Marketing Against The Grain: AI Agents in 2025 – Where to Start & What Really Works (No Hype)
Hosted by HubSpot Podcast Network
Release Date: February 13, 2025

Introduction to AI Agents 2025

In the February 13, 2025 episode of Marketing Against The Grain, hosts Kipp Bodnar and Kieran Flanagan delve into the evolving landscape of AI agents. Joined by Joe Mora, CEO and founder of Crew AI, the discussion centers on demystifying AI agents, exploring their practical applications, and understanding their impact on various business functions.

Defining AI Agents

Kieran Flanagan opens the conversation by addressing the pervasive buzz around AI agents:

Kieran Flanagan [02:04]: "This is the year of agents, right? What do we really mean by that, and where are these agents having an impact?"

Joe Mora provides a clear definition, differentiating AI agents from traditional large language models (LLMs):

Joe Mora [02:50]: "The definition of an agent is it got to have agency. Unlike LLMs that predict content, agents can autonomously navigate tasks through reasoning."

He emphasizes that AI agents possess agency, enabling them to perform tasks independently rather than merely responding to prompts.

Current State and Capabilities of AI Agents

The hosts discuss the current capabilities and limitations of AI agents. Kieran and Kip highlight OpenAI's Operator as a case study:

Kip Bodnar [05:52]: "Operator has a browser control agent that can make reservations and do research, but it's still slow and clunky compared to human performance."

Joe Mora concurs, noting that while agents aren't yet outperforming humans in speed, their ability to handle tasks autonomously offers significant potential:

Joe Mora [07:00]: "Agents can help with tasks like posting on socials consistently, something I struggled with manually. Although slower, they free up valuable time."

Use Cases in Business

The episode explores various use cases where AI agents are making tangible impacts across business functions.

Marketing and Sales

Joe Mora shares an advanced application in marketing and sales, where agents enhance customer engagement through data-driven personalization:

Joe Mora [27:02]: "Agents enrich customer data, create hypotheses on usage, and integrate structured data into CRM systems, enabling hyper-targeted email marketing."

Kieran adds that such agents can significantly boost a sales representative's productivity:

Kieran Flanagan [29:33]: "With agents handling research and preparation, sales reps can manage more meetings and be better prepared."

Legal Automation

Joe discusses how AI agents are transforming legal departments by automating contract analysis:

Joe Mora [16:33]: "In the legal realm, agents automate contract analysis by providing recommendations and red lines, scaling what was previously cost-prohibitive."

SEO and Conversion Rates

A notable use case involves automating the A/B testing process for SEO and conversion rate optimization:

Joe Mora [32:24]: "Agents can analyze website content, research competitors, generate A/B test hypotheses, and implement changes, streamlining the optimization process."

Adoption Trends in Enterprises

The conversation shifts to enterprise adoption, with Joe Mora presenting compelling statistics:

Joe Mora [12:08]: "We're seeing over 16 million crew executions per month, with enterprises alone accounting for 23% adoption in production environments."

He explains that enterprises are centralizing AI agent deployment to maintain control and ensure security:

Joe Mora [21:56]: "Central deployment under CIOs or CTOs allows companies to manage LLM usage, apply filters, and enable reusable, secure use cases."

UX Patterns and Human Comfort with Agents

A critical aspect of AI agent adoption is the user experience (UX). Kieran references user concerns about autonomy:

Kieran Flanagan [09:56]: "When agents perform tasks autonomously, should users see the entire process? How much control do they retain?"

Joe Mora addresses the balance between automation and user oversight:

Joe Mora [10:55]: "Humans currently prefer to visualize and control agent actions to feel reassured. While full autonomy may come in the future, human-in-the-loop remains essential."

Starting with Agents: Recommendations

For organizations looking to integrate AI agents, Joe Mora offers strategic advice:

Joe Mora [18:11]: "Start with low precision, low-risk use cases to test agents' effectiveness. For example, agents can draft sales presentations or automate initial customer research."

He further recommends leveraging no-code platforms for non-technical teams:

Joe Mora [20:38]: "Use no-code platforms like Crew AI for easy deployment. As teams become more comfortable, involve technical personnel to customize and scale agents."

Future of AI Agents in the Workforce

The hosts explore the future implications of AI agents on employment and hiring practices. Kieran references Satya Nadella's vision:

Satya Nadella's Perspective [31:11]: "Employees might be evaluated based on the agents they utilize to enhance their roles."

Joe Mora agrees, highlighting the importance of agent proficiency in job performance:

Joe Mora [31:11]: "Using AI tools effectively reflects on a candidate's ability to leverage technology, crucial for modern engineering roles."

Dispelling the Hype vs. Reality

Concluding the episode, Joe Mora cautions against the exaggerated claims surrounding AI agents:

Joe Mora [35:43]: "There's a significant human-in-the-loop aspect that won't disappear this year. Agents aren't ready to autonomously take over entire workflows, especially for high-precision tasks."

He underscores the ongoing need for human oversight and the current technical challenges in integrating agents with existing systems:

Joe Mora [35:43]: "Implementing complex use cases requires clear code and instructions, which presents a substantial challenge beyond simple browser-based tasks."

Conclusion

The episode provides a balanced perspective on AI agents, celebrating their current capabilities while acknowledging their limitations. Joe Mora’s insights offer actionable strategies for businesses eager to adopt AI agents effectively. By starting with manageable use cases and progressively integrating more complex tasks, organizations can harness the true potential of AI agents without falling prey to the surrounding hype.

For those interested in exploring AI agents further, Crew AI offers a platform with templates and no-code options to inspire and facilitate the creation of custom agents tailored to specific business needs.

Notable Quotes:

• Joe Mora [02:50]: "The definition of an agent is it got to have agency. Unlike LLMs that predict content, agents can autonomously navigate tasks through reasoning."

• Kip Bodnar [05:52]: "Operator has a browser control agent that can make reservations and do research, but it's still slow and clunky compared to human performance."

• Joe Mora [27:02]: "Agents enrich customer data, create hypotheses on usage, and integrate structured data into CRM systems, enabling hyper-targeted email marketing."

• Joe Mora [35:43]: "There's a significant human-in-the-loop aspect that won't disappear this year. Agents aren't ready to autonomously take over entire workflows."

This comprehensive summary captures the essence of the podcast episode, highlighting key discussions, practical insights, and expert opinions on the state and future of AI agents in the business landscape.