What do you need from us to set Pliney up?

A knowledge source (FAQs, SOPs, docs in any format), credentials for the tools you want Pliney to read from or act in, and a few sample queries so we can calibrate the agent's behavior. Most pilots start with a single workflow or surface (e.g. internal data querying or customer-facing resolution).

How do you make sure Pliney isn't making things up?

Three layers: grounded retrieval (Pliney can only answer from cited sources), a refusal default (out-of-scope questions get explicit 'I don't know'), and RAGAS (Retrieval-Augmented Generation Assessment, an open evaluation framework) scoring every turn for faithfulness, answer relevancy, and context precision. Failures are caught and reviewable.

Does Pliney replace our team?

No. Pliney handles the routine 60–80% of queries that follow a clear pattern. Complex, high-stakes, or out-of-scope cases route to your team with the full conversation, sources used, and recommended next action attached.

What if we have a custom internal system Pliney needs to read from?

We support REST APIs and Model Context Protocol (MCP) servers. If your system has an API, Pliney can use it. If it doesn't, we scope an integration as part of the engagement.

How does Pliney work with Horizon?

Pliney calls Horizon's forecasting and decision APIs as a tool. That means you can ask predictive questions in natural language (stockout probability, demand projections, cash position) and Pliney returns answers grounded in Horizon's live forecasts with full citations to the underlying data.

Can Pliney run on our infrastructure?

Yes. Default is hosted on GCP, tenant-isolated with AES-256 encryption. VPC and on-prem deployments available for enterprise customers. The vector store is per-tenant. Your data is never used to train shared models. SOC 2 Type II-aligned controls.

Pliney · The grounded AI agent for operations teams

The AI agent your operations team actually trusts.

Not a chatbot. Not enterprise search. Pliney is a grounded agent that queries your live ERP, CRM, and helpdesk, runs Horizon forecasts on demand, and resolves customer requests end-to-end. Every answer cites a source, every turn is RAGAS-evaluated, and out-of-scope questions get refused rather than guessed at. In production with retail and travel customers.

Book a demo How it works

Customer

Why was my refund declined? Order #84021.

Pliney

60-80%

Routine queries resolved end-to-end

Live ERP

Grounded answers, no static FAQs

RAGAS

Faithfulness scored on every turn

Per-tenant

Data isolation by default

Generic chatbot

Answers from a static FAQ. No live data.
Hallucinates outside its training corpus.
Routes anything complex straight to a human.
No memory of prior interactions in your tools.
Can't take actions in your booking, helpdesk, or CRM.
No evaluation. No way to know what it got wrong.

Pliney

Grounded retrieval over your sources, with citations.
Refuses to answer outside scope. No invention.
Triages, routes, and resolves end-to-end where possible.
Pulls live data from helpdesk, booking, ERP via MCP.
Takes actions: opens tickets, books slots, issues credits.
RAGAS evaluation on every turn. Faithfulness measured.

01 · LIVE FORECAST QUERYING

Ask your forecasts in plain language

For Planning, Finance & Commercial Teams

What Pliney does that ChatGPT, Glean, and copilots cannot: query a live forecasting model. "Which products have a high probability of stockout in the next 5 weeks?" "Expected demand for Product X next quarter?" "Projected cash position for Q3 if we delay this promotion?" Answers grounded in Horizon's live forecasts, with confidence intervals and the signals that drove the answer.

Integrated with Horizon forecasting engine via internal API
Stockout risk probabilities and demand projections
Treasury and cash flow forecasting queries
Confidence indicators and explainability per answer
Side-by-side scenario comparisons
Available standalone, or as a forecasting front-end if you've deployed Horizon

02 · OPERATIONAL DATA QUERYING

Ask your operations in plain language

For Internal Operations & Analytics

Ask questions in natural language and get answers grounded in live data: "Show me sales by region for the last quarter." "Which products had the highest returns last month?" "List the top 10 products running low on stock." No SQL and no analyst required.

Connected to ERP, e-commerce, and sales data in real-time
Returns visualisations and summary dashboards
Role-based access and data governance built in
Multi-language support (en, pt, es)
Per-tenant vector store, isolated by default
Citations surfaced inline for user and operator

03 · PROACTIVE ALERTS

Surface what matters before someone asks

For Operations & Leadership

Pliney doesn't wait for questions. It surfaces anomalies, flags risks, and recommends actions: liquidity warnings, demand surges, reconciliation mismatches. Your team sees the right information before they think to ask for it.

Configurable alert thresholds and notification channels
Automated anomaly detection across operational data
Action recommendations with confidence indicators
Slack, Teams, email, web widget surfaces
Per-action audit log
Smart escalation to human agents when needed

04 · CUSTOMER-FACING RESOLUTION

Resolve customer requests, don't just deflect them

For Support Teams & Customer Operations

For customer-facing teams: Pliney handles order tracking, returns, product questions, and account changes. Connects to your order system and CRM to resolve queries end-to-end via MCP servers and REST. Audit every action.

Real-time order status and tracking lookups
Returns and exchanges processed per your policies
Multi-agent triage, routing, resolution
Smart escalation to human agents with full context
MCP server integrations (Freshdesk, Zendesk, Jira)
RAGAS auto-evaluation on every turn

Deployment evidence

Where Pliney is in production today.

Anonymised production deployments. Each grounded in live data from the customer's operational systems.

Retail distributor

Replaced static reports with dynamic natural language querying across ERP, e-commerce, and sales data.

Fashion retailer

Finance teams query treasury and reconciliation data in natural language, connected to 5 banks and fintech platforms.

Horizon integration

Predictive queries powered by Horizon forecasting: stockout risk, demand projections, cash position forecasts.

Fashion & Retail

Finance teams query treasury and reconciliation in natural language. Customer-facing teams resolve order, return, and product questions grounded in live order data.

Treasury querying across banks
Reconciliation anomaly alerts
Order status and returns resolution
Multi-language out of the box

See how it works

FMCG & Distribution

Replace static reports with conversational analytics over ERP, e-commerce, and sales data. Surface stockout risks and recommend field-team actions before stockouts happen.