InXAP / GR / Overview
Promethix Cockpit.
Promethix Cockpit is InXAP’s agentic AI platform. It connects SAP, data platforms such as Databricks and your documents, runs specialist agents as one governed team, and turns their work into evidence-linked decisions that a person approves before anything is written back.
Illustrative views · Product demo on requestLet’s talkPROMETHIX COCKPIT
Where SAP context meets AI agent teams.
Promethix Cockpit is InXAP’s agentic AI platform. It connects SAP, data platforms such as Databricks and your documents, runs specialist agents as one governed team, and turns their work into evidence-linked decisions that a person approves before anything is written back.
Invoice exception warrants a human decision
The evidence supports an investigation, not an automatic booking. Context from SAP, documents and policy is attached; nothing executes before approval.
7 agents · 3 strategies · 11 evidence gaps
Follow an invoice exception from enterprise context to a human approval gate. An illustrative workflow, ready for discussion.
What is it? Workflows
A workflow is the path a business task takes from the first signal to a finished action: the trigger, the context, the plan, the approval and the execution. In most companies this path is spread across inboxes, spreadsheets and SAP screens.
How we solve it
- We model the process once, from signal to action, with a named owner for every step.
- Agents prepare the work; execution waits at an approval gate.
- Every run is recorded, so it can be repeated, measured and improved.
Explore how specialized agents could collaborate around a shared task, with a reviewer deciding what happens next.
What is it? Agent teams
An agent team is a group of specialist AI agents that share one objective. One agent gathers evidence, another analyses it against policy, another drafts the recommendation, much like a small project team.
How we solve it
- A coordinator splits the task and hands each part to the right specialist.
- Specialists run in parallel and return findings with their sources.
- A reviewer sees the combined result and decides what happens next.
Prepare a recommendation with references to the relevant system records and policies.
See how a proposed knowledge layer could bring SAP context, documents and business policies into the same evidence trail.
What is it? Knowledge
Knowledge is the business context an agent needs to be right: SAP master and transactional data, documents, policies and approved external sources. Without it, an AI answer is only a guess.
How we solve it
- We connect SAP, document stores and policy sources through governed connectors.
- Retrieval (RAG) returns only what the user is allowed to see, with a link to the source.
- Gaps are shown openly, so a missing fact is never filled in silently.
- Context agent started
- Analysis agent returned a response
- Recommendation drafted
- Execution paused: approval required
Controls are part of the workflow architecture, from context to decision.
Explore proposed controls for identity, tool access, human decisions and traceability throughout a workflow.
What is it? Governance
Governance is the set of rules that decide who may use an agent, which systems it can touch, what needs a human decision and what must be recorded. It is what makes AI acceptable in finance, supply chain and regulated work.
How we solve it
- Identity and role-based access define who can run which workflow.
- Tools are scoped per agent, with approvals before anything is written to a system.
- A complete audit trail keeps context, decisions and costs traceable.
- 01Hybrid agent teams with Databricks Genie
- 02SAP context through governed connectors
- 03Knowledge Hub with cited answers
- 04Human approval before every action
- 05Run timeline, cost and audit trail
- 06ABAP coding agents in VS Code
- 07Scheduled market and risk monitors
- 08Visual decision briefs
Portfolio overview. Product ownership, functionality, integrations, deployment model and availability require review. These concepts are not released products or live demos.
Enterprise agents. Real business questions.PROMETHIX COCKPIT · PLATFORM
Everything an agent team needs to run in production.
One control plane for agents, tools, knowledge, approvals and cost, instead of AI point tools scattered across the business.
- 1
Start from an industry team
Begin with teams for finance, supply chain, procurement, risk and SAP delivery, or compose your own specialists.
- 2
Connect your systems
Attach SAP, data warehouses, documents and MCP servers under policy, role-based access and tool-approval gates.
- 3
Review every output
Each output carries evidence, reviewer state and an audit trail your operators can verify before anything executes.
See the cockpit in use
Every page of the Promethix Cockpit, from the briefing to the war room. Browse them below.
Browse the pages:
Your business is talking.
Your intelligence team has something to show you.
4 signals requiring attention · Correlated margin risk and offset opportunities
Benchmark movement warrants exposure reconciliation
The evidence supports an investigation, not a quantified margin exposure. External benchmarks show material movement, but they are not supplier prices or realized savings.
7 agents · 3 strategies · 11 evidence gaps

Start a new investigation
Production agent teams
Run single agents or coordinated teams with session memory, human approvals, task handoffs and full run traces.
Knowledge Hub
Hybrid retrieval over your documents with access control, citations, grounding checks and links to the exact source page.
