Private AI for the Enterprise

St. Julien Hotel & Spa, Boulder

A private working session for a small group of board directors and senior executives examining the governance realities of AI inside modern enterprises. Through candid discussion and expert perspectives, participants explore fiduciary duty, legal exposure, financial oversight, and what true operational control of AI systems requires at the board level.

St. Julien Hotel & Spa  ·  Boulder, Colorado

Recognizing leaders advancing AI governance, innovation, and enterprise responsibility.

Conclude the day with drinks and networking, fostering connections and discussions sparked throughout the day.

Some of you were in the room yesterday afternoon. You heard the governance case — the fiduciary argument, the legal exposure, the question of who owns runtime authority when an AI system acts on its own at 2am. Today is the answer to the question that session was designed to raise: what does any of that actually look like inside a real enterprise system?
‍
For those joining us for the first time today: welcome. You're arriving at the right moment. The day turns on two irreversible shifts — AI autonomy and governance — and a third one most leadership teams aren't tracking yet: quantum sensing and computing, which are closer than the headlines suggest. The choices enterprises make in 2026 will determine whether they govern AI by 2027, or get governed by it.
‍
A rapid-fire P&L reality check follows: ten real, profitable AI use cases already driving measurable results across revenue, cost, fraud, compliance, and operations.

Jon Nordmark — CEO & Co-Founder, Iterate.ai · Co-Founder & Former CEO, eBags.com — scaled to $165M in annual revenue and acquired by Samsonite

Hans Peter frames AI not as just another a software upgrade, but as a societal reconfiguration — a supercharged cognitive revolution unfolding 10× faster with 10× the impact of the industrial revolution. What does it mean when AI increasingly runs society, from governments to corporations to schools and much more? What is our role in shaping that world?

Hans Peter Brondmo — former CEO, Everyday Robots, Google X, Start-up Advisor, Visiting Committee Member, MIT Media Lab

Autonomous vehicles are already operating on public roads. But when will they be everywhere — and what does it actually take to get there? Chris spent three years building the AI stack at Zoox, designing and training the multimodal language models that interpret driving behavior, handle long-tail scenarios, and support validation. He brings a rare, unvarnished view of what happens when AI systems move from prototype to persistent real-world operation — and what governance gaps look like when the cost of getting it wrong isn't a bad quarter, it's a life.

Chris Heckman, Ph.D. — Professor, University of Colorado · Former AI Stack Engineer, Zoox · Former Postdoctoral Fellow, Naval Research Laboratory

In January 2026, OpenClaw went viral. By February: 1.5 million agents created, 770,000 spawned in a single week, 1.49 million database records exposed, an $8 million crypto scam executed, and Cloudflare's stock moved 14%. This is not a hypothetical. A candid teardown of what commercial-scale agent swarms look like when governance is an afterthought.

Todd Sherman — Former CMO, Amplero · Established Amazon's Third-Party Marketplace

A practical look at how hybrid AI architecture can help governments and institutions scale AI without becoming fully dependent on cloud-first models. Based on an African government use case, this talk shows how AI-native devices, local edge execution, and selective cloud coordination can strengthen resilience, support public services, and create a more sovereign path to national digital acceleration.

Miguel Stief — Dominant Success, Global CEO, Executive Advisor, and Transformation Leader

Despite $30–40 billion in enterprise A.I. investment last year, 95% of organizations saw no measurable ROI. The problem isn't the technology. It isn't the architecture. It isn't even the governance framework. It's what A.I. memory CAN'T see. A.I. can capture what happens, but organizational greatness almost never lives in what happened. It lives in what didn't — the invisible acts of human restraint that no A.I. system, however sophisticated, can capture. In the rush to keep up, we've built systems that recall everything… and  forgot to ask what's actually worth remembering.

Josh Allan Dykstra — Optimistic Futurist & Keynote Speaker Hello Tomorrow, Board Member

Threat actors like Scattered Spider — thought to be roughly 1,000 loosely affiliated actors operating across the US and UK — are among the fastest adopters of AI on the planet. And they're not alone.

