Predictika · Power Generation Quote Check

Before an AI agent sends a generator quote, the quote is checked against the vendor’s sizing and sales rules

Halvern Power Systems, a fictional maker of standby generator sets, lets an AI quoting agent assemble quotes for sites that need backup power. Before the agent sends one, the call goes to Predictika’s Power Generation Quote Check. It sizes the site from the rating, the altitude and temperature derate and the units left in service, applies the use and emissions rules, checks paralleling, sound, fuel runtime and lead time, applies the discount authority, prices the quote, and checks that the total the agent stated matches the computed one. Then it returns one verdict: allow, needs approval, or block.

Illustrative only. Halvern Power Systems and its customers are fictional. Models, ratings, prices, limits and lead times are invented for this demonstration and are not technical or regulatory guidance.

Scope: one generator quote, judged against sizing, use, site and discount rulesSee everything this validator covers

This validator decides one thing: whether one generator quote may be sent, given what is in it. It is not a sizing tool or a price book, and it does not know the customer.

  • Actions — send one quote for Halvern HX-series generator sets, with their enclosure, fuel tank and paralleling options.
  • Rules — site capacity against the load after derate and redundancy, use and rating, Tier 4 Final, paralleling, combined sound, fuel runtime, lead time, discount authority, the price, and the stated total.

Within these rules the validator handles any combination of the listed choices, not just the seven scenarios on this page. It never sees the customer, the site or the order system; those facts are taken as stated.

What the validator sees: the quote the agent proposes, never the site or the order systemRead more

The validator takes the quote as inputs: the model and number of units, the rating and redundancy, the load, the use, emissions, enclosure and fuel tank, the site sound limit, altitude and temperature, the time available, the discount, and the total the agent stated. It computes the figures and applies the rules.

Two honest limits. The validator judges the quote it is given: if the agent gets a fact about the site wrong, the answer follows it. And a quote that never calls the validator is not checked; in this demo the agent calls it itself, and in a deployment the quoting workflow makes the call.

The AI agentProposes sending a quote, says why it is fine, then calls the validator before sending.
Predictika validatorReads the quote, sizes the site, prices it, applies the rules, then returns one verdict.

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 agent says

Illustrative agent statement. The validator receives the structured request below.

Proposing agent: AI agent (scripted)

Before sending, the agent must call Recorded run

Predictika validator


      
Waiting for the call.

Where quoting tools get it wrong

Click to expand

General statements about quoting tools, not about any named product. We rest this on public descriptions of how such tools work; anyone running a specific product will know it better than we do.

What the validator caughtTypical toolPredictika
Site capacity from the rating, the derate and the units left in serviceCannot do it
Capacity depends on the rating, a derate that combines altitude and temperature, and the units remaining after the reserve. Chaining those into one comparison with the load usually takes several rules that must run in order.
Every rule is checked on every call and the verdict lists each finding.
Combined sound level across unit countOnly with workarounds
The level of several identical units adds on a logarithmic scale. A rules-based quoting tool usually needs a lookup table kept by hand.
Every rule is checked on every call and the verdict lists each finding.
Use, rating and emissions rulesOnly if someone is vigilant
Restrictions by use category are simple validation rules. Keeping them complete as models and tiers are added is the work.
Every rule is checked on every call and the verdict lists each finding.
Discount authority that depends on the number of unitsOnly with workarounds
A distributor limit, a quantity-based maximum and an approval step need separate rules.
Every rule is checked on every call and the verdict lists each finding.
Checking the total the agent statedCannot do it
Comparing a stated total with a computed one needs a rule that does both, and price rules rarely run on stated figures.
Every rule is checked on every call and the verdict lists each finding.
Site capacity below the critical loadOnly with workarounds
A quoting tool can check a unit count. Sizing the site from the rating, the derate and the units left in service, then comparing it with the load and checking the stated total, takes rules written to read them together.
Every rule is checked on every call and the verdict lists each finding.
Standby rating used for non-emergency operationOnly if someone is vigilant
Use-category rules sit in a compliance note. A rule has to read the use, the rating and the emissions tier together.
Every rule is checked on every call and the verdict lists each finding.
Combined sound level above the site limitOnly if someone is vigilant
Sound adds up across units on a logarithmic scale, and fuel runtime depends on the tank. A rule has to read the enclosure, the unit count and the site limit together.
Every rule is checked on every call and the verdict lists each finding.
Discount above the distributor limitOnly if someone is vigilant
Derate and discount authority are separate approvals. A rule has to raise each one, so a person sees both.
Every rule is checked on every call and the verdict lists each finding.

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 request; approval findings need a person to sign off first.

14200 kW load exceeds site capacity 3600.0 kW (2 effective units x 1800.0 kW after 0 percent derate at Standby rating).
2Non-emergency operation (demand response or peak shaving) cannot use the Standby rating; size on Prime or Continuous.
3EPA rules require Tier 4 Final for non-emergency stationary operation.
4A Tier 4 Final aftertreatment package is not offered for the Halvern HX2800.
5More than one genset on a bus needs paralleling controls.
64 units with Weather enclosure produce about 100 dBA at 7 m, above the 95 dBA site limit.
7Fuel tank gives 6 h at full load; the site requires 24 h.
8Lead time 28 weeks exceeds the 27 weeks available.
910 percent exceeds the 9 percent limit for 1 units.
10The quote total the agent states ($1952000.0) differs from the total the validator computes ($2047500.0).
1112 percent is above the 4 percent distributor limit; approvable up to 14 percent.
12Site derate of 14 percent (altitude and ambient) needs Halvern application engineering review.

