Ecoservity · STREAM-Intelligence Platform

STREAM Agents: Agentic Execution on Live Supply Chain Operations

AI agents that monitor live SAP events, surface exceptions, and upon approval act. No warehouse copies. No batch review cycles. P2P Agent autonomous in production since November 2023.

Watch: STREAM Agents in action

See agents run on live SAP data, not batch copies

Three short videos: the batch-AI clog, the P2P Agent in autonomous production, and natural-language SAP execution via SAP Chat.

Overview

Why batch-AI cannot run supply chain operations

Watch on YouTube ↗
Live Demo

P2P Agent: autonomous exception handling in production

Watch on YouTube ↗
Deep Dive

SAP Chat: natural language SAP execution, no GUI

Watch on YouTube ↗

The gap between operational event and agent action, today.

Supply chains generate events every second. Batch-AI reacts in hours or days. STREAM Agents close the gap by running on live SAP, not warehouse extracts. These are the numbers that define the difference.

0
Data Copies
Agents read live SAP via zero-copy stream. No warehouse. No duplicate infrastructure. No stale copy governance.
24hr+
Batch Lag Eliminated
Warehouse pipelines land data 24 to 48 hours after the event. Agents on live streams act within seconds.
850K+
SAP Tables Live via MCP
Live access to SAP semantic models and tables. Discoverable via MCP Catalog. Ready for any agent that can call an API.
Nov 2023
Autonomous in Production
P2P Agent has resolved procure-to-pay exceptions autonomously since November 2023. Not a pilot. Not a demo.

Traditional AI agents run on warehouse copies of enterprise data. Warehouse copies lag the operational event by 24 hours or more. Agents that cannot see the event cannot act on the event. STREAM Agents run on live SAP via zero-copy streaming, so the sense-to-act loop closes in seconds, not batch windows.

The Problem

Four operational needs. All blocked by the same batch-AI clog.

The same root cause, latency between event and AI, shows up as four distinct failure modes across supply chain operations. Fix the latency, remove the failures.

Value Driver What Operations Needs What Batch-AI Delivers
01 · Exception Handling Detect and route exceptions as they happen. Route to the right owner, with full SAP context, and close the loop within the operational window. Weekly batch review. Exceptions rot in queues. Root cause lost by the time the exception is triaged. Compounding downstream.
02 · Decision Speed Agent acts within seconds of the operational event: inventory allocation, order prioritization, exception routing, transaction posting. Data warehoused overnight. Reports built in the morning. Decisions made after the event window has closed. The window is missed.
03 · Human Oversight Every agent action is reviewable, approvable, and auditable. Native SAP roles, native SAP audit trail, native SAP correction. Agent runs on a copy outside the SAP security model. Approvals in a separate tool. Audit trail split across systems. Correction after the fact is manual.
04 · Scale Economics Hundreds of agents on live data, one shared streaming layer. Cost scales with agents, not with copies of the same data. Each agent needs its own pipeline. Each pipeline duplicates SAP data, storage, and governance. Cost scales with copies, not value.
Full Briefing Deck

See the complete STREAM Agents story

The batch-AI clog, the sense-surface-approve-act loop, four executive personas, and platform architecture, in full. View below or download the deck (PDF) ↓

↓ Download briefing deck (PDF) SAP NetWeaver Certified · Microsoft Certified · USPTO Patented · P2P Agent autonomous since Nov 2023

Want to see STREAM Agents working on live data? Try the MCP Catalog, the SAP Chat sandbox, and the AI Insights API.

Try Live Agents See All Products ↗
► Key slides: full text (expand for searchable content)

Slide 1: The Batch-AI Clog

The gap between event and action is what defines supply chain performance in the AI era.

0 data copies for STREAM Agents: they read live SAP via zero-copy streaming. No warehouse. No duplicate infrastructure. 24hr+ batch lag eliminated: warehouse pipelines land data 24 to 48 hours after the event, so agents on those copies act after the window has closed. 850,000+ SAP tables live-accessible via MCP Catalog for any agent that can call an API. P2P Agent autonomous in production since November 2023: two years of continuous operation on live enterprise data. Not a demo. Not a pilot.

Slide 2: Four Operational Needs Blocked

Four operational needs, all blocked by the same batch-AI clog.

Exception Handling: Weekly batch review. Exceptions rot in queues. What's needed: detect and route as it happens, with full SAP context, closed within the operational window.

Decision Speed: Data warehoused overnight, reports built in the morning, decisions made after the event window has closed. What's needed: agent acts within seconds of the operational event.

