AMOS

An agentic marketing planning tool that converts business objectives into marketing plans and forecasts

Comment

Inspiration

I spent three years at Taneira, the ethnic wear brand from the Tata Group, where the festive season — Varamahalakshmi through Diwali into early wedding season — drives nearly half of annual revenue. I led campaign briefs for that window, and the single most painful part of the job was the two months of cycles it took to turn a business objective into a final plan the teams could actually execute on.

Most AI marketing tools today focus on execution: generate 20 Instagram captions, buy media more efficiently, automate A/B tests. Tools like Jasper, Smartly, and Basis Compass (launched nine days before this buildathon) are all excellent at the "do the thing" layer. But very few are working on the strategic reasoning layer that sits between "we have an objective" and "here's what to execute."

That's where AMOS lives — and it's genuinely underserved.

What it does

AMOS takes one sentence — a business objective with a target growth rate — and produces a complete, cohort-segmented campaign brief through a 5-agent pipeline with a deliberate human-in-the-loop checkpoint.

The system reads everything from a single Excel input file (brand context, cohorts, channel economics, brand guidelines, historic data, competitive landscape, model assumptions) and runs five specialized agents in sequence.

What you get:

  • A ranked list of customer cohorts labeled DEFEND or INVEST FOR GROWTH
  • Two brand-compliant creative routes per cohort, generated live via the Claude API
  • A media plan with 2–3 channels per cohort, budget allocation, and weighted channel effectiveness
  • A budget-driven reach model that tells you which cohorts you can saturate and which are budget-limited
  • A revenue uplift forecast with every assumption visibly row-referenced to the input file
  • A final verdict — ON TRACK WITH BUFFER or SHORT BY X% — with the exact gap to target

Live demo: https://amos-campaign-planner.vercel.app/

The 5-agent architecture

# Agent Role Type
1 Objective Intake Parses the business objective, loads baseline numbers Deterministic
2 Context & Strategy Consolidates 6 sheets of data, ranks cohorts via composite score, labels DEFEND/INVEST Deterministic
— Human Checkpoint Marketer approves strategy before execution Human
3 Creative Director Generates 2 brand-compliant routes per cohort Claude API
4 Media Planner Selects channels by best-fit, allocates media budget Deterministic
5 Impact Forecaster Computes budget-driven reach and risk-adjusted uplift Deterministic

Why only Agent 3 uses the Claude API: Creative variation is the one place where LLMs genuinely add unique value. Strategy and forecasting must stay deterministic

Built With

  • claude
Share this project:

Updates