quote states 480 MW of plant output at ISO conditions; the listed ratings give 495 MW.
19The quote states a net price of $316.501 million; the modeled price build-up gives $316.5 million.
20A hydrogen blend of 60 percent by volume is above the 45 percent the selected turbine is listed for.

  

  

    
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. Hydrogen blend and turbine rating: Two GT26 combined-cycle blocks with a hydrogen blend above the turbine's rating

    
The agent's request. Please quote 2 combined cycle 1x1 (one gas turbine and one steam turbine) blocks with the GT26 on a 50 Hz grid, firing natural gas with a hydrogen blend of 55 percent by volume. The site is at 30 degrees Celsius and the customer needs 800 MW at the site. The quote includes a full service agreement for 12 years. The quote states 1,080 MW of plant output at ISO conditions and a net price of $726.48 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Hydrogen blend above the turbine capability. A hydrogen blend of 55 percent by volume is above the 45 percent the selected turbine is listed for.
    • 
      
  • Hydrogen blend above 30 percent. A hydrogen blend above 30 percent by volume needs a combustion engineering review before it is promised.
    • 
    

    
2. Hydrogen blend and turbine rating: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 combined cycle 1x1 (one gas turbine and one steam turbine) blocks with the GT26 on a 50 Hz grid, firing natural gas with a hydrogen blend of 25 percent by volume. The site is at 30 degrees Celsius and the customer needs 800 MW at the site. The quote includes a full service agreement for 12 years. The quote states 1,080 MW of plant output at ISO conditions and a net price of $726.48 million. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Plant summary. 1,080 MW of plant output at ISO conditions, 966.6 MW at the site, modeled net price of $726.48 million.
    • 
    

    
3. Grid frequency and site output: An AE94.3A combined-cycle block for a 60 Hz grid at a hot, high site

    
The agent's request. Please quote 1 combined cycle 1x1 (one gas turbine and one steam turbine) block with the AE94.3A on a 60 Hz grid, firing natural gas. The site is at 38 degrees Celsius and 400 metres above sea level and the customer needs 350 MW at the site. The quote includes a full service agreement for 10 years. The quote states 495 MW of plant output at ISO conditions and a net price of $316.5 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Grid frequency not offered for this turbine. The selected gas turbine is listed for 50 Hz only; the AE64.3A is the model listed for 50 and 60 Hz.
    • 
    

    
4. Grid frequency and site output: The agent resubmits after fixing the errors

    
The agent's request. Please quote 2 combined cycle 2x1 (two gas turbines and one steam turbine) blocks with the AE64.3A on a 60 Hz grid, firing natural gas. The site is at 38 degrees Celsius and 400 metres above sea level and the customer needs 350 MW at the site. The quote includes a full service agreement for 10 years. The quote states 486 MW of plant output at ISO conditions and a net price of $329.4 million. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Plant summary. 486 MW of plant output at ISO conditions, 384.426 MW at the site, modeled net price of $329.4 million.
    • 
    

    
5. Discount and net price: A GT36-S5 2x1 combined-cycle plant with a 25 percent discount and a net price that leaves it out

    
The agent's request. Please quote 1 combined cycle 2x1 (two gas turbines and one steam turbine) block with the GT36-S5 on a 50 Hz grid, firing natural gas. The site is at 30 degrees Celsius and the customer needs 1,200 MW at the site. The quote includes a full service agreement for 15 years and a discount of $285.81 million. The quote states 1,600 MW of plant output at ISO conditions and a net price of $1,143.24 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Discount above the limit. The discount of $285.81 million is more than 20 percent of the $1,143.24 million gross price.
    • 
      
  • Stated net price does not match. The quote states a net price of $1,143.24 million; the modeled price build-up gives $857.43 million.
    • 
      
  • Discount above 5 percent. The discount is more than 5 percent of the gross price; a sales director should approve it.
    • 
    

    
6. Discount and net price: The agent resubmits after fixing the errors

    
The agent's request. Please quote 1 combined cycle 2x1 (two gas turbines and one steam turbine) block with the GT36-S5 on a 50 Hz grid, firing natural gas. The site is at 30 degrees Celsius and the customer needs 1,200 MW at the site. The quote includes a full service agreement for 15 years and a discount of $114.324 million. The quote states 1,600 MW of plant output at ISO conditions and a net price of $1,028.916 million. It is ready to send.

    
Verdict: Needs approval.

    
    
      
  • Discount above 5 percent. The discount is more than 5 percent of the gross price; a sales director should approve it.
    • 
      
  • Plant summary. 1,600 MW of plant output at ISO conditions, 1,432 MW at the site, modeled net price of $1,028.916 million.
    • 
    

    
7. Discount and net price: The agent resubmits after fixing the errors

    
The agent's request. Please quote 1 combined cycle 2x1 (two gas turbines and one steam turbine) block with the GT36-S5 on a 50 Hz grid, firing natural gas. The site is at 30 degrees Celsius and the customer needs 1,200 MW at the site. The quote includes a full service agreement for 15 years and a discount of $57.162 million. The quote states 1,600 MW of plant output at ISO conditions and a net price of $1,086.078 million. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Plant summary. 1,600 MW of plant output at ISO conditions, 1,432 MW at the site, modeled net price of $1,086.078 million.
    • 
    

  

  

    
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 a gas turbine power plant: a turbine from the Ansaldo Energia heavy-duty range (AE64.3A, AE94.2, AE94.3A, GT26 or GT36-S5) in simple cycle or combined cycle, the number of plant blocks, grid frequency, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model and returns a verdict, every reason and the figures it computed. The turbine names and ratings are from the public Ansaldo Energia brochure; the prices and several of the rules are modeled for this demonstration and are not Ansaldo Energia's.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote 1 combined cycle 2x1 (two gas turbines and one steam turbine) block with the GT36-S5 on a 50 Hz grid, firing natural gas. The site is at 30 degrees Celsius and the customer needs 1,200 MW at the site. The rest of the request is on screen. The quote states 1,600 MW of plant output at ISO conditions and a net price of $1,143.24 million.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with two findings and one approval, all in one answer. The discount of $285.81 million is more than 20 percent of the $1,143.24 million gross price. The quote states a net price of $1,143.24 million; the modeled price build-up gives $857.43 million.
    6. 
      
  6. The validator returned two 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 price that adds turbines, steam cycle, balance of plant and service, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the discount to 114.324 and the stated net price as to 1,028.916.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. Not blocked, but held for a person.
    12. 
      
  12. The two blocking findings are cleared; one approval remains for a person: discount above 5 percent.
    13. 
      
  13. The agent resubmits once more.
    14. 
      
  14. The revised request. It changes the discount to 57.162 and the stated net price as to 1,086.078.
    15. 
      
  15. It calls the validator once more.
    16. 
      
  16. This time the validator clears it.
    17. 
      
  17. The earlier finding is cleared, and the validator's total, $1,086.078M, matches the quote.
    18. 
      
  18. Three problems caught in one call, fixed in two passes, 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.
    19.