- The Opportunity Scanner
- Posts
- AI Agents in 2025 - Are they worth the hype?
AI Agents in 2025 - Are they worth the hype?
PLUS: How to build AI Agents as a non-tech person & 3 AI Agent Business Ideas
The Opportunity Scanner
January 08, 2025
What are AI agents, How do they work, and Why is everyone talking about them in 2025?
Think of AI agents as digital assistants on steroids - they're smart software that can actually think, act, and learn on their own. Unlike your typical computer program that just follows a strict set of rules (like if A happens, do B), these agents can figure things out as they go, kind of like how we humans learn from experience.
The magic happens through three main parts working together:
Think of "sensors" as their eyes and ears - they gather information from their environment
Their "brain" (the control center) processes this info and decides what to do
Then their "hands" (effectors) carry out the actions
Source: World Economic Forum
What's really got everyone excited is how far these agents have come. They've evolved from basic rule-followers to sophisticated systems powered by the same tech behind ChatGPT (large language models) and can now handle images and videos too.
They can work alone or team up with other agents to tackle complex problems - imagine having a whole team of digital experts working together seamlessly.
The buzz isn't just hype either. Companies are seeing real results - these agents are handling customer service inquiries, managing supply chains, and even helping doctors analyze medical data.
The really cool part? They get better at their jobs over time, learning from each interaction just like a human would. That's why industries from healthcare to logistics are betting big on this technology.
Is it all fluff? Are AI agents actually being integrated into businesses?
Let's talk numbers - LangChain recently surveyed over 1,300 professionals, and the results are eye-opening. AI agents aren't just buzzwords anymore - they're becoming essential business tools.
Here's what's happening on the ground:
Just over half (51%) of organizations are already using AI agents in real-world applications. Interestingly, it's the mid-sized companies (with 100-2,000 employees) who are really going for it, with 63% having agents in production.
For those who haven't jumped in yet? A whopping 78% are actively planning to do so. But here's the real kicker - 90% of non-tech companies are either using or planning to use agents, practically neck-and-neck with tech companies at 89%.
LangChain’s Survey of 1300+ Enterprises
So what are companies actually using these agents for?
Nearly 60% are putting them to work on research and summarization tasks
About 54% are using them as personal productivity assistants
And 46% have them handling customer service duties
LangChain’s Survey of 1300+ Enterprises
But companies aren't just throwing caution to the wind. They're being smart about it:
Most are keeping their agents on a tight leash with read-only permissions
Bigger companies (2000+ employees) are especially careful - 51% of tech firms use multiple safety measures, compared to 39% of other industries.
And everyone's biggest worry? Making sure these agents actually perform well - it's more than twice as concerning as costs or safety risks
LangChain’s Survey of 1300+ Enterprises
What's clear is that AI agents aren't just experimental toys anymore. Companies are getting serious about implementation, but they're doing it thoughtfully - especially when sensitive data is involved.
The focus has shifted from "What could AI agents do?" to "How can we make them work effectively and safely in our business?"
How Moody’s is Using AI Agents in Financial Services
Moody's is leading the charge in deploying AI agents for financial services, moving beyond basic chatbots to create sophisticated autonomous systems. The company has developed a four-stage GenAI maturity model: from assisted intelligence (basic chatbots) to autonomous intelligence (self-learning agent networks).
The company leverages three key frameworks:
Autogen for generating comprehensive credit reports (If you want to get started with Autogen here’s your guide - AutoGen Guide)
CrewAI for analyst-AI collaboration (CrewAI is a low code platform for developing AI Agents, literally anyone can build AI Agents with Crew AI. Get started here: CrewAI Tutorial for Beginners)
Lanraph for visualizing complex financial relationships
A Moody’s AI agent automatically plotting a trading graph with the last two years of stock prices with just one simple, unoptimized prompt — all in about 10 seconds:
Their latest product, ReconAI in the EDF-X platform, employs multiple agents to:
Monitor real-time Early Warning Signals for companies of any size
Track global news, controversies, and regulatory filings
Analyze the impact on specific companies
To ensure accuracy, Moody's implements a triple-verification system:
RAG evaluation using LLM judges
What are LLM Judges?Search engine performance metrics
Multiple agent voting system comparing results across GPT-4, Claude 2, Llama 3, Gemini, and Moody's own models.
This makes Moody's the first financial institution to launch a production-ready agentic platform in their suite.
How can you ride the AI Agent Wave?
Here are some business ideas to start in this space based on gaps identified after reading reports from Deloitte, Moody’s, World Economic Forum and LangChain.
AI Agent for Customer Testimonials | |
|---|---|
Problem | Marketing teams need to constantly repurpose customer testimonials into various formats like social proof, case studies, and sales decks. |
Solution | An AI agent that automatically converts customer testimonials into multiple formats, streamlining content creation. |
Revenue Model | $300/month |
Initial Customers | Marketing teams in small to mid-sized companies |
Workflow Duplication Monitoring Agent | |
|---|---|
Problem | Companies frequently run the same processes across multiple tools like Notion, Linear, and Asana, leading to inefficiencies. |
Solution | An AI agent that identifies and alerts teams about duplicated workflows across different tools, suggesting consolidation. |
Revenue Model | $2k/month to consolidate |
Initial Customers | Mid to large-sized companies using multiple tools |
Competitor API and Cost Monitoring | |
|---|---|
Problem | Product teams often miss competitor API changes and price hikes, leading to delayed responses or increased costs. |
Solution | An AI agent that monitors competitor API changes and pricing updates in real-time, alerting product teams proactively. |
Revenue Model | $2k/month per company |
Initial Customers | Tech startups and mid-sized SaaS companies |
Some Tutorials and Videos to get started with AI Agents as a Non-Tech Person
Thanks for reading! Stay tuned for the next edition, where we’ll dive into more complex insights and uncover fresh investment and startup opportunities to keep you ahead of the curve.
