The Green Sheet Online Edition
February 2, 2026 • 26:02:01
Agentic commerce, Part 2: Rethinking intelligence and trust
Digital assistants are hardly new to consumers and businesses. Technology-driven brands have been integrating virtual assistants, chatbots and interactive voice response (IVR) systems into phones, homes and websites for decades. Each integration has taken us farther away from checkout lanes, printed receipts and plastic credit cards.
This series explores agentic commerce, the most extensive integration of artificial intelligence (AI) and payments to date. Part 1 examined a new reality in which autonomous AI agents transact on behalf of humans. Part 2 looks at how orchestrating intelligence and trust will help agentic commerce reach its full potential and play an integral role in our businesses and lives.
Beyond instinct and intuition
Sam Sharma, founder of Elevate AI Tech, an enterprise AI solutions provider, said that AI is "computer-made, not man-made," which makes it different from other technologies. "Every time a new system enters our world, we get better at integrating it and making it human," he added. "AI is more human than any technology we've known, and it integrates naturally into everything we do."
Sharma pointed out that machines learn by separating thoughts into multiple options and neural-network pathways; it's similar to human learning but faster and more efficient. While humans are guided by cognitive bias and intuition, AI can recognize patterns at a scale and speed that no individual can match.
Before AI, when people relied on intuition and historical data to make decisions, Sharma recalled, some people could recognize patterns more easily than others. AI helped to level the playing field by enabling people to think and recognize signals in similar ways, he said. He cautioned, however, that AI may inherit biases from its human creators, which could become a trust issue as agents act autonomously.
"If two people use the same AI tool but are compensated differently, that can reflect a deeply embedded bias in the system," Sharma said. "When an AI values one professional's background or credential over another from an equally qualified applicant, we clearly have more work to do."
From tools to teammates
Sharma shared that AI can illuminate insights that humans frequently miss, which can be useful for businesspeople. In sales, for example, AI-powered solutions can pick up nuances in tone from a conversation with a prospective client by observing a person's choice of words and the subtext, emotions and political dynamics embedded in spoken and written language.
"After a meeting, you can run a sentiment analyzer on the transcript and uncover who felt insecure, who needed reassurance and what would make them more likely to open doors," Sharma said. "That's real intelligence that takes us beyond gut instinct and guesswork."
Having this type of intelligence frees people who are chasing a deal from depending on intuition and past performance, Sharma noted, adding that AI can recognize multiple patterns in seconds, giving it the potential to expand human knowledge, skills and decision-making.
Multilayered AI ecosystem
Sharma framed the AI ecosystem as a multilayered technology stack with five functional layers, which he described as follows:
- Agentic AI: Automate complex tasks and processes.
- AI agents: Execute tasks autonomously.
- Generative AI: Generate content at scale.
- Neural networks: Detect complex business patterns.
- AI and ML: Turn data into decisions using artificial intelligence and machine learning.
Generative AI, which Sharma said is the most well-known AI layer, creates content and images from prompts and is different from agentic AI, which executes tasks and processes on a user's behalf. Below that are deeper neural-network systems, where machine learning turns large volumes of data into decisions through advanced pattern detection.
"Most people stop at generative AI because they don't know how to go deeper," Sharma said. "But as you move down the stack, AI shifts from complementing us to augmenting us."
Rather than lumping AI into one category or pushing it into areas where it doesn't fit, Sharma recommended looking for places where AI and automation can deliver the highest impact at the lowest cost. A process he called "AI opportunity mapping" can help individuals and enterprises identify bottlenecks, pain points and communication gaps between systems, data and teams.
"When systems don't talk to each other, when data delivers analytics but hides insight, when teams resist change—that's friction," he said, calling it one of the leading causes of business failure.
He advised companies to embed AI into agentic workflows as early as possible in a project to close communication gaps and enable teams to test and validate agents before deployment.
From KYC to governing actors
Hartley Thompson III, CEO of Microblink, an adaptive identity intelligence platform, stated that trust is essential to agentic commerce.