State-sponsored AI teams from Russia, China, North Korea, and Iran are now using frontier models at every stage of the attack cycle: reconnaissance, phishing, malware development, and data exfiltration. North Korea uses AI to synthesize intelligence on targets at defense companies. Iran uses it to augment reconnaissance and map business partner networks. China uses it to conduct vulnerability analysis and penetration testing planning against US targets.

These groups are deploying AI across the full attack surface. Social engineering is one method — cloning executive voices, manipulating help desks, automating identity takeover at scale. But it doesn't stop there. The same groups use agentic AI scripts to infiltrate code repositories and accelerate source code theft, run automated reconnaissance that maps internal networks and locates SOPs faster than any human analyst, exploit leaked credentials to move laterally before anyone knows they're inside, and — increasingly — use prompt injection to hijack AI agents operating inside enterprise systems, redirecting them to exfiltrate data or execute actions their owners never authorized. When your AI agent can be told what to do by the content it's reading, the attack surface isn't just your network. It's every document, email, and data feed your agents touch.

Sibito has been inside this fight. His team built the predictive fingerprinting system — developed in collaboration with the International Crime Investigation Authorities — that detected the behavioral signatures left by AI-automated attack operations. What he found, and what he'll discuss this morning: identity is now the perimeter, and AI has made every attack vector faster, cheaper, and harder to detect.

The question isn't whether your organization is a target. It's whether your defenses were built for the threat that exists today.

Sibito Morley — Co-Founder and President, Veromesh.ai · Former Chief Data Officer, Sinch (~1T of Western-world mobile message traffic flows through this network) · Former CTO, Lumen, Century Link, Davita

Private AI is not a product category. It is an architectural commitment spanning data, models, and hardware. Brian introduces the Three Rings framework and explains precisely why control, containment, and accountability collapse under public AI architectures — and what it takes to build systems where governance is structural, not performative.

Brian Sathianathan — Co-Founder & Chief Technology & AI Officer, Iterate.ai · Former Engineering Leader, Apple (Secret Products)

Quantum computing has promised breakthroughs on some of the hardest computational problems for years, but noisy hardware, complex programming, and difficult calibration have kept it largely in the hands of specialists. AI is starting to change that. Across the stack, large language models are helping with algorithm design and code generation, while machine learning is improving circuit optimization, hardware tuning, and noise mitigation. The pattern is becoming clear: AI will not compete with quantum computing. It will make it usable. Rob Wamsley explains where that shift is already happening in practice, where the limits still are, and what this emerging model of human direction, AI assistance, and selective quantum execution means for enterprise teams thinking about next-generation compute.

Robert Wamsley, Ph.D. — Physicist & Quantum Algorithm Researcher, Quantum Rings · Building tools to simulate quantum futures before the hardware arrives

Quantum isn't only about computing — it's about sensing. Eva's company Flari uses quantum-grade optical frequency comb technology to detect disease biomarkers in breath with a precision that classical instruments can't match. The catch: the data these sensors generate is so complex that only AI can interpret it at scale. Eva will show how AI and quantum sensing are already converging in healthcare — and how DARPA-funded research is accelerating the path from lab to clinical deployment. This is what it looks like when quantum comes to life in the real world.

Eva Yao, Ph.D. — Founder & CEO, Flari · DARPA-Funded Quantum Molecular Sensing for Biomedical Research and Disease Detection

DARPA reviewed 18 of the world's most advanced quantum computing companies for its Quantum Benchmarking Initiative — a rigorous program designed to determine whether fault-tolerant, utility-scale quantum computers can be built by 2033. Only 11 survived to Stage B. Atom Computing, headquartered in Boulder, is one of them — alongside IBM, IonQ, and Quantinuum. In October 2025, Jensen Huang unveiled NVQLink — a new architecture directly connecting NVIDIA's GPU supercomputers to quantum processors — and Atom Computing was named one of its 17 founding hardware partners. NVIDIA's bet is straightforward: GPUs won't be replaced by quantum computers. They'll be the engine that runs them. Justin explains what this convergence means and what the AI-quantum hybrid era will actually look like for enterprise compute.