Human Oversight: Agent runs on a copy outside the SAP security model. Approvals in a separate tool. Audit split across systems. What's needed: every action reviewable, approvable, and auditable in the native SAP model.

Scale Economics: Each agent needs its own pipeline. Each pipeline duplicates SAP data, storage, and governance. What's needed: cost that scales with agents, not with copies.

Slide 3: Five Points in the Operational Cycle

Five points across the operational cycle where agents intervene, each on live data, not batch copies.

  • Point 1: Demand Signal to Plan. Agent monitors order intake, forecast deviation, and inventory position live. Surfaces plan-breaking events within the fulfillment window, not the next planning cycle.
  • Point 2: Plan to Procurement. Agent watches for supplier delays, PO exceptions, and price variances as they happen. Routes to the right buyer with full SAP context, not batch report inbox.
  • Point 3: Procurement to Production. Agent tracks goods receipt, quality holds, and production variance live. Reallocates inventory and reprioritizes orders before the shortage propagates.
  • Point 4: Production to Fulfillment. Agent monitors throughput, shipment status, and customer commitments live. Flags at-risk orders and proposes recovery before the SLA is missed.
  • Point 5: Fulfillment to Cash. Agent watches invoice cycle, dispute events, and payment status. Surfaces cash-blocking exceptions with full order-to-cash context, closed the same day.

Common thread: Every intervention is live, in the SAP security model, and reviewable in native SAP audit trail. Fix at the event, fix everywhere.

Slide 4: The Solution Loop

Sense, surface, approve, act. Every step on live SAP.

Batch-AI

Nightly warehouse extract · exceptions in report inbox · approval outside SAP · execution manual by operator

STREAM Agents

Live SAP via zero-copy stream · prioritized exception with context · native SAP roles for approval · SAP-certified execution

Sense: Agents monitor live SAP events via zero-copy streaming. No warehouse extract, no batch pipeline. Every state change is visible as it happens.

Surface: Exceptions arrive with full operational context: linked documents, upstream root cause, downstream impact. Not a row in a report.

Approve: Human-in-the-loop wherever business policy requires it. Native SAP application authorizations. No shadow approval tool.

Act: Agents execute via SAP-certified interfaces. Every action lands in the SAP audit trail. Correctable in the SAP correction model.

Slide 5: Meet the Agents

Ready-to-deploy agents, streaming live SAP data from day one.

P2P Agent (Supply Chain): Autonomous procure-to-pay exception handling. Monitors live operational events, surfaces exceptions, and resolves them without human batch review cycles. In production since November 2023.

SAP Chat (UI and API): Natural language SAP execution. Any TCODE or report accessed through conversation. Operators act without navigating SAP GUI. Programmatic API for embedding SAP execution into any application.

AI Execution (Supply Chain): Agents executing supply chain decisions on live data: inventory allocation, order prioritization, exception routing. At the speed of the operational event.

AI Insights API: Combined SAP, trade, and sales intelligence through a single API. End-to-end supply chain AI on live data. The complete operational picture for agents and apps.

Slide 6: Executive Personas

Your clog by function, and the STREAM Agents entry point for each.

COO: Supply chain exceptions caught weekly, should be seconds. Replenishment signals 24 to 48 hours late. Safety stock held against uncertainty that could be removed. Entry point: P2P Agent and AI Execution. Autonomous since November 2023. Question: "How many supply chain exceptions did your team handle manually last month?"

CFO: Cashflow forecasts on last week's receivables. Working capital tied to exceptions that could close today. Entry point: P2P Agent for O2C and AR exception handling. Question: "How much working capital is tied to exceptions your team could close today with better information?"

CDO / CAIO: AI running on hours-old warehouse copies. Every agent needs its own pipeline. Governance duplicated per copy. Entry point: STREAM MCP and AI Insights API. Live SAP for AI agents via zero-copy streaming. Question: "How many pipelines does each of your AI agents need today, and how fresh is the data on the other side?"

CIO: Will agents destabilize SAP? Where does data go? Another integration project? Entry point: STREAM Activate. SAP NetWeaver certified. No ABAP. Clean-core aligned. Data never leaves your environment. 100% native SAP authorizations. Question: "What is your current cost and risk of feeding SAP data to AI agents?"

Slide 7: Platform Architecture

Enterprise-grade. Zero-copy. SAP-certified. Microsoft-certified. Patented.

STREAM Agents run on STREAM Engine: live access to 850,000+ SAP tables and 450,000+ business-ready semantic models. No pipeline. No extractors. No data copy. USB4SAP is SAP-certified via NetWeaver. No ABAP. Clean-core aligned. In production since 2021.

