Enterprise AI Summit — Charlotte, NC · Oct 7–8, 2026

An urgent threat to a flagship product pushed NRC Health to rebuild its core platform with AI. The harder question came afterward: how could the business keep that speed without another emergency? Dustin Warner describes bringing sales, service, operations, product, and technology into a standing collaboration organized around shared product goals. He shares how working prototypes shortened decisions, how managers’ responsibilities shifted toward removing delays and building alongside their teams, and why margin, growth, and learning became more useful measures than delivery alone. Attendees will hear both the practices NRC Health kept from the crisis and the unresolved challenges of sustaining voluntary participation, redefining product roles, and accelerating customer adoption.

As AI accelerates software development, financial services firms must increase delivery speed while maintaining control and accountability. Khalid Elsawaf describes how Morgan Stanley is evolving its software development controls by breaking manual reviews into focused, reusable agent skills, bringing those checks into developers’ everyday tools, and generating evidence alongside the work. Drawing on architecture alignment, test quality, specification coverage, and infrastructure review, he explains how standardized development practices make controls easier to automate and how teams can question whether existing gates still address meaningful risks. Attendees will learn an approach to moving controls earlier in delivery while preserving human judgment for exceptions and consequential decisions.

What does it take for a CEO’s personal AI breakthrough to change how an entire organization works? Gene Kim sits down with Hyland CEO Jitesh Ghai to explore why he built Ziggy, a chief of staff agent using Claude Code, and what that experience revealed about AI’s relevance to his own role. They discuss why Jitesh devoted executive offsite time to giving his leaders their own hands-on breakthroughs, what he asked them to do next with their teams, and how leaders can turn experimentation into changes in daily work. The conversation tackles a concrete challenge facing technology leaders: when engineering accelerates but spends its time waiting on the rest of the business, how can leaders reshape the organization so that speed advances its most important goals?

The familiar practices of the Toyota Production System rest on layers of technology and specialized knowledge that visitors rarely see. Drawing on his experience in Toyota’s engine manufacturing operations, Art Smalley traces the machinery, controls, tooling, and materials expertise beneath the visible management system, then connects that history to today’s opportunity for domain experts to build small tools with AI. He examines work that is too specific for a vendor product and too small for a central IT project, but still needs a safe path into use. Attendees will explore how appropriate platforms, risk-based review, building skills, and clear ownership can help people closest to the work create and maintain useful applications.

Faster code generation makes it more important to know whether software will perform reliably in production. Christina Yakomin shares Vanguard’s shift from counting resiliency tests to asking whether those tests are valid, meet their goals, and cover the changes being released. She describes an AI-assisted workflow that generates tests from API specifications, configures them using production behavior, evaluates results against explicit intent, and iterates on scripts and test parameters while surfacing application changes for consideration. Attendees will learn how this approach distributes SRE expertise across more teams and how today’s resiliency automation informs a broader vision for continuously assessing production readiness.

What happens when AI makes it easier to cross the boundaries that once defined software expertise? Hilton’s Matt Klosterman and Jeremy Kooistra share how their teams moved from experimenting with coding agents to rethinking who can contribute to software delivery. Through hands-on workshops, shared tooling, and workspaces equipped with relevant context, they helped developers work across native iOS and Android platforms and opened new paths for product leaders and business colleagues to turn ideas into working changes. Their experience builds on years of architectural alignment and team design, showing why tools alone do not create a new operating model. The speakers trace the shift from platform specialization toward engineering judgment, context, and coordination. They share evidence of broader cross-platform contribution, explain how a regional APAC team brought a Japan-specific customer experience into production, and describe the roles of business teams, agents, and development enablers. Attendees will learn how access, structure, and permission help people begin; why shared capabilities and targeted engineering support matter; and how to broaden participation while reinforcing quality and the fundamentals of software delivery.

AI-driven code generation, Jeremy Eder argues, is not economically viable without discipline: token spending can outrun budgets before anyone notices. Eder, a Distinguished Engineer at Red Hat, describes the token budgets per engineer, centralized AI gateway, and factory model used there, and defines value per token as business value delivered divided by tokens spent. He explains how priority should come from systems of record such as the CRM, issue tracker, and security feed rather than from guesswork, and walks through an illustrative queue in which business rules and a tighter budget reorder three requests. He also shows why a model's price per token does not equal its cost per job, since retries and failed runs count. Attendees will learn how to rank work, route it to the cheapest model that clears the evaluation bar, and which roles, such as curator and trajectory miner, nobody yet owns.

AI has collapsed the cost of writing code — and moved the bottleneck. Drawing on flow metrics from hundreds of enterprise value streams, Mik shows where AI-assisted delivery actually speeds organizations up, where it silently piles up wait states, and what the Flow Framework reveals about running a product operating model in the AI-native era.

Thrive Market's Claude rollout began in February with a license and no structure. Michael Winslow and Marya Caples present two experience reports from it. Marya, a food safety manager who is not an engineer, describes building the Ingredient Vetting Portal on tools the company already licensed, replacing an email-and-spreadsheet review process. Michael describes the company-wide rollout: department Dojos that start with building and discarding a first version, a Science Fair, and an AI app marketplace. He then covers what changed once non-engineers could build, including questions of ownership, governance, and the gap between building software and operating it. Finally, he explains why engineering adoption in existing codebases was a different problem, and how an AI Guild took on measuring outcomes and standardizing tools. Attendees will hear the open questions the speakers bring to the audience.

