Agents – Research & Strategy
Case Study – Agentic Stack
One of the biggest bottle necks in getting to great work, is great direction. Strategists are looking for insights that help construct exceptional briefs, regardless if it’s a major integrated campaign or an ongoing social media retainer. The agentic stack is designed to automatically scour the internet on routine intervals for recent events in the competitive and social landscapes. It also looks to the future for potential cultural opportunities. All of this is routed through strategists that prioritize and summarize pages of information so that the user is armed with highly relevant information to tailor their briefs.
The Approach
Agentic Flow
Numerous agents are daisy-chained, performing specific tasks then passing the data to a strategist stack to interpret and prioritize based on company goals.
States and Outputs
Agentic outputs are transformed into states for review downstream. Final outputs as clean, organized text for users to add to briefs.
Prompting
Working from a simple proof of concept through to refined outputs, prompts were evolved for AI precision.
Learnings
The OpenAI workflow builder is less robust than seasoned platforms like n8n, but provided a low-cost option in a contained system. Working with ChatGPT, I was able to produce a state-driven agentic workflow that created an enormous amount of highly useful information. The strategy stack was a simple prioritization and interpretation flow to pass to a strategy team, but it had plenty of opportunity for growth. The workflow was built to house a creative director node, social/web output node and even ingest analytical data for learning over time.
Let’s Talk!
Got a fun project to discuss? Let’s have a chat and see if we can’t get you something!
