Why One-Third of Consumers Prefer AI Agents for Faster Service?

Welcome to AgentsX.AI!

Dear Operations Professionals,

In today's rapidly evolving business landscape, staying ahead of the curve is not just an advantage—it's a necessity. That's why we're thrilled to introduce AgentsX.AI, your new go-to resource for leveraging Generative AI agents to supercharge your operations.

In each issue, we'll dive deep into:

* Cutting-edge AI agent technologies and their practical applications

* Real-world case studies of successful AI implementation in operations

* Step-by-step guides for integrating AI agents into your workflow

* Expert insights and interviews with industry leaders

* Tips and tricks to maximize efficiency and productivity using AI

Whether you're just starting to explore the potential of AI or looking to optimize your existing systems, AgentsX.AI is here to guide you through the exciting world of Generative AI in operations.

Let's embark on this journey together and unlock the full potential of your operations with AI!

Stay ahead, stay efficient,

The AgentsX.AI Team

Why One-Third of Consumers Prefer AI Agents for Faster Service?

As the workplace evolves, companies are set to invest significantly in artificial intelligence (AI) agents. According to Gartner, agentic AI is positioned to be the most impactful strategic technology from 2025 onward. The firm predicts that by 2028, at least 15% of routine work decisions will be made autonomously by AI agents, a steep increase from 0% in 2024.

Agentic AI’s goal-oriented capabilities will bring about more flexible software systems capable of handling a wide range of tasks. However, Salesforce’s “State of the AI Connected Customer” survey, which gathered insights from 15,015 respondents across 18 countries, highlights potential challenges. Consumer trust in brands is at an all-time low, with AI raising expectations for trust and reliability.

The Growing Role of AI Agents

The Salesforce survey underscores some important trends, notably:

Trust in AI and Businesses

• 72% of consumers have less trust in companies than they did a year ago.

• 60% believe that advancements in AI make trust even more essential.

• 54% of AI users lack confidence in the data used to train these systems.

• 80% of consumers value customer experiences as much as the products themselves.

• 65% feel companies are careless with their data.

Consumers Expect Great Service Experiences

• 69% expect consistent interactions across different company departments.

• Over one-third say that difficulties like a cumbersome return process could lead to brand disloyalty.

• Consumers report spending as long as nine hours trying to resolve single issues with customer service.

• In the past six months, nearly one-third of customers ended service interactions without a resolution.

• Over half (54%) don’t mind how they interact with a brand, as long as their issues are promptly addressed.

• One-third would prefer to make purchases digitally rather than through a human.

• 34% would interact with an AI agent to avoid repeating information.

Trust and Transparency Drive AI Adoption

• 30% would choose to work with an AI agent for faster service.

• Fewer than a third of Gen Z and millennials are comfortable with AI agents making purchases on their behalf.

• 45% of consumers are more likely to use an AI agent if an escalation path to a human is provided.

• 44% prefer AI agents if the reasoning behind their responses is explained.

The “State of AI Connected Customer” report also notes that many customers are open to interacting with AI agents, provided they know they’re doing so and have a clear path to human support if needed.

Why Multi-Agent AI Solves Problems Beyond LLM Capabilities?

The introduction of ChatGPT has brought large language models (LLMs) into widespread use across both tech and non-tech sectors. This rapid adoption primarily stems from two factors:

1. LLMs as Knowledge Repositories: LLMs are trained on vast amounts of internet data and are updated periodically (e.g., GPT-3, GPT-3.5, GPT-4, GPT-4-turbo).

2. Emergent Abilities: As LLMs scale, they exhibit capabilities not found in smaller models.

Does this mean we’ve achieved human-level intelligence, or artificial general intelligence (AGI)? According to Gartner, AGI represents an AI that can understand, learn, and apply knowledge across diverse tasks and fields. However, the path to AGI remains lengthy, with one key challenge being the auto-regressive nature of LLMs, which predict words based on past sequences. As AI pioneer Yann LeCun notes, this approach can cause LLMs to deviate from accurate responses, exposing several limitations:

• Limited Knowledge: Despite vast training data, LLMs lack current world knowledge.

• Limited Reasoning: LLMs are effective at retrieving information but have constrained reasoning abilities, as noted by Subbarao Kambhampati.

• Static Nature: LLMs can’t access real-time information dynamically.

Overcoming these challenges requires a more advanced approach — this is where agents come in.

Agents to the Rescue

The concept of intelligent agents in AI has evolved significantly over the past two decades, adapting to new developments. Today, agents are discussed in the context of LLMs. An agent can be thought of as a multi-tool for addressing LLM challenges: it supports reasoning, can access up-to-date information from external sources (addressing the static nature of LLMs), and can autonomously complete tasks. Built on an LLM foundation, an agent generally consists of tools, memory, reasoning (or planning), and action components.

Components of an AI Agent

• Tools: Agents use external resources, such as the internet, databases, or APIs, to gather necessary data.

• Memory: Agents can store information temporarily (short-term memory) or retain it for longer durations (long-term memory).

• Reasoning: The reasoning component enables agents to break down complex tasks into simpler sub-tasks for effective processing.

• Actions: Agents perform actions based on the environment and their reasoning, iteratively solving tasks with feedback. ReAct is a popular approach for iterative reasoning and action-taking.

What Are Agents Good At?

Agents excel at complex tasks, particularly when employing a role-playing approach that leverages the LLM’s enhanced performance. For example, one agent might handle research while another focuses on writing when composing a blog. This multi-agent approach can address various real-world problems.

Role-playing helps agents concentrate on specific tasks, reducing hallucinations by clarifying the task through defined prompts. Structured frameworks like CrewAI formalize this process, supporting effective role-based agent functioning.

