March New York Roundtable Recap
Catch Up on What You Missed!
1. Setting the Context
On March 31st, we convened a cross-functional group of leaders from across the CardLinx ecosystem—loyalty platforms, financial
institutions, payment networks, data orchestration providers, and commerce media—to explore a rapidly emerging question:
What does agentic commerce mean for our industry, and what should we do about it now?
We entered the room with uneven levels of familiarity, but we quickly aligned on two realities:
• We are still early—both conceptually and operationally
• The pace of change is fast enough that “waiting” is itself a strategic decision
We also shared a posture that shaped the conversation: we are not yet experts in agentic commerce. We are participants in its formation.
2. How We Framed Agentic Commerce
We spent time grounding ourselves in a shared understanding of the concept.
At its core, we described agentic commerce as:
The delegation of commerce-related decisions and actions—from discovery through transaction and beyond—to AI agents acting on behalf of consumers.
However, an important distinction emerged:
• Today, most activity is happening in discovery and search
• Very little (well under 1%) is happening in end-to-end autonomous transactions
This led us to a critical reframing:
We are not yet in a world of autonomous commerce. We are in a world of assisted commerce that is becoming increasingly agent-mediated.
At the same time, what makes this moment different from past waves is that multiple layers of the commerce stack are being disrupted simultaneously—discovery, identity, payments, loyalty, and data infrastructure are all shifting at once.
That simultaneity is what gives this moment both its urgency and its ambiguity.
3. A Layered View of What Is Actually Emerging
Rather than treating agentic commerce as a binary (it exists / it doesn’t), we collectively built a more useful model: a layered progression of capabilities.
We discussed how the experience evolves:
• At the most basic level, agents surface relevant information (availability, sizing, pricing)
• Then they begin to filter and personalize options based on preferences
• Then they optimize across ecosystems (loyalty, rewards, coupons, shipping thresholds)
• Only at the far end do they execute transactions autonomously
What became clear is that:
These layers will coexist. Consumers will not all move to full automation at once—and many may never do so.
This has two implications for us:
1. We must design for multiple levels of engagement and trust simultaneously
2. The near-term opportunity is not autonomy—it is efficiency and relevance
4. Trust Is the Constraint (Not Technology)
If there was one theme that consistently surfaced across the discussion, it was this:
Agentic commerce is not blocked by capability. It is blocked by trust.
We unpacked what “trust” actually means in this context. It is not a single issue; it is a system:
• Verifiable intent — Can we prove the consumer authorized the action?
• Identity assurance — Do we know the agent is legitimate? (KYA – Know Your Agent)
• Liability frameworks — Who is responsible when something goes wrong?
• Auditability — Can actions be traced and explained?
• Revocation — Can the consumer easily undo or withdraw access?
We also heard very practical barriers:
• Issuers declining transactions because they look like fraud
• Lack of “click-to-activate” equivalents for agent-driven purchases
• Consumer discomfort with giving agents financial authority
This led to a shared conclusion:
Agentic commerce will not scale until trust infrastructure is built—and that infrastructure is still in progress.
5. Human-in-the-Loop Is Not Temporary—It Is Foundational
We explored whether humans are simply a transitional step in the system or a permanent feature.
What emerged was not a binary, but a spectrum of control:
• Fully supervised (agents recommend, humans decide)
• Guardrailed autonomy (pre-approved rules, limits, preferences)
• Full autonomy (agents act independently within defined authority)
Importantly, we recognized:
• Trust varies by category (we may trust an agent to reorder paper towels, but not to book travel or make large purchases)
• Users will want control over where they sit on this spectrum
So rather than designing toward a fully autonomous future, we aligned around this:
We need to design systems that allow users to move along a continuum of trust—not force them into a single model.
6. The Shift from Push to Pull Is Already Happening
One of the most significant structural shifts we discussed is this:
Commerce is moving from a push model to a pull model.
Historically:
• Brands pushed offers, ads, and messages toward consumers
In an agentic world:
• Agents pull information based on explicit or inferred intent
This changes everything:
• Visibility is no longer about placement—it is about accessibility and structure
• Offers must be:
o Real-time
o Machine-readable
o Context-aware
• Discovery becomes intent-driven rather than exposure-driven
We also connected this to a broader shift:
• Advertising dollars are moving toward performance models
• Agents respond to price, value, and structured signals—not display creative
The implication for all of us:
If our data, offers, and products are not accessible to agents, they effectively do not exist in the new discovery layer.