  • STREAM People@Work: Live SAP for operators in Excel, Power BI, and Fabric. The human side of the same live streaming layer the agents use.
  • STREAM Agents: P2P Agent (autonomous since Nov 2023), SAP Chat, AI Execution, AI Insights API. Agentic execution on live data.
  • STREAM Engine + USB4SAP: Zero-copy, SAP-certified, patented. Live access via MCP Catalog and REST for agents, apps, and Python workflows.
  • TradePro and SalesPro: Companion streaming intelligence for trade compliance and revenue signals. Same live foundation, different operational surface.

Certifications: SAP NetWeaver Certified · Microsoft Certified (Excel, Power BI, Fabric) · USPTO Patent · Azure Marketplace Published · P2P Agent autonomous since November 2023 · 90% reduction in SAP data access cost vs traditional replication (Microsoft co-innovation).

Value Chain · Five Points of Intervention

Where agents intervene, from demand signal to cash.

Five points across the operational cycle where a live-data agent can act before the exception compounds. Each on the same zero-copy streaming layer, each inside the SAP security model.

Stage Operational Event What the Agent Does What This Replaces
01 · Demand Signal to Plan Order intake, forecast deviation, inventory position. Surfaces plan-breaking events within the fulfillment window. Proposes replan with full context. Weekly planning cycle. Events discovered after the fact.
02 · Plan to Procurement Supplier delay, PO exception, price variance. Routes to the right buyer with linked docs, upstream cause, downstream impact. Recommends corrective PO. Batch report inbox. Buyer triages without SAP context.
03 · Procurement to Production Goods receipt exception, quality hold, production variance. Reallocates inventory and reprioritizes orders live. Escalates only where policy requires it. Manual reconciliation at end of shift. Shortage propagates.
04 · Production to Fulfillment Throughput signal, shipment status, customer commitment risk. Flags at-risk orders and proposes recovery: split shipment, alternate carrier, partial fulfillment. SLA miss discovered after customer complains.
05 · Fulfillment to Cash Invoice cycle, dispute event, payment status. Surfaces cash-blocking exceptions with full O2C context. Closes same-day disputes autonomously per policy. Month-end reconciliation. Working capital tied up in queues.

One streaming layer. Every intervention. The same USB4SAP zero-copy foundation that powers exception handling in P2P also powers order prioritization in fulfillment and dispute resolution in O2C. One live truth, many agents, no duplicate pipelines.

Why Change

Batch-AI was built for reporting. Supply chain is an execution problem.

Three structural failures, each compounding the others. Not an ops inconvenience: a scale and cost problem that gets worse with every additional agent.

# Structural Failure Why It Has Become Untenable
01 Warehouse-first AI cannot act on events it cannot see.
Traditional agents run on nightly extracts from SAP into a warehouse or lakehouse.		 By the time the extract lands, the event window has closed. Exception handling in daily batches is not exception handling: it is exception recording.
02 Per-agent pipelines duplicate SAP data, storage, and governance.
Each new agent needs its own extract, its own model, its own copy.		 Cost scales with copies, not agents. Governance, lineage, and access control are re-implemented for every copy. Each copy is a new attack surface.
03 Approval outside the SAP security model is not really approval.
Shadow approval tools disconnected from SAP roles and audit.		 What looks like an approved agent action is really an unlogged operator action. Auditors reject it. Correction after the fact is manual and error-prone.
STREAM Agents · The Solution

Sense. Surface. Approve. Act. Every step on live SAP.

STREAM Agents close the loop between operational event and action inside the SAP security model, on zero-copy live streams, with human oversight where policy requires it.

Loop Stage What STREAM Agents Deliver What This Replaces
Sense
Event detection on live SAP		 Agents subscribe to SAP events via zero-copy streaming. Every state change is visible as it happens. No batch window. No extract lag. Nightly warehouse extract. Events discovered 24 to 48 hours after they occur.
Surface
Prioritized exception with context		 Exceptions arrive with full operational context: linked docs, upstream root cause, downstream impact. Prioritized by policy and impact. Row in a batch report. Owner triages without SAP context. Root cause lost by the time the exception is reviewed.
Approve
Human-in-the-loop, native SAP		 Approvals routed through native SAP application authorizations. Policy determines what is autonomous, what needs review. Audit trail native. Shadow approval tool outside SAP. Split audit trail. Correction after the fact is manual.
Act
SAP-certified execution		 Agents execute via SAP-certified interfaces. Every action lands in the SAP audit trail. Correctable in the SAP correction model. Manual TCODE navigation by operator. Or shadow write from an outside tool with no SAP audit link.

