AI Agents on Both Sides: Brand NPCs Meet Shopping Bots

McKinsey projects agentic commerce — AI agents researching, shopping and transacting for humans — will generate $3 to $5 trillion annually by 2030. That's the number hanging over every CMO currently bolting an AI character onto their brand, and it cuts both ways: the marketer's agent and the shopper's agent are arriving at the same checkout simultaneously.

Adswerve CEO Tom Zawacki lays out the shift in a MarTech Cube essay on AI-enabled NPCs, those once-scripted background characters now promoted to brand ambassadors. The tension he identifies is blunt. IAB research found 30% of Gen Z consumers already describe brands that use AI in ads as "inauthentic," yet 94% of media decision-makers in the same report believe AI will lead to a future where every advertisement is personalized to the individual.

As Zawacki puts it: "The characters that once existed purely to advance a single aspect of a player's journey, or merely as background filler, are now among the most dynamic participants in the experience."

Modern NPCs learn from interactions, shift tone based on sentiment and act autonomously — personalizing offers and optimizing journeys without a human in the loop.

How are brands deploying these characters?

Zawacki identifies three levels of engagement:

• Viral, long-form character arcs. Duolingo turned its owl mascot Duo into an interactive AI guide inside the app. The character's unhinged personality drove Duo's staged "death" in early 2025, which generated 1.7 billion impressions in two weeks. He's fine, for the record.
• In-environment placement. Fortnite and Roblox host brand appearances from PUMA and CeraVe, where characters add value to the world rather than interrupting it with a digital banner.
• Personalized ambassadors at scale. Qatar Airways' "Sama," an AI-powered digital cabin crew member, answers traveler questions and plans trips inside the airline's app and its QVerse immersive platform.

Who's on the other side of the transaction?

Consumers are deploying their own agents, and the examples are already shipping:

• Amazon's Rufus chatbot now includes an "Auto Buy" feature that monitors products and purchases automatically when an item hits a target price.
• Delta's new AI concierge surfaces and books alternative flights on the spot when disruptions hit.
• Google's agentic checkout lets shoppers set a spending limit and criteria, then completes the purchase when the right product appears at the right price.

The collision is structural. Brand ambassador agents engage, persuade and convert; they're optimized to tell a story. Shopping agents evaluate, filter and transact; they're optimized for the lowest price, without sentiment. When the two meet, Zawacki argues, the winner isn't whoever tells the better story — it's whoever runs the more intelligent marketing operation.

What does that operation require?

Four foundations, per Zawacki:

• Unified data layer. Know who the user is — human or AI agent — across every touchpoint, with clean, structured product and pricing data that shopping agents can query during evaluation.
• Sophisticated modeling. The signals telling a brand agent what to say, to whom and on which channel come from models built on the data, not the data itself.
• Real-time activation. Orchestration engines and streaming pipelines deliver the right message, channel and moment — and keep pricing data current enough to compete when agents compare offers across sources simultaneously.
• Ubiquitous measurement. Track both human-driven and AI-mediated decisions, so brands can optimize as shopping agents claim a larger share of the journey.

The stakes are simple. Without that foundation, brand agents default to fixed responses and stale personalization — the exact scripted quality that makes them feel like, well, NPCs. Marketers building the predictive stack today, Zawacki writes, are the ones who will define what customer journeys look like tomorrow.

Zawacki brings 30-plus years in data, marketing and AI to the argument, including a prior role as President/CMO of Data Axle and SVP at Rocket Fuel, where he led an industry-first integration with IBM Watson.