There is fierce competition in this space, driven by the potential for AI to revolutionize every aspect of the investment process, from research and analysis to execution and risk management. Traditional financial firms, fintech startups, and financial information providers are all investing heavily in AI capabilities to stay competitive and relevant in this rapidly evolving landscape.

I have tested out codes + APIs to research unstructured data such as transcripts to research and analyze the market, but it would be greater if I can create agents to automate what I did more efficiently and routinely.

There are several commercial use cases where startups have built agents using APIs to serve various functions:

1. E-commerce and Shopping Assistants: Startups can use APIs like the Retrieve API to develop AI-powered personal shopping assistants. These agents can autonomously scan e-commerce sites, compare prices, and gather product reviews to provide personalized recommendations to users.
2. Lead Generation Tools: Startups can create automated lead generation tools that employ APIs to crawl company websites, social media profiles, and business directories, compiling detailed prospect lists for sales teams.
3. Content Creation: Web agent startups focused on content creation can utilize APIs to gather trending topics, relevant statistics, and reputable sources across the internet, aiding in generating data-driven articles and reports.
4. Business Intelligence: APIs are used in business intelligence applications to collect, store, and analyze data. This helps organizations make better business decisions by integrating various data sources and automating processes.
5. Healthcare: In healthcare, API integration allows for connecting different systems, improving patient care, reducing costs, and increasing efficiency by sharing data and automating processes.

I have also built a live Chat Bot to implement if agents are built in the backend. So similarly, for indexing business, I can use this chatting Bot UI to implement lead generation, content creation, intelligence creation, model creation, backtesting…

Now referencing advanced level of agents, 1. could have multiple agents, 2. each agent is equipped with tools such as function calling, web search, RAG, 3, agent could have multiple stages such as planning, thinking, checking, verification etc.

This is a good GIT open source lite library of agents to reference: https://github.com/salesforceairesearch/agentlite

For example, below codes to define an action of agent to use API to search Wiki content:

from agentlite.actions.BaseAction import BaseAction
from langchain_community.tools import WikipediaQueryRun

class WikipediaSearch(BaseAction):
    def __init__(self) -> None:
        action_name = "Wikipedia_Search" 
        action_desc = "Using this API to search Wiki content." # LLM uses action_name and action_desc to understand this action
        params_doc = {"query": "the search string. be simple."} # LLM uses this params_doc to understand the parameters in self.__call__() function
        self.search = WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper())
        super().__init__(
            action_name=action_name,
            action_desc=action_desc,
            params_doc=params_doc,
        )

    def __call__(self, query):
        return self.search.run(query)