Predictika · Cyber insurance placement

An AI sales agent's cyber insurance placement, checked before it goes out

An AI sales agent prepares a cyber insurance placement for a client: the program, the revenue, the primary and excess limits, the retention, the controls in place, recent ransomware and the total limit it states in millions. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the placement rules and returns a verdict, every reason and the figures it computed.

Scope: Cyber placements through three programs: Cyber Facility, Cyber Pathway and a bespoke placement with Cyber Catalyst terms, with limits and layers, retention minimums, required controls, approvals and the stated total limit. See everything this validator covers

This validator models cyber placements of this kind, not just the scenarios shown below.

  • Programs and eligibility — the revenue each program serves, its largest total limit, the completed self-assessment, and the controls and claims history it requires.
  • Layers and retention — the size of the primary layer, an excess layer that needs a large enough primary, and the minimum retention for the revenue band.
  • Approvals and the stated total — recent ransomware, missing incident response plans, regulated data at volume and very large limits that need a person, and the stated total limit compared with the layers.

Any placement that touches these rules is in scope — not only the scenarios pictured here. The program names and the purpose of each program follow public Marsh information; every limit, revenue band, retention, control requirement and approval threshold is modeled from comparable cyber programs and is not published by Marsh. No premiums are modeled. Not covered: other lines of business and other Marsh McLennan businesses.

The AI sales agentBuilds the cyber placement 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
Limits that depend on the programOnly with care
Each program has its own largest revenue and its own largest total limit, so every program needs its own rules.		The program, the revenue and the layers are fields of one placement and are checked together.
Controls and historyOnly with workarounds
Multi-factor authentication, offline backups, a completed self-assessment and recent ransomware matter differently in each program.		Every condition is evaluated on every call, whatever the program.
Layers and retentionOnly with care
An excess layer needs a large enough primary, and the minimum retention rises with revenue.		The primary, the excess and the retention are compared with every level on each call.
A total built from layersOnly with workarounds
The total limit is the primary plus the excess, so it must be computed before it can be compared with the total stated.		The total is computed on every call and compared with the total stated in the placement.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as a step is saved or 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.

    
1Cyber Facility serves small and medium-sized companies; revenue of 251 million is above the 250 million modeled limit.
2Total limit of 11 million is above the 10 million modeled for Cyber Facility.
3Cyber Facility needs multi-factor authentication and offline backups in place.
4A ransomware incident in the last 24 months makes the company ineligible for Cyber Facility; consider Cyber Pathway.
5Cyber Facility and Cyber Pathway start from a completed Marsh Cybersecurity Self-Assessment.
6Cyber Pathway offers limited coverage; total limit of 4 million is above the 3 million modeled.
7A bespoke placement is modeled for revenue of at least 25 million; the quote has 24 million.
8Excess limit needs a primary layer of at least 5 million; the quote has 4 million.
9The primary limit is 26 million; the maximum modeled for one primary layer is 25 million.
10The retention is 100 thousand; the minimum modeled for this revenue is 250 thousand.
11The retention is 24 thousand; the minimum modeled for this revenue is 25 thousand.
12Cyber Catalyst terms need at least one Cyber Catalyst designated solution deployed.
13Enhanced Cyber Catalyst terms apply to individually negotiated policies, not to this program.
14A bespoke limit above 10 million without multi-factor authentication and EDR needs underwriter review.
15No incident response plan is in place; a broker should confirm the insurer will quote.
16The company holds 5 million records of regulated data; a broker should confirm underwriting appetite.
17A total limit of 60 million needs a senior broker to confirm market capacity.
18The quote states a total limit of 25 million; the layers add up to 30 million.
19The primary limit must be at least 1 million.
20A ransomware incident in the last 24 months needs a remediation summary for the insurers.

  

  

    
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. Program, layers and history: A Cyber Facility placement for a 300 million company with a thin primary, no MFA and a recent ransomware incident

    
The agent's request. Please place Cyber Facility cover for a company with 300 million of revenue: a 3 million primary and a 12 million excess layer, with a retention of 100 thousand. Multi-factor authentication is not on every account. The company had a ransomware incident last year. The placement states a total limit of 12 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Excess layer without a large enough primary. Excess limit needs a primary layer of at least 5 million; the quote has 3 million.
    • 
      
  • Retention below the minimum. The retention is 100 thousand; the minimum modeled for this revenue is 250 thousand.
    • 
      
  • Revenue too large for Cyber Facility. Cyber Facility serves small and medium-sized companies; revenue of 300 million is above the 250 million modeled limit.
    • 
      
  • Limit above the Cyber Facility maximum. Total limit of 15 million is above the 10 million modeled for Cyber Facility.
    • 
      