Justin Ging — Chief Product Officer, Atom Computing · One of 11 Companies DARPA Selected as a Plausible Path to a Fault-Tolerant Quantum Computer · NVIDIA NVQLink Founding Partner · MIT Leaders for Manufacturing Fellow

Traditional security platforms were designed to protect networks, applications, and endpoints. They were not designed for AI systems that accumulate memory, call external tools, spawn subagents, and operate across enterprise infrastructure with minimal human oversight. The attack surface is different. The failure modes are different. The defense has to be different.

Rigor.ai has built a mathematically rigorous cyberdefense management platform — one designed to identify, verify, and remediate every known attack vector that matters, before exploitation occurs. Not reactive. Not probabilistic. Mathematically complete. This live demo shows what that looks like against an AI system operating in a real enterprise environment: the vulnerabilities that standard tools miss, the attack paths that autonomous agents open, and what preemptive remediation actually looks like in practice.

Jai Desai — Head of Business Development and Sales Rigor.ai, Founding Member, Strategic Advisor

The cloud-first era is not ending. It is maturing. Sixty-nine percent of enterprises are now actively considering moving workloads back from public to private cloud — and more than a third have already done so. AI is accelerating that shift. When an AI agent operates continuously across public infrastructure, it creates latency, sovereignty, and observability problems that no policy document can fix. Justen makes the case that up to 50% of AI workloads will ultimately run in private or sovereign environments — not because of ideology, but because of physics, compliance, and control. He introduces AgentWatch, a real-time agent observability layer, and shows what it actually looks like to see, track, and govern AI agents across a distributed enterprise infrastructure — before something goes wrong.

Justen Aguillon — Director, Technology Partner Ecosystem, Equinix · Architect of the Fabric Intelligence vision connecting AI providers and enterprise subscribers across 270+ global data centers

Avoiding the Strategy Trap: How leaders create the conditions for AI agents to deliver real value.

Many organizations struggle to turn AI pilots into real operational impact.

In this talk, Kevin Ertell, author of The Strategy Trap, explains why the challenge is rarely the technology itself. It’s the operating conditions around it.

Drawing on his Six Cs system — Co-Creation, Clarity, Capacity, Communication, Coordination, and Coaching — Kevin shows how leaders can design organizations where AI agents actually improve performance instead of becoming another abandoned experiment.

This session provides a practical lens for leaders responsible for turning AI from an idea into consistent execution.

Kevin Ertell — Author of The Strategy Trap: Why Companies Fail at Execution, CEO Mistere Advisory, former Global VP Retail Operations at Nike.

Agents are proliferating inside enterprises faster than leadership can track. The hardest part isn't choosing a model—it's controlling what agents remember, what they're allowed to do with that memory, and how quickly bad state spreads across tools, systems, and people. Where do real-world deployments actually fail? What guardrails work at runtime—not just on paper?

Brian Sathianathan — Co-Founder & Chief Technology & AI Officer, Iterate.ai

Sibito Morley — Co-Founder and President, Veromesh.ai · Former Chief Data Officer, Sinch (~1T of Western-world mobile message traffic flows through this network) · Former SVP, Lumen Technologies and CenturyLink

Hans Peter Brondmo — former CEO, Everyday Robots, Google X, Start-up Advisor, Visiting Committee Member, MIT Media Lab

The conversation around Artificial General Intelligence (AGI) is often dominated by scale: larger models, more data, and increasing compute. But in real-world systems, scale alone is not enough. This talk presents a grounded, systems-oriented perspective on building intelligent systems that actually work in production. It explores how smaller, specialized language models, when combined with structured knowledge bases and coordinated multi-agent architectures, can deliver reliable, explainable, and scalable intelligence. Drawing from real-world deployments across domains such as agriculture and enterprise systems, Srikanth will unpack: (a) Why smaller, domain-adapted models often outperform general-purpose models in constrained environments (b) How knowledge graphs anchor reasoning, improve traceability, and reduce hallucination (c) The role of multi-agent systems in orchestrating complex, real-world decision-making (d) What it truly takes to move from proof-of-concept (POC) AI to production-grade systems. Rather than asking “How do we build AGI?”, this session reframes the question: “How do we systematically assemble intelligence from modular, reliable components?” For leaders and practitioners, this represents a shift from chasing trends to designing systems that deliver consistent, measurable value.