7. Loyalty Is Not Disappearing—It Is Being Rewritten
We spent significant time wrestling with a tension:
If agents optimize for price, does loyalty collapse?
There were two competing perspectives:
The Risk
• Agents could commoditize everything into price comparisons
• Brand relationships could weaken
• “Best deal” logic could override long-term loyalty behavior
The Counterpoint
• Agents can incorporate:
o Loyalty status
o Rewards value
o Long-term benefits
• They can actually surface loyalty value earlier in the journey, not just at checkout
We also reframed loyalty more fundamentally:
• From program-centric (“earn points”)
• To consumer-centric (“what is the best outcome for this individual right now?”)
This led to a key insight:
The future of loyalty is not about programs—it is about personalized value at the moment of intent.
8. Data Is the Strategic Lever—but Also the Constraint
We aligned quickly on the importance of data—but also surfaced its complexity.
We distinguished between:
• Public data (widely available, increasingly commoditized)
• Private / first-party data (high-value, permissioned, differentiated)
We also identified several tensions:
• Many current systems (e.g., open banking, CLO) are batch-based
• Agentic experiences increasingly require real-time data
• Consent models are not yet designed for continuous agent access
We also raised a critical behavioral question:
• Consumers may share data for insights
• But are far less comfortable sharing financial control
This leads to a strategic implication:
The organizations that can combine high-quality data with strong permissioning and governance will define the next phase of the ecosystem.
9. Infrastructure Gaps Are Real (and Material)
We identified several structural mismatches that will shape near-term progress:
• Real-time decisioning vs. batch settlement systems
• Agent-driven flows vs. legacy fraud models
• Emerging protocols vs. fragmented standards
We discussed emerging approaches such as:
• Model Context Protocols (MCP)
• Early-stage industry frameworks (e.g., UCP)
But we also acknowledged:
There is no dominant standard yet—and waiting for one may mean losing influence over how it is defined.
10. Consumer Adoption Will Not Be Linear
One of the most important reframes in the room was this:
Adoption will not be driven by generation. It will be driven by persona.
We discussed:
• Value-driven users vs. quality-driven users
• Convenience-seekers vs. control-oriented users
• Voice-enabled users (including older populations) vs. visual interfaces
We also acknowledged:
• Many consumers are already using AI without realizing it
• Trust and simplicity—not technology—will determine adoption
The implication:
We should not design for “the future user”—we should design for multiple user types simultaneously.
11. Strategic Postures in the Room
We saw a clear spectrum in how organizations are approaching this moment:
1. Proactive Builders
• Engaging early in protocols and infrastructure
• Seeking to shape standards and maintain control
2. Pragmatic Operators
• Focusing on near-term value (discovery, personalization)
• Avoiding overinvestment in immature capabilities
3. Observers
• Monitoring developments
• Waiting for clearer signals before committing
Despite these differences, there was a shared recognition:
Doing nothing is not neutral—it increases the risk of becoming a passive data provider in someone else’s system.
12. Broader Implications We Touched
Beyond commerce mechanics, we surfaced wider implications:
• Workforce shifts toward supervision, orchestration, and judgment
• Regulatory lag, with parallels to BNPL and auto-renewal
• Advertising transformation, with budgets shifting toward performance
• The possibility of agent-to-agent economies over time
These are not immediate, but they are directional.
13. What We Left With
We did not leave with a single answer and that was appropriate.
Instead, we left with a clearer set of realities:
• We are early—but not too early to act
• Trust infrastructure is the gating factor
• Discovery is the immediate battleground
• Data and interoperability will determine long-term position
• Loyalty, marketing, and payments will all be reshaped—but unevenly
• The timeline is uncertain, but the direction is not
Perhaps most importantly, we left with a shared responsibility:
To participate in shaping this ecosystem—rather than reacting to it once it is defined by others.
If you missed this one, join us at our June New York Roundtable! Click the link for more details.
