Jack Wang is an enterprise AI delivery leader who established a forward-deployed engineering team within Accenture’s innovation practice, helping large, regulated organizations put agentic systems into production. His work addresses a practical obstacle that increasingly determines whether enterprise AI succeeds: models can move faster than the governance, funding, deployment infrastructure, and organizational trust surrounding them.
Earlier in his career, Wang worked on public-sector and transportation analytics at Deloitte and technology products and process automation at EY. At Accenture, he delivered transformation programs across banking and telecommunications, including work involving Lloyds Banking Group, BT, and Vodafone, before expanding into data-and-AI leadership. Within its RX innovation practice, he developed customer-facing engineering capabilities, reusable delivery frameworks, and relationships with AI companies including Anthropic. His professional background reflects sustained experience working inside the operational constraints of major institutions.
Credited as Accenture’s GenAI Lead at AI Engineer Europe 2026, Wang joined Jess Grogan-Avignon to examine why enterprise agent projects stall between prototype and production. Their example was unusually concrete: an agentic application built in approximately two weeks required another year to reach production while infrastructure, security, AI-gateway, governance, and application teams aligned.
What enterprise AI deployment actually requires
- Governance at machine speed. Coding agents increase software output, but manual reviews, approvals, deployment processes, and underdeveloped CI/CD simply relocate the bottleneck. Wang and Grogan-Avignon advocate engineering automation and executable controls that preserve oversight while reducing dependence on sequential human sign-offs.
- Portfolio-based AI investment. Agentic products often reveal their value through experimentation, making fixed upfront assumptions about scope, cost, and returns unreliable. Their approach spreads funding across multiple experiments and leaves room for new products and services beyond straightforward efficiency gains.
- Hypothesis-driven agent delivery. Because agent behavior is probabilistic and evolving, delivery teams need short build-evaluate-iterate cycles, measurable confidence, and the ability to translate technical results into decisions enterprise stakeholders can understand.
- Progressive autonomy and living memory. Agents should advance from observing human workflows to recommending actions, then to constrained autonomy with safeguards, as operational evidence accumulates. The feedback generated through customer behavior, corrections, and edge cases becomes living memory: proprietary signals that help deployed systems improve over time.
