ODSC AI NYC Brings Workflow Ownership to Its December Summit

The difficult moment in an AI workflow arrives when a plausible result needs somebody's approval. ODSC AI NYC: AI for Work Summit will examine that handoff on December 2–3, 2026, at Jay Conference Center Bryant Park in Midtown Manhattan. Its preliminary program brings engineering, business operations and leadership into a gathering organized around the work each group owns.

The summit has 4 tracks: Build, Business Workshops, Lead, and AI Engineering Fundamentals. Technical practitioners can work on agent reliability while business teams examine recurring tasks and leaders consider governance and measurement. For founders and operators planning an AI rollout, the useful question is how those separate efforts will fit together when the team returns to work.

Four tracks, one workflow to account for

ODSC's program offers a way to inspect an assignment from several seats. An engineer might start with retrieval quality, a process owner with an overdue report, and an executive with the cost of supervising the result. Those are different views of the same delivery problem. A successful handoff needs the people involved to agree on what counts as finished.

The Business Workshops program makes that agreement unusually concrete. Its planned full-day training covers task specifications, connections to documents and business systems, multi-step processes, and review before team rollout. The connections segment uses a prepared demonstration workspace, so attendees do not need to connect corporate accounts for that exercise. That makes the access discussion something participants can take back to their own technology and security teams.

For an operator, the preparation starts before travel: select a recurring task whose inputs and acceptable result you understand. A report with a disputed number is more useful workshop material than an ambition to make the entire business AI-powered. With a bounded assignment, colleagues can ask where context enters, which decisions need approval, and how an error becomes visible. The summit's track structure gives those questions somewhere specific to go.

The instructors put scrutiny inside the build

The current agenda lists Michael Levan for an agent-observability workshop and Bruno Gonçalves for work on an LLM evaluation harness. These sessions approach reliability through the evidence a system leaves behind and the tests used to assess its behavior. The distinction matters commercially: a customer-facing error creates both a correction task and a question about how the next error will be caught.

Mike Taylor's planned session turns to an evaluation suite a team can keep using. The ODSC Team's October 27 preview describes representative examples, explicit checks and scrutiny of false passes. For a business owner, that creates a useful conversation with engineering: which examples actually represent the work, and who has enough domain knowledge to recognize a superficially convincing failure?

On the business side, Greg Rosner's session focuses on defining an agent's job. Rosner is the founder of PitchKitchen, giving that discussion a direct business-operator context. Mark Kashef, founder of Prompt Advisers, is listed for a Claude Cowork workshop covering roles, context, process steps and review. ODSC AI founder Sheamus McGovern will lead the full-day business training. These are announced program assignments, with further instructors and details still being added.

A New York room with an operating question

The setting places this conversation in Midtown Manhattan, but geography alone cannot do the work of an agenda. The useful connection is between people evaluating AI systems and people accountable for the process those systems enter. DevCuration's related coverage of AI agent reliability and AI software architecture follows adjacent implementation questions. Those gatherings are separate events, with their own organizers and programs.

The leadership track adds another pressure: deciding what improvement looks like when measurement starts late. Michael Schrage is listed for a panel about proving ROI without a previously measured baseline. His MIT Initiative on the Digital Economy profile establishes his institutional affiliation. The session topic gives leaders a reason to arrive with an actual measurement problem, rather than treating a productivity estimate as a completed business case.

For founders, the cross-track format could expose the organizational work a product still requires from its buyer. For investors, it offers a program through which to examine implementation friction and customer responsibility. Those are editorial reasons to follow the summit, rather than promises of deal access or commercial returns. A conversation about the work someone will maintain can be more revealing than a polished product demonstration, especially when the maintenance owner is also in the conversation.

Choose access around the work you intend to do

The organizer's registration listing distinguishes limited Open Pass programming from Training and Leadership Plus access. The current summit site describes the Open Pass as covering keynotes, open talks, selected breakouts and networking, with hands-on workshops requiring an appropriate paid pass. Leadership Plus adds the speaker and leadership dinner. Readers should check the live registration terms and current prices before choosing; a free entry route and a workshop plan are different purchases.

ODSC lists attendee networking and an evening reception, alongside more curated interactions associated with eligible passes. The program identifies intended audiences, not a guaranteed roster of customers, investors or employers. Bringing colleagues across functions could make the conversations more useful because they can compare the same task after attending different tracks. The outcome remains work for the team to produce, rather than a benefit a badge automatically delivers.

The venue is Jay Conference Center Bryant Park, 109 West 39th Street, New York, NY 10018. The official site lists doors at 8 AM; individual session times are still being finalized. Check the current program and passes before attending December 2–3. A team that arrives with a report, customer process or internal assignment can carry its unresolved decisions through the workshops, then put names beside the review points before the next run lands back on somebody's desk.