Dr. Srikanth Thudumu — Institute of Applied Artificial Intelligence and Robotics (IAAIR)

This is what vibe coding looks like at its purest: no budget, no team, no marketing plan — just a teenager, agentic tools, and an idea. Highlandrr built Not Cute Anymore Tower on Roblox in a single day using AI-assisted development, then watched 18.9 million people play it in three weeks. Along the way he built a community of 1.3 million Roblox players in under two months — without a marketing department, a growth team, or a dollar of paid acquisition. No enterprise process. No governance framework. No sprint planning. The question this raises for every organization in the room: if a 17-year-old can build and ship a product used by millions in 24 hours, what does that mean for who builds what next — and what happens when your competitors figure that out before you do?

Jon Nordmark — CEO & Co-Founder, Iterate.ai · Co-Founder & Former CEO, eBags.com — scaled to $165M in annual revenue and acquired by Samsonite

Highlandrr had no guardrails. That's fine for a Roblox game. It is not fine for a hospital, a financial services firm, or a public company. This session is the enterprise answer to the question the previous demo raised: how do you move fast with AI without losing control of what it builds? Blake shows what responsible AI development actually looks like from the first line — scope, guardrails, and governance defined before a single line of code is written. Most enterprises skip this step entirely. This demo shows exactly what it costs them, and what it looks like when you get it right.

Blake Stenstrom — AI Engineer, Iterate.ai · Princeton-Trained Algorithms Expert

Agriculture generates enormous amounts of data — soil conditions, weather patterns, yield histories, supply chain timing, equipment telemetry — and almost none of it has been accessible in a form that drives real decisions. This demo shows how private AI is being applied to that problem at the World Food Bank: turning ground-level data into forecasting, resource allocation, and operational decisions that directly affect food security outcomes. The governance requirements are different from enterprise IT — the infrastructure is distributed, the connectivity is unreliable, and the stakes are measured not in revenue but in whether communities eat. Richard brings a perspective that no other presenter today carries: what it looks like to deploy AI in environments where the margin for error is human, not financial.

Richard Lackey — CEO & Chairman, World Food Bank

Jason Fisher — Founder and Executive Director, Institute of Applied Artificial Intelligence and Robotics (IAAIR)

What happens when brand engagement becomes semi-autonomous — but still brand-safe, measurable, and governed? This demo showcases the AI-driven engagement engine built for large brands, designed to monitor, respond, and amplify engagement across TikTok, Instagram, and Facebook in real time. Unlike simple chatbots or auto-replies, this system ingests high-velocity social signals, understands context and brand voice, generates compliant responses, escalates when human intervention is required, and tracks performance and sentiment impact end to end. This is not AI posting randomly. It is instrumented, observable, constrained engagement at scale — and a live example of what governed autonomy looks like in a consumer-facing environment.

Magnus Tagtstrom — Corporate VP, Iterate.ai · Former Global VP of Innovation, Circle K - Couche-Tard (16,000 stores)

AI systems don't store information the way IT systems do. They accumulate state through working memory, retrieval, tool outputs, and long-lived context — each with different persistence, cost, and risk profiles. This session explains where memory actually lives, how it decays, and why it behaves nothing like a database.

John Selvadurai, Ph.D. — VP R&D, Iterate.ai · Former SAP Architect

Search is changing. Discovery is moving inside AI interfaces. And commerce is following. This demo shows what happens when product information, pricing logic, availability, and checkout workflows are inserted directly into ChatGPT and Gemini — enabling consumers to discover, evaluate, and complete purchases entirely inside a conversational interface. It's early, but the parallel is worth taking seriously: this may be what messaging-based commerce in China looked like before WeChat became a super app. Using structured commerce protocols — UCP and ACP — the demo walks through product discovery inside AI, inventory verification, dynamic pricing logic, secure checkout orchestration, and transaction completion without a single redirect. The result is commerce without a website — and a preview of where consumer expectations are heading.

