Predictika · AI platform orders

An AI sales agent's AI Refinery order, checked before it goes out

An AI sales agent prepares an order for Accenture AI Refinery: the platform components (Agents, Trusted Agent Huddle, Knowledge, Models, Governance), an AI Refinery for Industries solution, the Physical AI SDK or the Sovereignty offering, a deployment option and an edition. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of Accenture's public descriptions plus modeled rules, and returns a verdict, every reason and the figures it computed.

Scope: Accenture AI Refinery orders: components, industry agent solutions, Physical AI SDK, sovereignty, deployment, edition limits and support. See everything this validator covers

This validator models AI Refinery orders, not just the scenarios shown below.

  • Components — Agents and Agent Builder, Trusted Agent Huddle, Knowledge, Models (customization) and Governance, with the dependencies between them.
  • Industry solutions — the six AI Refinery for Industries solutions Accenture lists: B2B Marketing, Hyper-personalization for Beauty, Revenue Growth Management, Intelligent Sales Coach, Clinical Trial Companion and Asset Troubleshooting.
  • Physical AI SDK and Sovereignty — where each can be deployed.
  • Deployment — public cloud, private cloud, on-premises or sovereign infrastructure.
  • Edition, limits and support — agents, users and data sources per edition, a stated agent limit, and approvals for Strategic orders and many data sources.

Any AI Refinery order of this shape is in scope — not only the scenarios pictured here. Not covered: prices (this validator produces no price build-up), models and model providers, and Accenture services engagements. Source status: component names, the six industry solutions, the Physical AI SDK, the Sovereignty offering and the four deployment options are taken from Accenture's public pages. The Pilot, Enterprise and Strategic editions and their limits, the dependencies between components, the on-premises support requirement and the approval thresholds are modeled from comparable enterprise AI platforms and are not Accenture's official rules.

The AI sales agentBuilds the AI Refinery order from the customer's request 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
Components that depend on each otherOnly with care
Dependencies such as Huddle needing Governance and industry solutions needing Knowledge, kept as separate rules.		All dependencies are checked together on every call.
Where an offering may be deployedOnly with care
A matrix of offerings against four deployment options to maintain by hand.		Each offering is checked against the chosen deployment on every call.
Edition limits for agents, users and sourcesOnly with workarounds
Three limit tables per edition, each with its own rule.		The limits of the chosen edition are looked up and returned with the verdict.
A stated limit that disagreesOnly with workarounds
The figure the agent states is a separate field and needs an extra comparison rule.		The stated limit is part of the request and is compared on every call.
Orders that need a personOnly with workarounds
Pass or fail rules need special approval rules for Strategic orders and large integrations.		They go to a person as needs approval, neither passed nor blocked.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the order 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.

    
1AI Refinery for Sovereignty keeps data and processing in a sovereign cloud; the order deploys elsewhere.
2The Physical AI SDK is ordered on public cloud; this order models it on private cloud, on-premises or sovereign infrastructure only.
3Trusted Agent Huddle needs the Governance component on the same order.
4An AI Refinery for Industries solution needs the Knowledge component to hold its data.
56 agents is over the 5 the edition takes.
626 users is over the 25 the edition takes.
74 data sources is over the 3 the edition takes.
8The Pilot edition is modeled for public cloud only.
9Model customization is not part of the Pilot edition.
10AI Refinery for Industries solutions are not part of the Pilot edition.
11101 agents is over the 100 the edition takes.
125,001 users is over the 5,000 the edition takes.
1326 data sources need an integration review by a person.
1450 data sources need an integration review by a person.
1551 data sources is over the 50 the edition takes.
1651 data sources need an integration review by a person.
17An on-premises deployment is modeled with Premier support only.
18A Strategic edition order is scoped with a person before it is released.
191,001 agents is over the 1,000 the edition takes.
20100 data sources need an integration review by a person.
21The order states a limit of 5 agents; this edition is modeled at 100.
22The order states a limit of 100 agents; this edition is modeled at 1,000.
23200 agents is over the 100 the edition takes.
246,000 users is over the 5,000 the edition takes.

