I was on the ground today in New York City, attending the AI Agent Conference, taking the pulse of where the AI industry is heading, and gathering takeaways for Mainfactor and music/entertainment e-commerce. The overarching theme of Day 1 wasn’'t deep in the technical weeds; instead, it was highly strategic. The focus was on how to transform organizations to be truly "AI-native," navigating complex organizational shifts, and preparing for an entirely new era of operations, security, media, marketing, and commerce.

‍

Here are the highest-level insights and future predictions from the ground following day 1 of the conference. The following are notes from the vairous panels and sessions, combined with my takeaways.

‍

1. The Era of Embedded Workflows (and the Rapid Reactive Evolution of Traditional SaaS)

We are moving past reactive, one-off AI enablement. The future relies on seamlessly embedded workflows.Because building software is increasingly commoditized, the real difference-maker for businesses today is the ability to ship AI tools and products at scale without breaking them, and crucially, driving internal adoption.

‍

Software as a Service (SaaS), as we've known it, has changed forever. We are entering an era where companies build a lot of their own SaaS, and pay for what they cannot build. Furthermore, pricing models for the remaining valuable SaaS platforms will change drastically. The entire ecosystem could shift away from flat ARR and monthly set fees toward actual usage-based pricing driven by AI consumption.

‍

2. Unlearning SaaS: The "Porsche in the Driveway" Problem

Transitioning to an AI-native infrastructure requires aggressively unlearning the traditional SaaS mindset. There is a brutal reality of AI implementation: it is still a massive hurdle. Just like legacy software, simply buying a powerful AI tool doesn't magically solve your operational bottlenecks. Many enterprises are "buying the Porsche but never taking it out of the driveway." Integrating AI into your specific tech stack remains inherently difficult work. You cannot just purchase your way into being AI-native; you have to actively learn, build, and integrate.

‍

3. Organizational Redesign & "Training" the Machine

Upgrading individual capabilities with AI copilots only scales so far. True operational transformation requires top-down organizational redesign that empowers teams to deploy autonomous agents with humans in the loop. 

‍

4. The Absolute Prerequisite: Clean Data

However, that training is completely useless without clean data. A single, pristine source of truth is the foundational requirement for agentic success.If you unleash properly trained agents onto clean data, you can accomplish an unprecedented volume of work. Conversely, unleashing agents onto disparate data silos without clear guardrails will rapidly create cascading operational disasters. Clean your data first, or don't build the agents.

‍

5. Marketing's Return to Storytelling

This delegation of execution to agents is going to fundamentally change specific verticals, most notably marketing. AI is enabling marketers to return to their creative roots. Because agents can now handle the relentless, repeatable execution of campaign deployment and optimization, the premium is no longer on pushing buttons in ad platforms. Marketers are now more free to focus on high-level strategy and compelling storytelling.

‍

6. True AI-Enablement: Process Orchestration

When it comes to implementation, process orchestration is the most important element to become truly AI-enabled. You have to build foundational workflows before unleashing AI on complex operational processes. The typical enterprise relies on a varied mix of legacy software and APIs. True AI-enablement is a hybrid approach: blending automation, agentic decision-making, and replacing select legacy systems with newly AI-built tools where it makes sense.

‍

7. The Boardroom "Build vs. Buy" Debate

With tools like Claude Code and Perplexity, enterprises are heavily weighing whether they still need external SaaS. The primary hesitation to "building your own" is the "fear of God" regarding auditability and compliance. However, infrastructure around insurance and compliance for self-built tools will evolve rapidly to meet this new market demand.

‍

8. From ERP to ARP (Agentic Resource Planning)

We are entering the era of ARP: Agentic Resource Planning. Static models of work are out. Enterprises are being advised to return to a "best-of-breed" mentality, cobbling together the most effective ways to solve problems by connecting autonomous agents, internal AI software, and the best existing SaaS, but the glue between all of it is being AI-enabled, on top of a clean data set.

‍

9. Consumer-Facing Adoption and the "Human Premium"

Currently, AI adoption is predominantly internal. The next phase is full adoption for consumer-facing needs. As consumer-facing AI reaches saturation, we will see an explosion of "anti-AI" businesses. Companies that are purely human-powered will emerge, charging a unique premium for authentic human interaction.

‍

10. E-Commerce Platforms Have a Moat

Platforms like Shopify and other major POS systems have built an incredibly powerful moat that isn't going anywhere. They have optimized the backend infrastructure to a point where they are too far ahead to be easily displaced. Instead of disrupting them, AI simply amplifies their existing capabilities.

‍

11. The End of the Search/Social Duopoly

Historically, Google Search and Meta have functioned like the "government of e-commerce” in that you had to pay them to operate. That era is ending. As LLMs and agentic shopping evolve, platforms like ChatGPT, Gemini, and Claude will implement paid agentic search. Optimizing for and acquiring these agentic placements will quickly become more critical for merchants than traditional social media and search engine marketing.

