AI agents and what they mean for the enterprise [Q&A]
Artificial intelligence is creeping into more and more areas of business and consequently we’re seeing it used for ever more complex work.
We spoke to Trey Doig, CTO and co-founder at Echo AI, to find out more about AI agents and how they can help solve problems and carry out tasks.
BN: There is a lot of talk about 'AI agents' at the moment, can you break through the hype and tell us what they are?
TD: AI agents fundamentally, utilize Large Language Models (LLMs) to orchestrate the execution of complex, multi-step tasks. By linking the outputs of one LLM prompt to another, they navigate problems or perform tasks that are unattainable within the confines of a single prompt.
This approach to utilizing LLMs autonomously, coupled with the appropriate application software, not only paves the way for automation of tasks traditionally done manually but also dramatically increases the volume of work that can be generated.
This method enhances the efficiency and effectiveness of these tasks, enabling AI Agents to tackle a broader array of challenges with precision and adaptability.
Industries ranging from customer service to content creation are already experiencing a transformative shift with agent technology, where tasks that once required extensive human intervention can now be executed swiftly and at scale, thereby unlocking new possibilities for innovation and productivity.
BN: How are they being used in a real business setting? What are some of the top industries using them?
TD: I think AI agents as a general paradigm for most process automation may still be a ways off, but that’s not to say that we aren't already seeing its potential and substantial savings to real businesses.
Everything from GTM research, to data analytics, to personal coding assistants are coming to market everyday. Naturally, the tech industry is leading the way in adoption, however lots of growth can be seen happening in contact centers, retail, and D2C brands where AI Agents are achieving deeper levels of customer intelligence than ever before.
Contacts are a prime example where agents are making a dent. In this setting, AI agents are enabling a deeper understanding of customer needs and preferences, leading to more personalized and effective customer service.
BN: 'Agents as analysts' seems to be an interesting concept. Are they better than humans at analysis?
TD: For most analysis, especially the sort that can be done in a vacuum, LLMs have already achieved human-levels of accuracy and are able to produce analysis several orders of magnitude greater than manual processes.
Even for highly unique domain specific types of analysis, LLMs have demonstrated a remarkable ability to be fine-tuned for those exact use cases. Fine-tuning models that are pre-trained on documents for a specific domain such as legal materials, coding, or customer conversations, questions previously thought too big to answer or nebulous in nature, are now being understood.
BN: What are their limitations?
TD: Aside from cost, and the common need for complicated fine-tuning processes, LLMs and AI Agents are limited mostly by the scale in which they can be deployed for a task. A company seeking to analyze 100 percent of their customer conversations for instance, might require thousands of agents to analyze each of their conversations in a single day.
For real time insights, each agent would then require access to LLMs performant enough to stay at pace with the rate of incoming data. Without this high availability of LLM inference, AI Agents can be drastically encumbered in their ability to deliver results in a timely manner.
There are other challenges worth mentioning as well. For example, there are many models out there today, making it challenging to select the right model for specific needs due to varying performance and quality. Also, the lack of memory in current models poses privacy benefits but challenges continuity in applications like call centers, requiring comprehensive inputs for each interaction.
Context window sizes and security concerns, such as 'hallucinations' and 'jailbreaking', are also hurdles companies are dealing with.
None of this is insurmountable, but the technology is nascent and, in some ways, we're all 'building the ship as we go.'
BN: How is Echo AI making agents useful for business?
TD: Echo AI has built a platform from the ground up to leverage this powerful nature of AI Agents so that for the first time ever, companies of all sizes are able to finally understand what their customers are telling in everyday conversations. This gold mine of information and data has been previously untouchable purely through the time and effort required to extract meaning, but now with the power of AI Agents and LLMs, those invaluable insights are now able to be understood.
To break it down, Echo AI has two primary components. The first is 'Pathlight Conversation Intelligence' (CI), which marks a significant advancement in the field of conversational analysis.
CI harnesses the most advanced large language models (LLMs) to provide human-level analysis of customer interactions at an unprecedented scale and speed. CI overcomes the limitations of traditional conversational analysis tools, enabling organizations to gain deep and actionable insights from customer conversations.
By capturing, transcribing, tagging, and categorizing every conversation across all channels, Pathlight CI offers businesses the ability to uncover insights, track trends in real-time, and deliver personalized guidance. The result is that customer-facing teams can more quickly address issues and also provide a comprehensive understanding of customer needs and emerging trends, facilitating informed decision-making and enhancing service quality.
The second is 'Insight Streams', which uses Generative Agent technology to analyze millions of customer conversations, transforming them into valuable business insights and trends. Insight Streams acts like 'autonomous analysts', providing executives with real-time, comprehensive overviews of customer interactions, requiring minimal setup and offering customizable streams for specific insights.
Insight Streams ultimately bridges the gap between leadership and customers, allowing businesses to quickly respond to issues, opportunities and customer feedback.
Image credit: phonlamai/depositphotos.com