AI Applications
AI that answers from your own documents and links each answer to its source.
Oil Market AI: answers and price scenarios from OPEC, IEA and EIA reports
Built for oil-market analysts, Oil Market AI puts monthly agency reports and market data behind one chat. Answers list the report passages they used; scenarios state probability, Brent impact and method.
- Status
- Live at orpitek.com/oilmarketai
- Users
- Oil-market analysts and researchers
- Data
- OPEC MOMR, IEA OMR, EIA STEO; market data from the EIA, FRED, JODI, the World Bank and IMF PortWatch
- Our role
- Product design, report pipeline, AI answers, web application
- Problem
- OPEC, IEA and EIA monthly reports run long, change every edition and often disagree, so one question can mean reading several reports by hand.
- What we built
- A live web application: a chat that answers from OPEC, IEA and EIA reports, a library of Brent price scenarios and a market dashboard.
- Result
- Analysts get a structured answer tied to its source, without reading each report in full.
The problem
Three monthly publications track oil supply and demand: the OPEC Monthly Oil Market Report (MOMR), the IEA Oil Market Report (OMR) and the EIA Short-Term Energy Outlook (STEO). Each runs to dozens of pages. Comparing OPEC and IEA demand forecasts for 2026 means opening several editions. Turning an event such as a Gulf of Mexico hurricane into a Brent view also needs market data and history.
What we built
The chat answers oil-market questions from the report library. The scenario library covers what-if questions on sanctions, shipping chokepoints, OPEC+ policy, weather and demand. The dashboard tracks prices, OPEC output, OECD stocks, supply and demand, refining margins and market events. An MCP server lets analysts query the report library from their own AI assistant, with an API key from their account page.
Reports and data
Reports are converted from PDF, including scanned pages, and split into sections. Sections are indexed twice: by meaning for plain-language questions and by exact keyword for a specific figure or series name. Market data comes from the EIA, FRED, JODI, the World Bank and IMF PortWatch. Each scenario is computed from these figures as they stood on the day it was run.
Answers with sources
Every chat answer has a Sources panel: report edition, section and match score for each passage, plus web sources when the reports fall short. A scenario answer states the probability, mechanism, historical analogy and supply-demand effect, with the Brent impact in percent and US dollars per barrel. “How this was computed” gives the formula and each assumption. Scenario pages are marked as scenario modeling, not investment advice.
The result
An analyst can ask how the latest OPEC and IEA demand views differ and get one comparison table. Each scenario shows its Brent impact and reasoning on one page. The same design works on internal material: an energy company, an investment team or a ministry could apply it to its own reports, research notes and data.
What the product does
- Chat: plain-language answers from OPEC MOMR, IEA OMR and EIA STEO
- Web sources: added when the reports do not cover a question
- Scenario library: what-if questions on sanctions, chokepoints, OPEC+ policy and demand
- Probability: a blended market and expert estimate for each scenario
- Stated method: formula, assumptions and dated base figures behind each price estimate
- Market dashboard: prices, OPEC output, stocks, refining margins and market events
- MCP server: the report library inside the analyst’s own AI assistant
Related solutions
Data Platforms
One clean data model, so dashboards, search and AI show the same figures.
Systems Integration
Connect AI and new software to the systems you already run.
Related industries
More Case Studies
Education · Public data platform
Dubai School Atlas
Dubai’s private schools on one map with KHDA ratings, fees and curricula. Families ask questions in plain language and every answer links to its source report.
Send us the reports your analysts read every month
List the questions they help your team answer. We will propose which reports and data a first pilot should cover.