Making time-sensitive questions easier to answer
At The Washington Post, questions about content, subscriptions, and advertising performance are frequent and often time-sensitive. The insights needed to answer them can live across multiple datasets and tools, and they often require deep institutional knowledge, while turnaround time can directly affect revenue conversations. For the Analytics team, responding to recurring questions can become a capacity bottleneck.
To address that challenge, the Post’s Analytics team is using ChatGPT and OpenAI APIs to build a network of agents that support questions about content, subscriptions, and advertising analytics. The agents reference internal documentation, metric definitions, aggregated tables, and permission logic, enabling more efficient querying while helping ensure that only relevant information is shared.
The work focuses on a practical goal: help teams get to useful answers faster, apply metric definitions more consistently, and reduce the time analysts spend responding to repeat questions.
Four ways The Post is applying OpenAI technology
The Content Agent
The Content Agent lets users ask natural-language questions about content performance, drawing on trusted, governed content datasets. Scoped access and predefined logic shape what the agent can share and how it responds.
For example, a user might ask which articles had the highest average engaged time last month, or whether a given section performed better this quarter than last. The agent applies the requested metrics and timeframes to approved data, giving teams faster answers to questions that would otherwise require an analyst to query and interpret the data directly.
The Advertising Agent
Advertising data is complex, and answering questions about it often requires deep institutional knowledge—knowledge that can be hard to scale across a team. Because turnaround time can directly affect revenue conversations, the Advertising Agent provides guided access to advertising performance data and applies consistent logic across questions.
The agent is designed to reduce back-and-forth on common requests, speed up answers during live planning and reviews, and support more consistent framing of advertising performance.
The Subscriptions Agent
The Subscriptions Agent is an internal analytics assistant built to help teams answer recurring questions about subscription acquisition and conversion in natural language. It provides consistent responses based on approved business definitions and trusted data sources, reducing reliance on manual analysis for routine questions.
The agent is being developed with clear guardrails around data quality and unsupported requests. Its initial focus is on validating whether this approach can improve speed to insight and expand self-service analytics.
AI Assistant
Alongside the data-enabled Content, Advertising, and Subscriptions agents, The Post is also developing AI assistants that support specific campaign and strategic workflows. Tools such as the Brand Metrics Assistant, Brand Voice Assistant, and Insights Agent connect users with curated knowledge bases containing established frameworks, guidance, research, and institutional knowledge.
Rather than primarily retrieving performance data, these assistants help teams apply that knowledge day-to-day work. For example, creating brand lift surveys, applying brand standards, and supporting the development of decks and presentations.
Adding concise insights to reports and dashboards
The Analytics team is also using OpenAI APIs to add automated, AI-powered insights to email and PDF reports. Each summary uses four or five concise sentences to explain the report’s contents and highlight notable trends, helping stakeholders grasp the key takeaways without reviewing every table or chart first.
The team has expanded these automated insights to dashboards, where they help users interpret trends across different time periods. The team continues to refine the prompts that guide the summaries based on stakeholder feedback.
Grounding answers in The Post’s own analytics context
The agents are designed around the context The Post already uses to answer analytics questions: internal documentation, metric definitions, aggregated tables, and permission logic. For the Content Agent specifically, trusted and governed datasets, scoped access, and predefined logic are central to how it works.
That grounding matters because a good answer depends on more than finding a number. It also depends on applying the right definition, using the appropriate logic, and sharing information relevant to the person asking.
What other news organizations can take from this approach
- Across the four efforts, The Post is applying OpenAI technology to a specific operational challenge: making complex analytics easier to query and understand.
- The Content Agent, Advertising Agent and Subscriptions Agent address different questions, datasets, and business needs. Keeping them distinct allows each agent to apply the logic and access appropriate to its use case drawing on the documentation, metric definitions, tables, and permission logic that already shape analytics work inside The Post.
- The automated insights bring a similar approach to reports and dashboards: a short explanation helps stakeholders identify the key takeaways and notable trends before reviewing the underlying detail.
