The Architecture of Action: Think, See, Do

Amidst the high-decibel energy of the NRF 2026 show floor, the conversation at the Manhattan Associates booth cut through the typical trade show buzz. Standing on the front lines of retail's biggest stage, Manhattan Associates’ CTO Sanjeev Siotia and VP of Field Marketing Liz Sophia discussed a fundamental shift in the industry: the move from "Assistant AI" to "Agentic AI."

While the last few years were defined by the excitement of Generative AI’s ability to summarize data, a new reality has set in for those in the trenches: visibility without execution is just more work.

Click the video below to watch the discussion.

The Architecture of Action: Why Built-In Wins

Many tech vendors are currently racing to build "data lakes"—external reservoirs where retail data is extracted, processed, and then fed back into a chatbot overlay. According to Siotia, this approach is fundamentally flawed for an industry that moves at the speed of retail.

“Agents living inside the product have real access to the data within the product,” Siotia explained. “Them being in the system makes it a lot more easier for these agents to work. They don't have to pull data, take an action, and then try to come back.”

By embedding AI directly into the Manhattan Active® Platform, these agents inherit the system’s native security, governance, and—most importantly—its ability to execute.

The "See, Think, Do" Framework

To understand the power of a true AI workforce, Siotia breaks it down into three core capabilities:

1. See: The agent observes live data from the heart of the supply chain.
2. Think: It evaluates that data against a specific objective.
3. Do: It executes the solution autonomously within the core system.

“The agents are supposed to be autonomous and they’re supposed to actually achieve an objective,” said Siotia. He pointed to the Labor Optimizer Agent as a prime example. While a human manager might spend hours trying to figure out how to reassign warehouse staff to cover a bottleneck, an autonomous agent can identify the gap and execute the move in minutes.

From Complexity to "Agentification"

Retailers today are overwhelmed by complexity, particularly in the labor and omnichannel space. Manhattan’s response is the Agent Foundry™, a groundbreaking platform that allows businesses to take their unique, messy operational problems and "agentify" them.

Whether it is a Shipment Tracking Agent identifying a late load or a Contact Center Agent processing a complex return and exchange start-to-finish, the goal is to move beyond the "Fancy Librarian" model of AI.

The 24-Month Mandate

Before heading back into the NRF crowd, Siotia left retailers with a surgical mandate for the next two years: Demand open API access.

“For AI to work, you need access to the data and you need access to be able to execute actions,” Siotia noted. “Retailers should be demanding open API access so that AI can actually do the work it needs to do.”

The era of settling for AI that just "assists" is over. As supply chains become more volatile, the only AI that matters is the AI that acts.

Ready to see a true AI workforce in action? Read the full announcement on the commercial availability of Manhattan AI Agents.

Video Reference Timestamps

• [00:21] – The API-First Advantage: Why built-in beats data lakes.
• [01:25] – Beyond Assistance: AI that takes real-world action.
• [02:37] – Architecture Matters: Why agents must live inside the product.
• [03:00] – Complexity Solved: Using the Labor Optimizer out-of-the-box.
• [03:55] – Agent Foundry™: Turning unique problems into AI agents.
• [04:03] – The 2-Year Outlook: Demanding open API access.