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. Site capacity: Three generators that cannot carry the load

The agent's request. I'm sending Brightfield Hosting a quote for 3 HX1800 generator sets with one in reserve, to carry a 4200 kW load. Paralleling is not selected. The quote total is 1952000.

Verdict: Blocked.

  • Site capacity below the critical load. 4200 kW load exceeds site capacity 3600.0 kW (2 effective units x 1800.0 kW after 0 percent derate at Standby rating).
  • Paralleling controls missing. More than one genset on a bus needs paralleling controls.
  • Stated total differs from the validator total. The quote total the agent states ($1952000.0) differs from the total the validator computes ($2047500.0).

2. Site capacity: Larger units with paralleling selected

The agent's request. I'm sending Brightfield Hosting a revised quote for 3 HX2300 generator sets with one in reserve, for the 4200 kW load, with paralleling controls selected. The quote total is 2647500.

Verdict: Allowed.

3. Non-emergency use: A generator that also runs during peak demand events

The agent's request. I'm sending Northvale Cold Storage a quote for 1 HX1400 generator set on the Standby rating with Tier 2 emissions, for a 1200 kW load that also runs during peak-demand events. The quote total is 535500.

Verdict: Blocked.

  • Standby rating used for non-emergency operation. Non-emergency operation (demand response or peak shaving) cannot use the Standby rating; size on Prime or Continuous.
  • Tier 4 Final required for non-emergency operation. EPA rules require Tier 4 Final for non-emergency stationary operation.

4. Non-emergency use: Prime rating with Tier 4 Final

The agent's request. I'm sending Northvale Cold Storage the same quote on the Prime rating with Tier 4 Final emissions. The quote total is 705500.

Verdict: Allowed.

5. Site conditions: A quiet site and a long outage

The agent's request. I'm sending Ridgeline Medical a quote for 4 HX1400 generator sets with weather enclosures and 6 hour fuel tanks, one in reserve, for a 4000 kW load. The site has a 95 dBA sound limit and needs 24 hours of runtime. The quote total is 2242000.

Verdict: Blocked.

  • Combined sound level above the site limit. 4 units with Weather enclosure produce about 100 dBA at 7 m, above the 95 dBA site limit.
  • Fuel tank runtime too short. Fuel tank gives 6 h at full load; the site requires 24 h.

6. Site conditions: Quiet enclosures, 24 hour tanks, a 12 percent discount

The agent's request. I'm sending Ridgeline Medical the same quote with Quiet enclosures and 24 hour tanks, and a 12 percent discount. The quote total is 2322960.

Verdict: Needs approval.

  • Discount above the distributor limit. 12 percent is above the 4 percent distributor limit; approvable up to 14 percent.

7. Approval authority: A hot site with a discount

The agent's request. I'm sending Marlowe Water District a quote for 2 HX1800 generator sets with paralleling and 24 hour tanks, for a 3000 kW load at a site that reaches 52 C. The discount is 8 percent and the quote total is 1360520.

Verdict: Needs approval.

  • Discount above the distributor limit. 8 percent is above the 4 percent distributor limit; approvable up to 9 percent.
  • Site derate needs an engineering review. Site derate of 14 percent (altitude and ambient) needs Halvern application engineering review.

Narration transcript

Click to expand

The text of the narrated walkthrough, step by step.

  1. Welcome to a demo of Predictika's Validator. Halvern Power Systems, a fictional maker of standby generator sets, lets an AI quoting agent assemble quotes for sites that need backup power. Before the agent sends one, the call goes to Predictika’s Power Generation Quote Check. It sizes the site from the rating, the altitude and temperature derate and the units left in service, applies the use and emissions rules, checks paralleling, sound, fuel runtime and lead time, applies the discount authority, prices the quote, and checks that the total the agent stated matches the computed one. Then it returns one verdict: allow, needs approval, or block.
  2. I will start with the first request by the agent.
  3. The agent says. I'm sending Brightfield Hosting a quote for 3 HX1800 generator sets with one in reserve, to carry a 4200 kW load. Paralleling is not selected. The rest of the request is on screen.
  4. Before it goes out, the agent calls the validator.
  5. Blocked, with three findings, all in one answer. 4200 kW load exceeds site capacity 3600.0 kW (2 effective units x 1800.0 kW after 0 percent derate at Standby rating). More than one genset on a bus needs paralleling controls. The quote total the agent states ($1952000.0) differs from the total the validator computes ($2047500.0).
  6. The validator returned three 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 request are checked against each other.
  7. What this shows: for site capacity below the critical load, a tool can do this only with workarounds.
  8. Now the agent resubmits.
  9. The corrected request. It changes the model and units to 3 × HX2300 (2300 kW standby), the paralleling controls to selected and the stated quote total to 2647500 USD.
  10. It calls the validator again.
  11. This time the validator clears it.
  12. All three earlier findings are cleared.
  13. Three problems caught in one call, fixed in one pass, and a clean request. That is the value of Predictika's validator: catching mistakes before the action runs.