Most AI use in the enterprise stops at efficiency: summarizing, drafting, and polishing. Owen Parker and Kurtis Cochran argue that this is table stakes, and that the greater value lies in rebuilding core processes and in greenfield systems. They describe what Arkana Laboratories, a 250-person kidney pathology lab, built in-house in under two years: automated billing-code correction, digitized intake, templatized physician reports, and a physician application, Arkana Connect, that replaced a long-running vendor build. They walk through one rebuild in detail and share early results from training models on the lab's own case data, along with what remains unvalidated. They also describe the seven-week internal cohort that started it and how value is moving to the edges of each role. Attendees will leave with a practical test for finding problems that off-the-shelf software has never solved, and a case for why a small, regulated company can build them itself.

Agentic AI promises speed, but its nondeterminism, and unbounded agent workflows in particular, can bring risk, rework, and frustration. JD Black, Sr Director and Chief Engineer in Northrop Grumman's Chief Information & Digital Office, describes a hypothesis that the known elements of engineering success, including rigorous process, version control, deterministic feedback, and open MBSE architecture-as-code, can be automated and amplified for agents. He walks through a five-phase agentic software and IT lifecycle, from architecture and system engineering through design, planning, test-driven development, and release, in which agents execute steps with deterministic exit criteria and humans hold formal review gates. He shares what the team has observed about small-batch execution, versioning agents, skills, and workflows, and tracing requirements to SysML v2 architecture and automated acceptance tests. Attendees will leave with his requests of the community: a standardized AI package manager and deterministic agentic workflow standards.

When a pilot goes down, rescue teams depend on a gridded reference graphic (GRG) with every building labeled, and building it by hand for a large urban area can take eight or more hours and still be cut short when time runs out. Kevan Mordan and Kenny Slater describe how Rise8 was asked to build a TAK plugin that detects and labels buildings with AI, runs fully offline on older Android phones and low-spec Windows laptops, meets cyber compliance, and ships in six weeks with six people. They explain how the team worked in areas it had not touched in years, ran a bake-off of detection models, decided what counts as a building and how labels should read, and handed AI the mechanical work while they owned quality. Attendees will learn how starting from the mission, keeping a firm quality bar, and moving fast fit together when the product is in operators' hands.

Asked whether the company uses Chinese AI, most teams lack a clean answer, and a country-of-origin ban says little about what actually makes a model risky. Nicky Pike, Field CTO (Americas) at Coder, proposes inspecting the model as a house in three layers: the wiring (what was baked in during training, including research-demonstrated backdoors), the contractor (what model files and setup scripts can run when loaded, including confirmed malicious models on public hubs), and the doors (what data can leave and what an agent is allowed to do). He sets out practical steps for pulling from public hubs, shows where bans fit and where they do not, and describes how Coder routes every agent through workspaces and one gateway so a model can be shut off or swapped. Attendees will leave with a way to answer, for one model, everywhere it runs.

Getting an AI software factory to work once is a different problem from running it every day. Justin McCarthy frames the first stage as Day 1, validation: design a loop, build a validator, and keep running and fixing when it gets hard. He then turns to Day 2, the operations of industrialized software factories, where the question becomes what a team does after automating software production. He offers a hint for the transition: if the model can just barely do something, the team must stop doing it. He then presents seven patterns for organizations in this phase, including how roles are divided, how effort is split between offense and defense, how to demand tokens, how competence spreads, and how teams backslide into stasis. Attendees will leave with a vocabulary for the patterns that follow a first successful deployment, and a forecast of where these factories are heading.

When an AI agent writes the code, it can do exactly what was asked and still miss what was meant. Andrew Stellman opens with a vibe-coded bus tracker that sent him to the wrong stop, then uses it to show how requirements, code, and tests can all agree while the original intent is lost. Drawing on his experience managing large development teams and on decades of quality engineering, he frames the problem as specification, delegation, and verification. He describes Quality Playbook, his open-source skill for Claude Code, Cursor, and Copilot, which reconstructs intent from documentation, project discussions, and users' reports, turns it into requirements, checks the code against them, and confirms findings with failing tests. Attendees will see findings from open-source projects such as Gson and the Linux kernel, including a long-standing defect found through user bug reports, and learn how to check that intent survived.

Executives keep asking where to invest in AI, and the answers often go wrong in predictable ways. Joe Beutler of OpenAI draws on five recent executive conversations to name the mistakes: copying a peer's AI strategy, promising value without a plan to realize and measure it, trying to do everything yourself, stopping at engineering, and making every team wait for one mandatory platform. He asks where each business actually wins, what a build decision will still differentiate by launch, and how to measure cost per successful workflow. He also covers what an agent harness does, how to compare complete provider systems, and how delegation and enablement spread beyond engineers. Attendees will leave with questions to put to their own investment decisions, including how to give strategic teams an approved path alongside a shared platform, and a forecast of when these capabilities become table stakes.