P2P Agent has been running this loop autonomously in production since November 2023. Not a demo. Not a pilot. Continuous operation on live SAP data, inside the SAP security model, with native audit and correction. This is what production agentic execution looks like.

Meet the Agents

Ready-to-deploy agents. Streaming live SAP data from day one.

No configuration project. 100% extensible via secured APIs. Each agent runs on USB4SAP: zero-copy, SAP-certified, live data. Native SAP authorizations. Native SAP audit.

P2P Agent · Supply Chain

Autonomous procure-to-pay exception handling, in production since Nov 2023

Monitors live procure-to-pay events. Surfaces exceptions with full context. Resolves them without human batch review cycles, per policy.

  • Autonomous in production since November 2023
  • Runs inside native SAP authorization model
  • Escalates only where policy requires human review
SAP Chat · UI and API

Natural language SAP execution, without the GUI

Any TCODE or report accessed through conversation. Operators act without navigating SAP GUI. Programmatic API for embedding SAP execution into any application.

  • Any TCODE, any report, in natural language
  • Native SAP authorizations, no shadow permissions
  • API for embedding SAP execution into apps and workflows
AI Execution · Supply Chain

Agents executing supply chain decisions on live data

Inventory allocation, order prioritization, exception routing. At the speed of the operational event, not the batch window.

  • Inventory allocation and order prioritization live
  • Exception routing with linked SAP context
  • Policy-driven autonomy, human oversight where required
AI Insights API

End-to-end supply chain AI on one API

Combined SAP, trade, and sales intelligence through a single API. The complete operational picture for agents and apps. Same live foundation, unified surface.

  • SAP + trade + sales in one API
  • Live via MCP Catalog, no warehouse copy
  • For agent builders, ISVs, and enterprise AI teams
Agent Works When Fails When
P2P Agent Procure-to-pay exception volume is high and manual review is the bottleneck. Policy for autonomous resolution can be defined. Procurement is fully manual, low-volume, and every case requires case-by-case human judgment with no repeatable policy.
SAP Chat Operators spend meaningful time navigating SAP GUI or waiting on IT reports. Field or floor users need SAP access without training on TCODEs. All operational actions already flow through purpose-built apps and no operator ever touches SAP directly.
AI Execution Inventory allocation, order prioritization, or exception routing are currently batch or spreadsheet-driven with visible operational cost. Volumes are small enough that manual coordination is faster than defining and maintaining agent policy.
AI Insights API You are building enterprise AI on SAP, trade, and sales data and each new use case is currently a new pipeline project. You have no AI or agent roadmap and no plan to consume SAP data programmatically.
Platform Architecture

Enterprise-grade. Zero-copy. SAP-certified. Microsoft-certified. Patented.

STREAM Agents run on STREAM Engine: live access to 850,000+ SAP tables, 450,000+ semantic models, and any REST source. No pipeline. No extractors. No data copy.

STREAM People@Work

Live SAP for operators

The human side of the same live streaming layer the agents use. O2C and P2P process visibility in Fabric, Power BI, and Excel. No IT dependency.

STREAM Agents

Agentic execution on live data

P2P Agent (autonomous since Nov 2023), SAP Chat, AI Execution, AI Insights API. Sense, surface, approve, act, on zero-copy live SAP.

STREAM Engine · USB4SAP

Zero-copy · SAP-certified · Patented

Live access via MCP Catalog and REST for agents, apps, and Python workflows. 850,000+ SAP tables on-demand. In production since 2021.

TradePro and SalesPro

Companion streaming intelligence

Trade compliance, tariff modeling, and revenue signals on the same live foundation. Different operational surface, same zero-copy layer.

Certifications: SAP NetWeaver Certified · Microsoft Certified (Excel, Power BI, Fabric) · USPTO Patent · Azure Marketplace Published · In Production since 2021 · P2P Agent autonomous since November 2023 · 90% reduction in SAP data access cost vs traditional replication (Microsoft co-innovation)

Deployment Options

Three ways to start. Each begins with your biggest operational clog.

Value-from-Velocity from day one. No configuration project required for STREAM Now. No data leaves your environment under STREAM Activate.

STREAM Now

Subscribe and Consume

Pre-built STREAM Agents on-demand. SaaS, on-premise, or customer cloud. Value from day one. No configuration project required.

For: Operators and AI leaders seeking live agentic execution now.

STREAM Activate

Deploy in Your Infrastructure

STREAM Engine in your cloud or on-premise. Data never leaves your environment. Full sovereignty. 100% native SAP application authorizations. No duplicate infrastructure.

For: Regulated industries and enterprises with strong data sovereignty requirements.