  

  

    
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. Components: A life sciences order with the Clinical Trial Companion and Trusted Agent Huddle but no Knowledge or Governance

    
The agent's request. Please quote an AI Refinery Enterprise edition on public cloud with 30 agents, 800 users and 8 data sources, on standard support, including Trusted Agent Huddle and the Clinical Trial Companion industry solution. The order states a limit of 100 agents. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Trusted Agent Huddle without Governance. Trusted Agent Huddle needs the Governance component on the same order.
    • 
      
  • Industry solution without Knowledge. An AI Refinery for Industries solution needs the Knowledge component to hold its data.
    • 
    

    
2. Components: The agent resubmits after fixing the errors

    
The agent's request. Please quote an AI Refinery Enterprise edition on public cloud with 30 agents, 800 users and 8 data sources, on standard support, including Knowledge, Governance, Trusted Agent Huddle and the Clinical Trial Companion industry solution. The order states a limit of 100 agents. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. The edition takes up to 100 agents, 5,000 users and 50 data sources; this order has 30 agents, 800 users and 8 data sources.
    • 
    

    
3. Deployment: A manufacturer order for the Physical AI SDK on public cloud

    
The agent's request. Please quote an AI Refinery Enterprise edition on public cloud with 40 agents, 1200 users and 10 data sources, on standard support, including Knowledge, Governance and the Physical AI SDK. The order states a limit of 100 agents. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Physical AI SDK on public cloud. The Physical AI SDK is ordered on public cloud; this order models it on private cloud, on-premises or sovereign infrastructure only.
    • 
    

    
4. Deployment: The agent resubmits after fixing the errors

    
The agent's request. Please quote an AI Refinery Enterprise edition on private cloud with 40 agents, 1200 users and 10 data sources, on standard support, including Knowledge, Governance and the Physical AI SDK. The order states a limit of 100 agents. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. The edition takes up to 100 agents, 5,000 users and 50 data sources; this order has 40 agents, 1,200 users and 10 data sources.
    • 
    

    
5. Pilot: A pilot order for 8 agents and 40 users with model customization

    
The agent's request. Please quote an AI Refinery Pilot edition on public cloud with 8 agents, 40 users and 3 data sources, on standard support, including Knowledge, Governance and model customization. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Model customization in a Pilot. Model customization is not part of the Pilot edition.
    • 
      
  • Too many agents for the edition. 8 agents is over the 5 the edition takes.
    • 
      
  • Too many users for the edition. 40 users is over the 25 the edition takes.
    • 
    

    
6. Pilot: The agent resubmits after fixing the errors

    
The agent's request. Please quote an AI Refinery Pilot edition on public cloud with 5 agents, 25 users and 3 data sources, on standard support, including Knowledge and Governance. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Order summary. The edition takes up to 5 agents, 25 users and 3 data sources; this order has 5 agents, 25 users and 3 data sources.
    • 
    

  

  

    
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 order for Accenture AI Refinery: the platform components (Agents, Trusted Agent Huddle, Knowledge, Models, Governance), an AI Refinery for Industries solution, the Physical AI SDK or the Sovereignty offering, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of Accenture's public descriptions plus modeled 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 quote an AI Refinery Pilot edition on public cloud with 8 agents, 40 users and 3 data sources, on standard support, including Knowledge, Governance and model customization. It is ready to send.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with three findings, all in one answer. Model customization is not part of the Pilot edition. 8 agents is over the 5 the edition takes. 40 users is over the 25 the edition takes.
    6. 
      
  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 quote are checked against each other.
    7. 
      
  7. What this shows: for edition limits for agents, users and sources, a rules-based CPQ can do this only with workarounds.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the agents to 5 and the users to 25. It removes the model customization.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All three earlier findings are cleared.
    13. 
      
  13. Three 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.