Tim McCue — SVP Global Operations, Jockey International

These aren't prototypes. They're production agents running inside real organizations today — observable, controllable, and fully private. A Clinical Documentation Assistant accelerating a 150-doctor radiology practice without touching PHI. A Contract and Compliance Reviewer that flags risk and produces audit-ready summaries. An Operational Workflow Optimizer that automates internal processes with human escalation built in. An Executive Intelligence Assistant that delivers role-based summaries with controlled memory. Dr. Dan Reed presents the clinical reality — what it actually means to deploy AI in a regulated medical environment where a mistake isn't a bad quarter, it's a patient. The closing question: where does this actually live? The IBM presenter connects the demo to infrastructure reality — why private AI is a physical architecture decision and not just a policy choice

Dr. Dan Reed — Radiology Oncologist · Owner, Radiology Oncology Practice · 150-Physician Practice

Most enterprises are sitting on decades of accumulated data — contracts, clinical records, financial documents, compliance archives — that has been completely inaccessible to the organization that owns it. Not misplaced. Not lost. Stored, paid for, and invisible. IBM and Deloitte both put the figure at roughly 90% of enterprise data falling into this category — unstructured, unanalyzed, and until now, unreachable. NetApp's intelligent storage platform changes that. At approximately $200,000, it brings a generative AI interface directly to on-premises mass storage — purpose-built for hospitals, financial services firms, and SLED organizations where data sovereignty and compliance are non-negotiable. This demo shows what it looks like when an enterprise finally turns its dark data into a working asset: private, governed, auditable, and without a single byte leaving the building.

Jeff Liborio — Director, Global Strategic Alliances, AI/Technology Partners, NetApp

Most AI architecture assumes something that retail environments can't always guarantee: a reliable connection. Point-of-sale systems in basements, pop-up locations, outdoor markets, festivals, and remote stores can't wait for a signal — and they can't afford to fail when one isn't available. This demo shows what AI looks like when it runs fully offline on constrained hardware, with no cloud dependency, no latency from a round trip to a data center, and no single point of failure tied to connectivity. The enabling hardware is Qualcomm's Snapdragon 6490 chip — a platform Iterate.ai has built on specifically to bring inference to the edge without sacrificing governance or control. The architecture handles real retail workloads: inventory awareness, customer interaction, transaction processing, and exception handling — all running locally, all observable, all stoppable. For retail operators managing hundreds or thousands of locations, this isn't a future capability. It's a deployment decision available right now.

Ynjiun Paul Wang, Ph.D. — SVP, USI · Inventor of the 2D Barcode, Found on Every Driver's License in America

Most document AI fails the moment it meets the enterprise. Contracts, compliance documents, clinical records, and classified files don't belong in a public model — and in many industries, sending them there isn't just a bad idea, it's a regulatory violation. Private AI is the answer, but most implementations get it wrong: slow retrieval, poor accuracy, no audit trail, and architectures that quietly phone home when no one is watching. This demo shows what actually works — intelligent data routing that keeps sensitive documents inside the perimeter, accurate extraction and summarization without cloud dependency, and governance built into every step of the pipeline. The use cases span regulated industries where air-gapped deployment isn't optional: legal, clinical, financial, and government environments where the data never leaves the building and the system still has to perform.

David Richard — Director, Digital Strategy & AI Automation, Terralogic · Specialist in Regulated Data and Private AI Deployments

Every governance conversation focuses on models, prompts, and agents. Almost none of them focus on the network layer that those agents run across — which is where latency, data sovereignty, and observability actually break down at scale. This demo goes there. Justen shows how to pause, redirect, and terminate an AI agent in real time based on policy, cost, or security constraints. Not after the fact. Mid-action. This is what "human in the loop" looks like when it's built into the architecture rather than written into a policy document. The infrastructure making it possible: Equinix's Private Service Exchange — currently in beta — a governed private fabric connecting AI inference providers and enterprise subscribers across 270+ data centers in 77 markets. Fabric Intelligence adds the control plane on top: automated routing decisions, live telemetry, and dynamic segmentation. KPIs from real deployments include reduced time-to-inference, fraud detection acceleration, and complete elimination of public internet exposure for AI workloads. The demo is live. The agent is real. The pause button works.