‍

12. How Shopping Behavior is Evolving (And What it Means for Merchants)

The consumer journey is actively shifting from human-led search to agent-led delegation:

• Phase 1: Search (The Past/Present): "I look for products." Involves keyword searches. Merchants must Be Found FIRST via SEO.
• Phase 2: Intent (The Transition): "I describe what I need." Consumers use natural language via LLMs. Merchants must Be Relevant to a NEED via AEO (Answer Engine Optimization) and GEO.
• Phase 3: Delegation (The Next Wave): "I tell it what to do." The AI acts autonomously. Merchants must Be Selected by the AGENT. Brands will compete on hard, machine-readable data: verifiable user reviews, trust metrics, flawless shipping policies, and programmatic incentives.

‍

13. A New Commerce Ecosystem

This shift results in a completely new, AI-native ecosystem. It bridges Consumer Experiences and Merchants through a centralized Agentic Core that acts as the intelligent routing and reasoning layer, matching intent to action.

‍

14. The 90s File-Sharing Era of Agents: A Warning on Security

We are in a "Wild West" phase of AI adoption that mirrors the 1990s file-sharing era.

While frontier models like Perplexity Computer and Claude Cowork essentially offer the same capabilities as open-source tools like OpenClaw, they are vastly more secure and should be prioritized. To safely scale, organizations must enforce Access, Identity, and Governance. Provision agents with their own logins and strictly delegated access. Reduce permissions, reduce token usage, monitor relentlessly, and only then expand.

‍

15. The Expanding Agent Framework

Organizations must prepare for a multi-tiered framework of agentic identities:

• Your Own Agents: Personalized, individual-level assistants.
• Company Agents: Internal, department-level tools.
• Enterprise Agents: High-level orchestrators crossing silos.
• Consumer Agents: External-facing brand representatives, and also agents that consumers themselves deploy to shop.
• Agents' Agents: Specialized agents designed to transact autonomously with other agents, or to create other agents to oversee.

‍

16. Building the "Agentic Data Plane" and the Kill Switch

Enterprises require predictability, not open-ended discovery. You want agents operating within a fixed, governed set of tools. To achieve this, architectures must rely on a central Gateway that proxies models and tools, enforcing static rules and cost-controls. Most importantly, every action must be recorded sequentially in an immutable audit log. A separate "Grader" process must constantly consume this log asynchronously. If the Grader detects anomalous behavior, it triggers an automated Kill Switch, terminating access at the Gateway level.

‍

17. Stop Asking, Start Tasking

Perhaps the most actionable piece of advice from the late afternoon was a simple shift in mindset: Stop asking and start tasking. We need to move past using AI merely to ask questions about how to build things, and start building agentic workflows that actually execute the work.

‍

18. Paving the Roads for the Future

AI enablement is the equivalent of paving roads for cities before the roads existed. Right now, it feels messy, disruptive, and expensive. You have to break existing infrastructure to lay the new groundwork. But once the roads are paved, the velocity of commerce and operations changes forever.

‍

19. The Agentic Commerce Timeline

Just how fast is this happening? It took roughly 30 years for traditional e-commerce to capture 20% of total retail transactions. By 2030—just a few years from now—it is projected that 25% of all commerce retail transactions will be agentic or executed through AI. The runway to pave your roads is incredibly short.

‍

20. Conclusion: The AI-Native Mandate

If there is one cohesive narrative to take away from Day 1, it is that the "wait and see" era of AI is officially over. We have moved past using LLMs as parlor tricks or basic search assistants. The future belongs to organizations that treat AI not as a software overlay, but as a fundamental architectural and operational shift.

‍

To survive the incoming wave of agentic commerce and workflows, businesses must internalize these core directives:

• Embrace the Build: The traditional SaaS model is fading. Prepare your organization to build, govern, and host your own agentic tools.
• Fix the Foundation First: Clean your data, enforce strict identity governance, and build an "Agentic Data Plane" with asynchronous kill switches before you attempt to scale.
• Optimize for the Agent, Not the Human: The consumer journey is rapidly shifting toward delegation. If your brand isn't machine-readable and optimized for agentic discovery (AEO/GEO), you will become invisible to the bots making the buying decisions.
• Redesign Your Workforce: Stop asking AI for answers and start tasking it with execution. Redesign your operational workflows to manage autonomous systems with humans strictly in the loop.

‍

Paving these roads now will be difficult, but the timeline is unforgiving. The companies that do the hard work of true AI-enablement today will be the untouchable market leaders of 2030.

‍

Mainfactor is actively building and utilizing AI to become the leading AI-enabled e-commerce-as-a-service company. Please reach out to us if we can help you with your e-commerce or merchandise business.

