Up-Level Your AI with Agents and Recipes

AI has the power to transform entire industries – and one could argue that it’s already doing that. Look no further than the spike in AI adoption across sectors and across the federal government. Global research firm Gartner held its Data and Analytics Summit last week, which many Finch AI team members attended, and noted that AI pilots are becoming increasingly ubiquitous and today the focus is shifting toward consistent execution of AI-driven initiatives. They’re right.

But in an environment where everyone – from your competition to your adversaries – is using AI, how can organizations stay ahead? How can they leverage the transformative potential of AI while also innovating with it at the same time?

Our work at Finch AI involves developing tools that accelerate analyst workflows. Every analyst wants to work more quickly, more thoroughly and more accurately. Those aren’t destinations. They’re moving targets. So, in the constant pursuit of “better,” we bring solutions to analysts that up-level their capabilities in meaningful ways.

Two such examples are AI Agents and AI Recipes.

I’ll explain.

AI Agents are autonomous systems that scan, analyze, reason and report what they find. We’ve built multiple agents for specific tasks. These include ones for initial research, hybrid RAG search, graph search, deep research, summarization, risk assessment, entity discovery, proofreading, rewriting, and citation verification. These agents are large language model (LLM) agnostic, working with any language model you have.

Imagine you lead a research institution’s efforts to detect foreign influence in grant applications. Autonomous AI agents can scan global databases, past grants, research papers and institutional collaborations and collaborators, for risk indicators. Its findings could surface undisclosed foreign connections or risks for nefarious influences – better and faster than a human would be able to.

Or imagine you’re a national security analyst tasked with combing through a never-ending feed of social media content from multiple persons of interest. Agents can surface relationships or connections and can continually update their findings as new content comes in.

Clearly, these agents are powerful tools in an analyst’s arsenal – and with the capability to become even more powerful over time.

Similarly, AI Recipes are important up-levelers for analysts.

AI Recipes are pre-defined but adaptive workflows that integrate multiple tools and datasets for deeper insights. Here, too, we have built multiple types of AI recipes at Finch AI to support multiple use cases. We use them to build Finch Insight Reports across a number of domains including due diligence, risk management, situational awareness and more. Not only have we created recipes for multiple report types, we’ve also built a recipe developer kit to help our customers build their own recipes using the same AI agents and workflows.

In action, these recipes can direct agents on how to comb government watchlists, networks and financial transactions to detect threats. Essentially, the recipes are designed to ask domain-specific questions of these massive datasets and then to leverage a retrieval augmented generation (RAG) pipeline to ask secondary and tertiary questions that are dependent upon the results.

We see agents and recipes as key parts of automated risk detection and reporting going forward. Our use of them now – and the ability to continue to refine and operationalize them – means they will soon become foundational elements of the next wave of AI innovations. That will be around driving down false positives and negatives; incorporating explainability and trust into reports; and effectively integrating human in-loop oversight.

And rest assured we’re already breaking new ground on those fronts as well.

To learn more about AI agents and recipes, or about what’s on the horizon for us, please visit www.finchai.com or get in touch at hello@finchai.com.