STREAM Build

Build on the Platform

SIs, ISVs, and enterprise teams build net-new agents on STREAM Engine, MCP Catalog, and AI Insights API. 850,000+ SAP tables on-demand via REST or Python.

For: Technology platforms and SIs building AI practices on live enterprise data.

Executive Leave-Behinds

Your clog. Your entry point. Your question.

The same root cause (batch-AI on stale copies) shows up differently by function. Start with the question your executive will ask.

Persona Their Clog STREAM Agents Entry Point Ask Them
COO Supply chain exceptions caught weekly, should be seconds. Replenishment signals 24 to 48 hours late. Safety stock held against uncertainty that could be removed. P2P Agent + AI Execution. Autonomous since November 2023. Exceptions closed in seconds, not weeks. Order and inventory decisions on live SAP. How many supply chain exceptions did your team handle manually last month?
CFO Cashflow forecasts on last week's receivables. Working capital tied to exceptions that could close today. AR aging on batch reports, not live signal. P2P Agent for O2C exceptions + AI Insights API. Live cash and exception picture across the O2C cycle. How much working capital is tied to exceptions your team could close today with better information?
CDO / CAIO AI running on hours-old warehouse copies. Every new agent needs its own pipeline. Governance duplicated per copy. Attack surface grows with every use case. STREAM MCP + AI Insights API. Live SAP for AI agents via zero-copy streaming. 90% reduction in SAP data access cost vs traditional replication. How many pipelines does each of your AI agents need today, and how fresh is the data on the other side?
CIO Will agents destabilize SAP? Where does the data go? Another integration project? Another security perimeter to defend? STREAM Activate. SAP NetWeaver certified. No ABAP. Clean-core aligned. Data never leaves your environment. Native SAP authorizations. 59% faster than replication. What is your current cost and risk of feeding SAP data to AI agents today?
CIO and IT · Frequently Asked Questions

The questions IT asks before saying yes to agents.

STREAM Agents are SAP-certified and P2P Agent has been autonomous in production since November 2023. These are the concerns we hear most from IT, security, and audit teams, with direct answers.

Will letting agents act on SAP destabilize our production environment?

No. STREAM Agents access SAP through USB4SAP, which is SAP NetWeaver certified and uses standard, supported interfaces. There is no ABAP development, no custom transports, and no modification to standard SAP objects. Agents act inside the same interfaces SAP is designed to expose. P2P Agent has been in autonomous production since November 2023 without incident.

Do agents copy or store our SAP data outside the system?

No. STREAM Agents read live SAP via zero-copy streaming. There is no replication, no external pipeline, and no copy stored outside your infrastructure. Under STREAM Activate, data never leaves your cloud or on-premise environment. Agents see live SAP through the streaming layer at query time and act back through SAP-certified interfaces.

How does access control work? Do agents inherit SAP roles?

Yes. Every agent action is governed by 100% native SAP application authorizations. Agents inherit the same SAP role and object permissions used for human users. There is no shadow permission model, no new identity layer, and no separate access boundary. If an agent cannot see a record in SAP, it cannot act on it. Everything lands in the native SAP audit trail.

How do humans approve, override, or roll back agent actions?

Approval policy is defined per agent, per action class. Where business policy requires human review, approvals route through native SAP application authorizations, not a shadow approval tool. Where policy allows autonomous action, agents act and every action lands in the SAP audit trail, correctable through standard SAP correction. This is how P2P Agent has run in production since November 2023.

Is this another long integration project? What does deployment actually require?

No pipeline to build and no extractors to configure. USB4SAP provides live access to 850,000+ SAP tables via REST and MCP. Under STREAM Now, pre-built agents are available on-demand. Under STREAM Activate, deployment is within your own infrastructure. P2P Agent reached autonomous production in November 2023 without a prior integration project.

What are the performance implications for our SAP system?

USB4SAP is 59% faster than traditional replication-based approaches and reduces SAP data access cost by 90% compared to conventional methods (validated in Microsoft co-innovation). Because there is no replication layer, there is no ETL load on SAP. Reads are on-demand and governed by SAP's own query engine, with no bulk extraction competing with transactional load.

More questions about agent governance, platform architecture, or security? See the full FAQ or talk to our technical team.

Where is your biggest operational clog?

Start with the clog you feel most acutely: procure-to-pay exceptions in weekly review, SAP GUI as the operator bottleneck, AI agents starved by warehouse lag, or every new use case demanding a new pipeline. STREAM Agents remove it. Live SAP throughout. No pipeline. No batch.

Talk to Agent Engineering Try Live Agents