Multi-Agent vs. Single-Agent Systems

A single-agent system, such as a retrieval-augmented generation (RAG) approach, can be effective for domain-specific queries by accessing indexed documents. However, single-agent RAG has limitations, like retrieval performance or document ranking. A multi-agent RAG model addresses these limitations by employing specialized agents for different tasks, such as document understanding, retrieval, and ranking.

In a multi-agent setting, agents collaborate similarly to distributed computing patterns (e.g., sequential, centralized, decentralized, or shared message pools). Frameworks like CrewAI, Autogen, and langGraph+langChain enable complex problem-solving with multi-agent approaches. This article uses CrewAI as a reference framework for exploring autonomous workflow management.

Workflow Management: A Use Case for Multi-Agent Systems

In industries, workflows are essential, whether for loan processing, marketing campaign management, or DevOps. These workflows often involve sequential or cyclic steps that need expertise at each stage. In a multi-agent system using CrewAI, each workflow step is managed by a "crew" of agents. For instance, one agent might verify a user’s identity, while another checks financial details in a loan application.

Can a single crew manage all loan processing steps? While possible, this approach complicates memory management and risks goal deviation. Instead, it is more effective to treat each step as a separate crew, viewing the entire workflow as a sequence of crew nodes (using tools like langGraph) that operate in cycles or sequences.

Since LLMs are still developing, workflow management isn’t yet fully autonomous. Human oversight remains necessary for certain stages, such as verifying loan application results. Over time, some steps might become autonomous, but for now, AI-based workflow management serves an assistive role, easing repetitive tasks and reducing processing time.

Production Challenges

Bringing multi-agent systems into production presents challenges:

• Scale: As the number of agents increases, managing them becomes more complex. Frameworks like Llamaindex provide scalable, event-driven workflow management.

• Latency: Agents often face latency as tasks require multiple LLM calls. Self-hosted LLMs with GPU control can mitigate this, reducing reliance on slower, managed models.

• Performance and Hallucination: Due to the probabilistic nature of LLMs, agent responses may vary. Techniques like output templating (e.g., JSON formatting) and providing prompt examples can improve response consistency and reduce hallucinations.

As noted by Andrew Ng, agents represent the future of AI and will advance alongside LLMs. Multi-agent systems will continue to improve at processing multimodal data (text, images, video, audio) and handling more complex tasks. While AGI and fully autonomous systems are still aspirational, multi-agent frameworks help bridge the current gap between LLMs and AGI, making strides toward enhanced AI capabilities.

Spot AI Raises $31M and Introduces Video AI Agents to Revolutionize Surveillance

Spot AI, a company focused on AI-driven camera systems, announced it has raised $31 million in new equity funding, increasing its total funding to $93 million. Qualcomm Ventures joined as a new investor in this round, alongside existing supporters such as Scale Venture Partners, StepStone Group, Redpoint Ventures, and Bessemer Venture Partners, with additional contributions from GSBackers, MVP Ventures, and Cheyenne Ventures.

The company also introduced its Video AI Agents, marking a major advancement in applying agentic AI capabilities from digital environments to the physical world. Video AI Agents empower organizations to instantly detect and address incidents across safety, security, and operational domains, generating significant return on investment for sectors like manufacturing, education, retail, and automotive services. Traditionally, video surveillance was limited to passive recording, basic search functions, and manual monitoring. Video AI Agents, however, transform cameras into active problem solvers, capable of autonomously identifying issues and initiating responses without human intervention, including alerting stakeholders, providing analytics, activating lights and sounds, playing voice messages, and controlling machines in cases of incidents or injuries.

"Eighty percent of our perception of the world is through vision. That’s why we’re committed to helping our customers turn their video cameras into proactive teammates," stated Rish Gupta, CEO of Spot AI, who co-founded the company with Sud Bhatija and Tanuj Thapliyal. "Our AI Agents observe physical spaces, utilize advanced AI to interpret events, and take real-time actions. This funding enables us to push the limits of what’s possible and bring AI from the digital realm into physical environments where we live and work." Spot AI currently serves 1,000 clients across 17 industries, processing over twice the daily video volume uploaded to YouTube. The company makes video data as accessible and actionable as other types of data for industries facing safety hazards, operational inefficiencies, and productivity losses due to workforce shortages and security concerns.

Cornel Stewart, an engineer at industrial foam producer Elite Comfort Solutions, shared, "Spot AI transformed our surveillance from mere video recording to a video AI system that drives results. We’ve seen a decrease in injuries and incidents, and the system’s hazard detection has allowed us to address issues proactively. It’s like having an expert monitor our facility 24/7." Tushar Gupta, Senior Director of Qualcomm Ventures at Qualcomm Technologies, Inc., noted that Spot AI’s solutions enable businesses to rapidly derive insights from video feeds, converting them into valuable tools for enhancing safety, security, and daily operations. "This aligns closely with Qualcomm’s vision of using edge AI to tackle real-world challenges," Gupta commented.

Thank you for joining us on this exciting journey with AgentsX.AI. We hope this newsletter becomes an essential part of your toolkit as you navigate the ever-changing landscape of operations with the power of Generative AI. As you continue to explore new technologies, implement innovative strategies, and drive efficiency within your organization, remember that we’re here to support you every step of the way.

Stay connected with us for the latest insights, practical guides, and expert advice to ensure you stay ahead of the curve. Together, we can unlock new levels of productivity and success in your operations.

Until next time, keep pushing the boundaries of what's possible with AI!

Best regards,

The AgentsX.AI Team