ProductOnDemand Media Planning Agent

Before you start

Demo purposes only.

This tool is a demonstration of ProductOnDemand's media planning capabilities. Recommendations are AI-generated from sample benchmark data and should not be used as the basis for actual budget, media, or campaign decisions.

Acceptable use.

This tool is intended for exploring media planning scenarios. Abusive, hostile, or excessive use — including prompt injection attempts, offensive language, or automated querying — may result in access being revoked.

No liability.

ProductOnDemand provides this tool "as is" with no warranties. ProductOnDemand assumes no liability for decisions, losses, or outcomes arising from use of this tool or its outputs.

Interested in building an AI-powered planning agent like this for your business? Let's talk.

What is this agent?

The ProductOnDemand Media Planning Agent is an AI-powered assistant that helps advertisers and publishers plan publication-based media — compare publications, allocate budgets across formats, and understand pricing and availability benchmarks — all through a simple conversation.

How it works

You share your goals. The agent asks about your campaign objective, budget range, and industry vertical so it can tailor recommendations to you.

It searches ProductOnDemand's knowledge base. Your question is matched against curated media benchmarks, pricing data, and campaign playbooks using semantic search.

An LLM crafts your answer. The retrieved data is combined with your conversation context and sent to a large language model, which generates a tailored media plan or recommendation.

You iterate. Ask follow-ups, adjust budgets, swap channels, or drill into specifics. The agent keeps your full conversation context so each answer builds on the last.

Technology

RAG Pipeline Semantic Search LLM-Powered pgvector

The agent uses Retrieval-Augmented Generation (RAG): your questions are embedded as vectors, matched against ProductOnDemand's proprietary data via cosine similarity, and the best-matching context is passed to the LLM alongside your conversation history. This means answers are grounded in real data — not hallucinated.

Good to know

All recommendations are AI-generated from sample benchmark data and are for demonstration purposes only.
The agent only handles media planning topics — it will politely redirect off-topic questions.
No personal data is stored. Conversations exist only in your browser and are cleared on page reload.
There is a rate limit of 20 queries per hour to keep the service available for everyone.

What data powers this agent?

The agent draws from ProductOnDemand's curated media intelligence database, covering publications, pricing, availability, events, and lead generation programs. Here's what's inside:

Publications Master List

A comprehensive directory of media publications across key verticals, including audience size, editorial focus, content formats offered, and vertical relevance.

B2B SaaS/Tech AdTech/MarTech Consumer/General Business

Formats & Pricing

Detailed pricing for ad formats across publications — display, native, sponsored content, newsletter sponsorships, webinars, and more. Includes CPM and CPL ranges, minimum spends, and format-specific benchmarks.

Q4 Availability

Real-time inventory status for Q4, including fill rates and availability flags (Available, Limited, Sold Out, On Hold) across publications and formats. The agent won't recommend sold-out placements.

Events Calendar

Upcoming industry events and conferences with dates, locations, expected attendance, sponsorship tiers, pricing, and current availability.

Lead Generation Programs

Content syndication and lead gen programs with guaranteed lead volumes, CPL ranges, targeting options, and Q4 availability status.

How the data is used

When you ask a question, the agent converts it into a vector embedding and searches the database for the most relevant data chunks. Only the top matches (those above a relevance threshold) are passed to the LLM to inform its answer. This means:

Answers cite real publications, real pricing, and real availability — not generic guesses.
Budget recommendations respect minimum spends and actual cost ranges.
Channel comparisons use actual CPL/CPM benchmarks from the database.
When data coverage is limited for a specific query, the agent will tell you and note it's using general benchmarks.