‍

Day 1 Contributors & Panelists (Alphabetical):

• Ali Alkhafaji – CEO, Apply Digital
• Ameet Talwalkar – Chief Scientist, Datadog
• Anju Kambadur – Head of AI Engineering, Bloomberg
• Ankit Arya – Head of AI, Inscope
• Apoorv Agarwal – CEO, [Stealth Mode]
• Ariel Shulman – Chief Product Officer, Bright Data
• Arvind Jain – Founder and CEO, Glean
• Asaf Bord – GenAI Product Leader, [Fortune 500 Company]
• Atin Sanyal – Co-founder, CPO, Galileo
• Atul Tulshibagwale – Senior Director, Continuous Identity Strategy, CrowdStrike
• Bridget Shea – Chief Customer Officer, 1mind
• Bryan Tsao – Chief Product Officer, Jasper
• Cassie Young – General Partner, Primary Venture Partners
• Chad Tetreault – Chief Technology Office AI & Governance, Zscaler
• Chang She – CEO, LanceDB
• Dan Balaceanu – Co-Founder, SVP & Chief Product Officer, Druid AI
• Daniel Vassilev – Co-Founder and Co-CEO, Relevance AI
• David Shim – Co-Founder and CEO, Read AI
• David Treat – Global Chief Technology Officer, Pearson
• Deepak Shrivastava – CEO, Sunrise AI
• Deepak Turaga – Chief Technology Officer, Oden Technologies
• Derrick Choi – Codex Deployment Engineering Lead, OpenAI
• Dylan Munro – COO + Co-Founder, Spot & Tango
• Eric Skinner – VP Strategy, TrendAI
• Ethan Ding – CEO, TextQL
• Eve Psalti – Sr. Engineering Director, Microsoft
• Felipe Romano – AI Product Leader, PayPal
• Gillian Chin – Global Head of Data Artificial Intelligence, Bloomberg
• Humayun Sheikh – Founder and CEO, Fetch.ai
• Igor Faletski – Co-founder & CEO, Superpilot
• Ilan Kadar – Co-Founder & CEO, Plurai
• Irina Denisenko – CEO, Knox Systems, Inc.
• Jai Das – Co-founder, President, and Partner, Sapphire Ventures
• Jeffrey Norman – VP Low Carbon Ammonia, Air Products
• Jenna Flateman Posner – CEO, Chief Digital Agency
• Jim Nguyen – Co-Founder & CEO, InFlow
• Joe Moura – Co-Founder & CEO, CrewAI
• Jyoti Kunal Shah – Director of GenAI Applications Development, ADP
• Karun Appapogu – Head of AI Technologies Architecture - CAI, Vanguard
• Kathy Kam – Director of Engineering for Kiro, AWS
• Kaushik Ghosh – Staff Software Engineer, Intuit
• Kyle Lui – General Partner, Bling Capital
• Masha Sharma – VP Merchant Experience, Groupon
• Matt Glickman – Co-Founder, CEO, Genesis Computing
• Matt Stearns – Chief Technology Officer, FABco LLC
• Michael Sinoway – CEO, Lucidworks
• Michele Franceschini – Head of AI Engineering - Conversational Agents, Bloomberg
• Nuwan Bandara – Head of Technology, Fintech & Capital Markets, AWS
• Omer Cohen – CSO, Descope
• Peter Day – General Partner, super{set}
• Peter Yared – Founder and CEO, AgentCloak by InCountry
• Phoebe Liu – Reporter, Forbes
• Prittam Bagani – VP, Product Management, Chargebee
• Qingyun Wu – CEO, AG2ai
• Rachyl Jones – Tech reporter, Semafor
• Rama Krishna Raju Samantapudi – Sr. Staff AI/ML Architect, ServiceNow
• Ray Grady – Founder, B2B Advisor
• Rob Wisniewski – CTO - Credit and Insurance, Blackstone
• Robin Chiang – Chief Growth Officer, OpenTable
• Saahil Jain – CTO, You.com
• Sarah Hoffman – Director of AI Thought Leadership, AlphaSense
• Sarmad Qadri – Software Engineer, Meta
• Scott Keane – Operating Partner, Invictus Growth Partners
• Sean Neville – CEO and Co-Founder, Catena Labs
• Sean Roberts – VP, Applied AI, Netlify
• Sharon Goldman – AI reporter, Fortune
• Shipali Jangra – Director- Global Digital Product Management, American Express
• Simon Chan – General Partner, FirsthandVC
• Sonja Sierra – Director of Marketing, The Skin Center
• Stephen Sharr – VP Procurement & Logistics, Footprint
• Tim Sanders – Chief Innovation Officer, G2
• Tom Moor – Head of Engineering, Linear
• Tom Petit – Co-founder, Didero
• Tom Ronen – VP of Customer Success, Harvey
• Tomas Reimers – CPO and Co-founder, Graphite
• Vatsal Modi – Data Science Manager, DraftKings
• Venky Veeraraghavan – Chief Product Officer, DataRobot
• Vrushali Paunikar – Chief Product Officer, Carta
• Zhou Yu – Co-Founder, Arklex.ai

(Source data provided by agentconference.com)

‍