Dentsu Exec Urges Shift to Agent-Operated AI Workflows

A Call to Rethink AI at Work

At the Seedance Global Tour's Singapore Main Forum, Dentsu's Jianhao Chen delivered a keynote arguing that the current wave of workplace AI adoption is aimed at the wrong target. Boosting an individual employee's productivity with a chatbot or copilot, he argued, does not automatically translate into enterprise-wide growth 1. His prescription: companies need to move beyond treating generative AI as a personal add-on and instead build what he calls "Human-Led, Agent-Operated" workflows — AI-native processes where autonomous agents handle execution while humans set direction and judgment calls 1.

This framing lands amid a broader industry reckoning over what AI adoption actually delivers versus what it promises. Across sectors, the pattern Chen describes — scattered tool adoption without structural redesign — appears repeatedly in how companies are experimenting with automation.

Productivity Gains, Measured and Unmeasured

The tension between visible AI enthusiasm and measurable business outcomes shows up elsewhere in recent coverage. Meta CEO Mark Zuckerberg has publicly credited AI with accelerating internal software development, saying engineers are building applications faster than before 6. Yet Meta has not disclosed which tools its teams use or published concrete productivity metrics, leaving the claim more anecdotal than verified 6 — precisely the kind of individual-level productivity narrative Chen warns should not be mistaken for organizational transformation 1.

At the individual level, tangible workflow gains are easier to document. One widely tested use case involves AI-powered voice dictation, with one long-term user reporting having dictated more than 120,000 words using a handful of favored tools, prized for accuracy, privacy, and minimal disruption to daily work 3. In logistics, fulfillment company ShipBob has rolled out a suite of AI tools, including a Claude-based integration, letting merchants manage orders, shipping performance, and inventory through conversational prompts rather than traditional dashboards 2. These examples illustrate genuine workflow automation taking hold in specific niches, even as broader enterprise-level transformation remains harder to quantify.

The Risks Riding Along with Adoption

As AI tools embed themselves deeper into daily operations, downsides are surfacing alongside the productivity pitch. Newsweek reports a rise in "algorithmic retaliation" litigation, with new lawsuits questioning whether AI-driven workplace systems are being used to mask discrimination or retaliatory decisions against employees 5. Separately, backlash erupted over X's Grok AI tool after users discovered it could generate deepfake content, prompting outrage over the platform's safeguards and raising broader questions about AI governance 4.

Why the Distinction Matters

Taken together, these developments underscore Chen's core argument: scattered AI tools — whether for dictation, fulfillment, coding, or social platforms — can each deliver localized value or localized harm, but neither guarantees coherent enterprise strategy. The real test for organizations, per the Seedance forum keynote, is whether they redesign workflows around agents and human oversight jointly, rather than simply layering AI onto existing individual tasks and hoping the gains compound on their own 1.