Alan Bekker, CEO of eSelf.ai and a former Head of Conversational AI at Snap, teams up with Jozef Soja, an ARK Invest analyst focused on enterprise AI. They explore the rapid rise of AI agents in customer service and their potential to revolutionize education. Alan shares insights on how large language models are transforming learning, emphasizing the importance of an engaging visual presence in AI tutoring. They also discuss evolving pricing models for AI services and the impending agent-versus-agent competition in the marketplace.
55:31
forum Ask episode
web_stories AI Snips
view_agenda Chapters
auto_awesome Transcript
info_circle Episode notes
insights INSIGHT
Custom AI Apps Disrupt SaaS
Software layers below SaaS, like Platform as a Service, gain importance as AI democratizes app-building.
Custom AI apps become cheaper and more aligned to specific enterprise needs than traditional SaaS options.
question_answer ANECDOTE
Alan Bekker's AI Journey
Alan Bekker founded Voca, providing voice AI agents for call centers including American Express and AT&T.
Voca was acquired by Snapchat, where Bekker led conversational AI during the emergence of LLMs.
question_answer ANECDOTE
From Graphs to GPT Models
Early conversational AI was graph-based with decision trees tailored per customer.
OpenAI invited Snap to be a design partner for GPT models in 2021, opening a new era for conversational AI.
Get the Snipd Podcast app to discover more snips from this episode
Brett Winton and ARK analyst Jozef Soja dive deep into the rapidly evolving world of AI agents—software entities that are increasingly automating enterprise functions like customer support. They explore why AI agents are gaining traction, how they’re priced, and the potential for a new kind of agent-versus-agent arms race between companies and consumers. Later in the episode, they’re joined by Dr. Alan Bekker, founder of eSelf.ai and former Head of Conversational AI at Snap, who shares his journey from building voice agents for call centers to launching a real-time, face-to-face AI tutoring platform. Alan offers insights into how the rise of large language models (LLM) is reshaping education, what makes a great AI tutor, and why a visual, embodied presence is crucial for learning.
Key Points From This Episode:
00:00:00 What enterprise AI agents actually do and how companies like Salesforce are pricing them
00:03:41 Why $2 per AI conversation may already undercut human support costs
00:05:04 The Return On Investment (ROI) model behind agent adoption and enterprise productivity
00:06:41 Why agent-based software may retain higher pricing power than other AI tools
00:09:11 The coming arms race: AI agents negotiating with other AI agents
00:12:30 Scaling demand for customer service with intelligent automation
00:15:04 Vertical vs. horizontal Software as a Service (SaaS) in the AI agent ecosystem
00:16:43 AI’s impact across the software stack—SaaS, Platform as a Service (PaaS) , and Infrastructure as a Service (IaaS)
00:17:56 Why building your own AI apps may soon be cheaper than onboarding SaaS
00:20:01 ARK's internal hackathon and how non-engineers are becoming developers
00:20:29 Guest: Dr. Alan Bekker joins to discuss the evolution of conversational AI
00:22:04 The journey from decision trees to LLMs: Lessons from Snap’s AI team
00:27:32 Seeing GPT’s impact from inside: OpenAI’s early partner outreach
00:31:47 Why face-to-face AI tutors found strong product-market fit in education
00:33:59 eSelf’s go-to-market strategy: Partnering with publishers as a business to business to consumer (B2B2C) wedge
00:36:24 Pricing real-time AI tutoring tools in a margin-conscious market
00:40:00 Business to consumer (B2C) aspirations: Moving toward a direct-to-student tutoring product
00:44:56 What’s still missing for real-time AI to match human-level teaching
00:48:03 The psychological impact of avatars: Building trust through embodied agents
00:51:43 Why personalization—not just LLM knowledge—matters in tutoring
00:54:20 Democratizing learning: LLMs as the end of expert-driven education