Governed tools, MCP, SQL & SAP
Turn MCP servers, SQL systems, OpenAPI endpoints and SAP ADT into permissioned, observable agent capabilities.
Skills & output contracts
Author reusable skills and require structured outputs that can be rendered, reviewed and exported.
Governance & observability
Org policies, tool-approval gates, redaction, audit trails and cost tracking on every agent, tool and answer.
Interactive outputs
Agents return dashboards, timelines, scorecards, evidence views and decision briefs, not just chat text.
A2A & external agents
Combine native agents with provider-hosted agents, such as Databricks Genie, in one governed team.
Where people already work
Operate teams from the web, Microsoft Teams, Slack, WhatsApp and Telegram, and coding agents inside VS Code.
Memory & connectors
Long-term agent memory plus connectors for SharePoint, OneDrive, Blob storage, Confluence, GitHub and the web.
- Full audit trail
- Human-in-the-loop approvals
- Role-based agent boundaries
PROMETHIX AI PLATFORM
AI for your SAP business and IT teams
Promethix connects AI agents to SAP and enterprise data. It helps business teams make decisions and SAP teams configure, test and develop faster.
- BUSINESS OPERATIONS
Commodity & Margin Intelligence
See how commodity price shifts and supply risks affect costs, margins and customer commitments.
- FINANCE & PLANNING
Planning & Variance Analysis
Explain gaps between budget, forecast and actuals across markets, including currency, volume and product mix.
- ENTERPRISE DATA
SAP & Data Integration
Connect SAP and platforms such as Databricks so teams can ask questions and see the sources behind answers.
- SAP CONFIGURATION
AI-Assisted SAP Customizing
Use AI agents to help configure SAP processes across modules.
- SAP QUALITY ASSURANCE
Automated SAP Testing
Run SAP test scenarios autonomously and surface issues for review.
- SAP DEVELOPMENT
AI-Assisted SAP Development
Use AI agents to build SAP developments and extensions with clean-core principles.
Teams review and approve critical actions.
Request a demoPROMETHIX IN ACTION
Agent teams for the decisions that run on SAP.
Four Promethix workspaces for food and manufacturing enterprises. Each one is a governed AI agent team working on an SAP-shaped business question, from cocoa exposure to ABAP delivery.
Commodity & margin war room
When cocoa moves or a supply hazard appears, four Promethix specialists and four Databricks Genie analysts work out which products, promotions and customer commitments are exposed, then put a decision brief in front of its owner.
- Detect
- Exposure
- Options
- Decision brief
- Human review
- Market & risk scout
- Procurement & supply planner
- Commercial response analyst
- Commercial decision lead
- + 4 Genie analysts
OutputExecutive decision brief with options, owners, evidence and a validation plan.
- SAP finance & supply
- Databricks Genie
- Market & hazard feeds
- ECB FX
Seasonal planning & FX variance
Budget, forecast and actuals across twelve countries, with Easter and Christmas sell-in timing, FX effects and recipe cost sensitivity. The team explains why a market drifts from budget and what to do about it.
- Variance
- Country & product
- FX vs. volume vs. mix
- Scenario
- Decision
- Market & consumer scout
- Cocoa & seasonal supply planner
- Premium growth analyst
- Commercial decision lead
OutputA country-level variance explanation and a reviewed forecast adjustment.
- SAP CO-PA & BOM
- Databricks Genie
- FX rates
- AI/BI dashboards
SAP coding agents in VS Code
Twenty-one coding specialists for SAP, applications, cloud and review take one SAP change from requirement to developer hand-off inside VS Code, through SAP ADT, with every SAP action approved.
- Requirement
- Design
- ABAP change
- Tests
- Developer review
- Functional analysis
- ABAP development
- Tests & checks
- Review & hand-off
OutputA reviewed, tested SAP change handed to the developer, with evidence for every step.
- SAP ADT
- ABAP
- VS Code
- Approval gates
Contract profitability & traceability
Industrial customers, contracts and recipes measured in tonnes. Five Genie analysts check price pass-through, contract margin, fill rate and cocoa-lot evidence, and flag where a contract needs attention.
- Contract signal
- Margin & pass-through
- Service & quality
- Origin evidence
- Renegotiation brief
- Finance & FX
- Cocoa & commodity
- BOM & operations
- Customers & service
- Sustainability & traceability
OutputContract-level margin view with flagged fill-rate, inventory and origin-evidence gaps.
- SAP SD & MM
- Recipes in tonnes
- Databricks Genie
- Lot traceability
Company names are withheld. Demonstration data combines public reporting with synthetic operational data; nothing shown here is a customer result.
Start with the right questions.