  • Cyber Facility needs MFA and offline backups. Cyber Facility needs multi-factor authentication and offline backups in place.
    • 
      
  • Recent ransomware not eligible for Cyber Facility. A ransomware incident in the last 24 months makes the company ineligible for Cyber Facility; consider Cyber Pathway.
    • 
      
  • Stated total limit does not match. The quote states a total limit of 12 million; the layers add up to 15 million.
    • 
    

    
2. Program, layers and history: The agent resubmits as a bespoke placement with the senior broker approval

    
The agent's request. Please place bespoke cover for a company with 300 million of revenue: a 10 million primary and a 5 million excess layer, with a retention of 250 thousand. The company had a ransomware incident last year. A senior broker has approved it. The placement states a total limit of 15 million. It is ready to send.

    
Verdict: Needs approval.

    
    
      
  • Recent ransomware on a bespoke placement. A ransomware incident in the last 24 months needs a remediation summary for the insurers.
    • 
      
  • Placement summary. 15 million total limit, 250 thousand retention, 300 million revenue.
    • 
    

    
3. Pathway and Cyber Catalyst: A Cyber Pathway placement with a 5 million limit, no self-assessment and Cyber Catalyst terms

    
The agent's request. Please place Cyber Pathway cover for a company with 100 million of revenue: a 5 million primary, with a retention of 100 thousand. The self-assessment has not been completed. Please apply Cyber Catalyst terms. The placement states a total limit of 5 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Cybersecurity Self-Assessment not completed. Cyber Facility and Cyber Pathway start from a completed Marsh Cybersecurity Self-Assessment.
    • 
      
  • Limit above the Cyber Pathway maximum. Cyber Pathway offers limited coverage; total limit of 5 million is above the 3 million modeled.
    • 
      
  • Catalyst credit without a designated solution. Cyber Catalyst terms need at least one Cyber Catalyst designated solution deployed.
    • 
      
  • Catalyst terms only on a bespoke placement. Enhanced Cyber Catalyst terms apply to individually negotiated policies, not to this program.
    • 
    

    
4. Pathway and Cyber Catalyst: The agent resubmits with a 3 million limit and the completed self-assessment

    
The agent's request. Please place Cyber Pathway cover for a company with 100 million of revenue: a 3 million primary, with a retention of 100 thousand. The placement states a total limit of 3 million. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Placement summary. 3 million total limit, 100 thousand retention, 100 million revenue.
    • 
    

    
5. Bespoke placement size: A bespoke placement for a 20 million company with a small primary under a large excess layer

    
The agent's request. Please place bespoke cover for a company with 20 million of revenue: a 4 million primary and a 20 million excess layer, with a retention of 25 thousand. The placement states a total limit of 25 million. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Excess layer without a large enough primary. Excess limit needs a primary layer of at least 5 million; the quote has 4 million.
    • 
      
  • Revenue too small for a bespoke placement. A bespoke placement is modeled for revenue of at least 25 million; the quote has 20 million.
    • 
      
  • Stated total limit does not match. The quote states a total limit of 25 million; the layers add up to 24 million.
    • 
    

    
6. Bespoke placement size: The agent resubmits for a 40 million company with matching layers

    
The agent's request. Please place bespoke cover for a company with 40 million of revenue: a 10 million primary and a 10 million excess layer, with a retention of 25 thousand. The placement states a total limit of 20 million. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Placement summary. 20 million total limit, 25 thousand retention, 40 million revenue.
    • 
    

  

  

    
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 cyber insurance placement for a client: the program, the revenue, the primary and excess limits, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the placement rules 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 place Cyber Facility cover for a company with 300 million of revenue: a 3 million primary and a 12 million excess layer, with a retention of 100 thousand. Multi-factor authentication is not on every account. The rest of the request is on screen.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with seven findings, all in one answer. Excess limit needs a primary layer of at least 5 million; the quote has 3 million. The retention is 100 thousand; the minimum modeled for this revenue is 250 thousand. Cyber Facility serves small and medium-sized companies; revenue of 300 million is above the 250 million modeled limit. The other four are on screen.
    6. 
      
  6. The validator returned seven 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 quote are checked against each other and against the stated total.
    7. 
      
  7. What this shows: for layers and retention, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the program to a bespoke placement, the primary limit to 10 millions, the excess limit to 5 millions, the retention to 250 thousands and the stated total limit to 15 millions. Two more changes are on screen.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. The required approval is recorded, so the validator accepts the release.
    12. 
      
  12. The required approval is recorded, so the validator accepts the quote.
    13. 
      
  13. Seven problems caught in one call, fixed in one pass, and the corrected quote now goes to a person for approval. That is the value of Predictika's validator: catching mistakes before they reach the customer. Now you can try the other scenarios yourself.
    14.