RFP Essential Knowledge: Fluent AI Capabilities
Essential knowledgeCore knowledge to understand Fluent Commerce
Author:
Michael Wholohan
Changed on:
5 June 2026
Overview
Fluent Commerce is built as an Agent Orchestration Platform — an intelligent OMS where AI is embedded in workflows, not bolted on from the outside. Intelligence is applied where commerce decisions are actually made: order promising, sourcing, fulfilment, customer service, and integration.This article maps Fluent's AI capabilities to the evaluation categories commonly assessed in enterprise OMS RFQs and RFIs, distinguishing clearly between what is live in production, what is in pilot, and what is on the committed roadmap.
Key points
- Product Area: AI & Intelligence
- Audience: Sales, Pre-Sales, Partners
AI Capability Summary
What are some of the AI powered OMS use cases today:
Production Capabilities
Estimated Delivery Dates (EDD) — ML-Powered Order PromisingFluent's EDD capability uses machine learning models to predict delivery times shown to customers at the pre-purchase stage — on product listing pages, product detail pages, and at checkout.
Model inputs: Historical carrier scan data, store performance by day and time, inventory accuracy by product type, weather forecasts along transit routes, store/warehouse calendars and operating hours, carrier cut-off times, and node lead times
Channels: Real-time EDD published across web, call centre, and store at order capture
Why it matters: Accurate delivery promises at the point of purchase improve conversion rates and reduce post-purchase WISMO enquiries
Order Management MCP — WISMO AI Agent
Fluent OMS acts as an MCP (Model Context Protocol) server, enabling AI agents to connect directly to order management data and handle customer enquiries via natural language.
The problem it solves: 60% of customer service enquiries are WISMO ("Where Is My Order") calls, with a handling cost of approximately $5.50 per call.
How it works:
1. Customer asks about their order via natural language through any AI agent surface
2. The AI agent queries the Fluent OMS MCP server to retrieve relevant order data
3. The agent converts the data into a natural language response
Outcomes:
- Up to 77% of enquiries resolved by the AI agent
- Cost per interaction reduced to <$1 vs ~$5.50 for a human-handled call
- Supports future growth without increasing customer service headcount
- Compatible with leading AI agent platforms
What the agent can surface: Order details, item information, shipment status, return eligibility and policy, warranty information, and item utilisation guidance.
Agentic Commerce via Universal Commerce Protocol (UCP)
Fluent exposes standardised commerce capabilities through the Universal Commerce Protocol, enabling AI agents to discover and execute commerce transactions directly.
Capabilities exposed via UCP:
- Inventory availability
- Fulfilment options
- Create order
- Create return
Purpose: OMS must remain model-, protocol-, and surface-agnostic so retailers can meet customers wherever discovery and purchase happens — chat interfaces, voice assistants, agentic checkout surfaces, and emerging AI storefronts.
Inventory Allocation & Rebalancing Agent
Fluent OMS acts as a control plane for AI agents handling inventory allocation and rebalancing recommendations across the fulfilment network.
- Inputs: Orders, inventory positions, demand signals, sell-through velocity, margin targets
- Output: Recommended optimal store transfers and allocation decisions; transfer orders generated with human approval
- Governance: Human-in-the-loop approval before actions are taken; all recommendations observable and reversible
Business outcomes targeted:
- Increased product availability for future demand
- Reduction in split shipments
- Cost optimisation between store transfers and missed demand
Fluent Connect — Connector AI Studio
Fluent's proprietary AI-assisted integration builder creates OMS connectors from natural language prompts.
- Build time reduction: 75%+
- TCO reduction: 3–5x lower support and maintenance cost
- How it works: Describe the integration in natural language → AI scaffolds the connector, maps schemas, aligns data fields, and generates code → human validation → deploy to production
- Important: AI is used only in the build and maintenance phases. AI is never used in the runtime of a connector.
See the Fluent Connect knowledge article for full detail.
Builder MCP Server — Developer Co-Pilot
An AI co-pilot for implementation teams that configures, extends, and deploys Fluent faster using platform-aware guidance and best-practice patterns.
- Effort reduction: 30%+ in build and configuration
- Capabilities: Scaffold and deploy Fluent projects, generate and run GraphQL queries, manage modules and dependencies, debug and troubleshoot, build rules and workflows (Beta), query documentation and validate patterns (Beta)
- Governance: All code goes through multiple stages of human review plus static code analysis and vulnerability scanning
In-Pilot and Roadmap Capabilities
Conversational Analytics (In Pilot)Enables operational and business users to query OMS data using natural language rather than building reports manually.
- Ask questions in natural language → receive metrics, visualisations, and insights
- Build personalised dashboards, generate custom charts, and schedule reports via natural language
- Built on top of the Fluent Analytics platform
See the Fluent Analytics knowledge article for the full reporting roadmap.
On-Platform Agents (In Development)
A configurable agent framework enabling AI agents to operate directly inside Fluent OMS applications, under the same governance, workflow controls, and access model as human users.
Named use cases:
- Order Sourcing Optimisation
- Customer Service Co-Pilot
- Store Capacity Optimiser
- Order Exception Handling
Every agent action is observable and reversible. The framework supports both insight/recommendation agents (with optional user actions) and conversational co-pilots.
Builder AI Skills (In Development)
Packaged AI skills extending the Builder MCP Server capabilities:
- UI component configuration and build
- Rules analysis and workflow configuration
- Orchestration audit analysis
- CLI command mapping
AI Architecture and Governance
How AI is DeployedAI is embedded directly in Fluent OMS workflows — not a separate analytics layer or third-party bolt-on. Foundation models are accessed via cloud partners (AWS and GCP), including Gemini, Anthropic Claude, Amazon Nova, and others. The choice of model may vary by use case.
Each AI agent is configured with:
- Agent type — the purpose and scope of the agent
- Accessible data entities — e.g. Orders, Inventory, Fulfilments, Locations, Customers
- System prompt — where behaviour, thresholds, and decision-making rules are defined
Transparency and Explainability
- Show Reasoning: Agents support an expandable section revealing chain-of-thought and data sources used to reach a recommendation
- Configuration: Decision-making process, confidence thresholds, and available user actions are all configurable per agent via system prompt
- Human-in-the-loop: All agents support a human review and approval process before any action is executed
- Auditability: All agent actions are logged, observable, and reversible through cloud provider audit tooling (AWS/GCP)
Data Usage and Model Training
- Fluent uses foundation models only and currently does not train proprietary models on customer data
- Customer data is not used across tenants or in shared/global model training
- In any future scenario where Fluent trains models, the Data Privacy Addendum will be updated to reflect the scope and any related procedures
- Development uses private instance models hosted on AWS/GCP with Fluent-configured data retention policies; retained assets are contained within Fluent's private cloud
- No customer data is used during agentic coding processes
Regulatory Compliance
Fluent's Agent Orchestration Platform is built to EU AI Act standards (full enforcement August 2026):- Every AI agent action is logged, observable, and reversible
- Zero-trust principles and data minimisation applied across the platform
- All AI-driven changes follow the same SDLC process as standard product development, including formal test evaluations for model changes
Full compliance documentation is available via the Fluent Trust Center: https://docs.fluentcommerce.com/landing/trust-center
AI Governance in Engineering
- Code development: Private cloud AI tools only; no real customer data; multi-stage human review + static analysis
- Testing: Unit, contract, integration, and performance testing; AI used to generate additional test cases and sample data
- Incident detection: Logs and metrics shared with AI agents for root cause analysis; PII excluded from logs by design
- Tool governance: Only tools approved by the AI Governance team may be used; all validated for privacy and compliance