"You can't have an agent acting on your behalf unless there's real trust," Thompson said, adding that trust is not a one-and-done event but a dynamic process involving continuous review, confirmation and verification. Integrating that kind of always-on trust into a broader digital identity operating system, he noted, requires three foundational elements. "First, every automated action must tie back to verified human intent," he said. "Today, that means confirming the person who authorized the agent. Second, trust requires continuous behavioral evaluation, not a single credential check. Systems must monitor whether actions remain aligned with permissions over time. Third, trust decisions must be consistent across channels."
Most importantly, Thompson added, agentic and human trust must be held to the same standard, whether it's a human transacting on a mobile app or an agent buying something online. The evolution from human-centric, know-your-customer (KYC) compliance to governing digital actors is a major shift, especially for traditional fraud models that tend to look for anomalies after a transaction has been completed.
Thompson portrayed that shift in financial services as moving from confirming trust at a single touchpoint to continuously reviewing and verifying activity across the lifecycle of agent-driven transactions, with scrutiny escalating only as confidence degrades.
"In the old model, you checked someone's ID once at the door," he said. "Now it's about understanding how they continue to behave throughout their visit."
Omnichannel identity protection
Thompson urged enterprises and teams to invest in what he called "trust infrastructure," to align with agentic protocols in payments, technology and commerce. As Google, Microsoft and other major players build protocols that allow AI agents to initiate and complete transactions, he noted, Microblink and other technology platforms are building digital identity frameworks that verify the human behind the agent.
These digital identity frameworks are always on and always connected, Thompson said, and designed to reduce friction by monitoring, detecting and preventing fraud, deepfakes and synthetic identities across multiple channels and media, which he described as follows:
- Document verification: Extract data in real time and in multiple languages using AI-powered non-identity document extraction and classification.
- Payment card verification: Perform liveness checks for card-present transactions, reducing chargebacks and ensuring PCI-DSS compliance.
- Orchestration and risk intelligence: Combine document, device and behavioral signals, using real-time analytics to assess pass rates, risk signals and operational performance.
- Continuous AI innovation: Combine on-device processing with cloud-based AI models to maximize speed, privacy and fraud detection while supporting continuous learning.
He pointed out that the framework functions as a platform enterprises can leverage to balance security, compliance and growth while delivering an exceptional user experience. It also safeguards agentic commerce, he added, by keeping agents transparent and auditable through clear records of who did what, when, where and under whose authority. Less friction, more freedom "Every major evolution in commerce brings new risk and new opportunity," Thompson said. "The organizations that thrive will build trust, not just defenses."
Sharma concurred, advocating an AI-first approach to building sustainable agentic commerce infrastructure.
"As we move through the AI integration lifecycle, from discovery to prototypes to production, agentic workflows and automations become part of the project," Sharma said. "These agents are present in the prototypes and pilot builds that validate an integration before it moves into production and deployment, replacing isolated tools with embedded intelligence to remove friction and improve decision-making and relevance."
When intelligence and trust are in sync, the payoff is human freedom, Thompson added. For example, he mentioned an agent could be shopping for better insurance and switching carriers late at night while a policyholder sleeps; other authorized agents could be autonomously rebooking flights or trading assets, triggering requests for biometric authentication only for higher-risk actions. Continuous trust, he noted, allows friction to appear when needed without slowing workflows or payment flows.
"This gives me more time to spend with my family, build a business or travel," Thompson said. "The net positive is humanity having more freedom instead of scrolling endlessly to transact."
As Sharma and Thompson made clear, when intelligence and trust operate as core components of a dynamic, adaptive system, agentic commerce can move from a technical construct to a material reality. Both agreed autonomy is within reach, but success will depend on how deliberately intelligence and trust are designed to work together, not as parallel capabilities but as a single, integrated foundation. 
Dale S. Laszig, content strategy director at The Green Sheet and founder and CEO at DSL Direct, is a payments industry journalist, creator and consultant. Connect via email at dale@dsldirectllc.com and LinkedIn at www.linkedin.com/in/dalelaszig.
Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.