Justen Aguillon — Director, Technology Partner Ecosystem, Equinix · Architect of the Fabric Intelligence vision connecting AI providers and enterprise subscribers across 270+ global data centers

Rob Taylor — Enterprise Technology Attorney, Carstens Allen & Gourley LLP · Advising companies on AI governance, data liability, and emerging technology risk

Most AI governance conversations use hypothetical stakes. This one doesn't. Tarian manages physical security operations across thousands of locations — and when an AI system makes a scheduling error, a coverage gap doesn't mean a missed deadline. It means an unprotected facility, a delayed response, or a person in the wrong place at the wrong time. This demo shows how private AI optimizes workforce scheduling, coverage gaps, and response times across 7,000 security personnel in real time — handling the complexity that no spreadsheet or rules-based system can manage at this scale. The governance requirements here are not theoretical: every autonomous decision the system makes has to be observable, auditable, and stoppable. Mark walks through what that looks like in practice — how the system earns trust, where human override is built in, and what the operational impact has been since deployment.

Mark Stephens — SVP IT & BI, The Tarian Group

Agriculture generates enormous amounts of data — soil conditions, weather patterns, yield histories, supply chain timing, equipment telemetry — and almost none of it has been accessible in a form that drives real decisions. This demo shows how private AI is being applied to that problem at the World Food Bank: turning ground-level data into forecasting, resource allocation, and operational decisions that directly affect food security outcomes. The governance requirements are different from enterprise IT — the infrastructure is distributed, the connectivity is unreliable, and the stakes are measured not in revenue but in whether communities eat. Richard brings a perspective that no other presenter today carries: what it looks like to deploy AI in environments where the margin for error is human, not financial.

Richard Lackey — CEO & Chairman, World Food Bank

Every demo today has shown AI working inside well-resourced organizations — hospitals, retailers, security firms, global data centers. This one is different. Community hubs serve people who need help navigating benefits, housing, employment, and services — and they do it with thin budgets, volunteer staff, and no dedicated IT infrastructure. This demo shows what responsible AI deployment looks like when the user isn't a knowledge worker with a corporate laptop — it's a parent trying to find childcare, a veteran navigating benefits, or a senior who needs help understanding a medical bill. The governance requirements here are the most human of any demo in this act: no data leaves the community, no decision is made without a human in the loop, and the system has to work for people who didn't choose to interact with AI. Randy shows what that looks like when you get it right.

Randy Kohn — President, AppBrew & Adventure Mind

This is not a summary. The day has already made the case. This is the moment where six people who have spent their careers making consequential decisions at scale tell you what they are taking out of this room — and what they think you should do before you leave.Each presenter shares their five takeaways and why they matter. Not talking points. Not consensus. The unfiltered view of executives who have lived the consequences of getting technology strategy wrong — and right

Frank Kollmar — Former Global Deputy Managing Director, L'Oréal Dermatological Beauty Division · Former President & CEO, L'Oréal Canada

Chris Smith — EVP International & Chief Customer Officer, Jockey International

Brian Tilzer — Former EVP, Chief Digital, Analytics & Technology Officer, Best Buy · Former SVP, CVS and Staples · Board Member, Signet Jewelers

Hans Peter Brondmo — former CEO, Everyday Robots, Google X, Start-up Advisor, Visiting Committee Member, MIT Media Lab
‍
Lynda Pak — Former SVP Technology, Estée Lauder

Diane Randolph — Board Director, Dollar Tree (NASDAQ: DLTR) · Board Director, Shoe Carnival (NASDAQ: SCVL) · Former CIO, Ulta Beauty

‍Sibito Morley — Co-Founder and President, Veromesh.ai · Former Chief Data Officer, Sinch (~1T of Western-world mobile message traffic flows through this network) · Former CTO, Lumen, Century Link, Davita