The AI Revolution in Loyalty:
From Complexity Cost to Strategic Advantage
A conversation with Emil Sarkissian, CEO of Loyalty Methods, on how platform architecture determines AI readiness in enterprise loyalty programs
Most enterprise loyalty vendors are calling themselves AI companies. Few have changed anything underneath. The platforms that actually deliver on AI's promise weren't redesigned for it. They were built, years earlier, with the architectural discipline that AI now requires. At the 2026 Loyalty Summit in Chicago, Loyalty Methods CEO Emil Sarkissian sat down with Bill Hanifin of The Wise Marketer Group to explain the difference — and what it means operationally.
The Gap Encore Launch: Treating Loyalty as a Brand
Gap’s Encore loyalty program went live in February 2026. It wasn’t the mechanics. It was the framing: Gap treated the program itself as a brand.
“The Encore itself, the program itself, is treated as a brand. They already have the four brands that sit within the Gap ecosystem. But I think rebranding the program and treating it as its own brand and experience creates a different kind of emotional bond with it.”
This reveals a broader shift: the most successful programs are no longer transactional systems. Gap’s approach integrates rewards directly into the shopping bag experience, layers in gamification, and ties unique burn opportunities to the brand’s cultural assets. The program shows how loyalty is becoming inseparable from the core retail experience.
Sarkissian also pointed to Zac Posen, Gap Inc.’s Creative Director, and signed merchandise in the marketplace as examples of burn opportunities that are genuinely differentiated — things that cannot be easily copied by other retailers. Emil emphasized that the February launch wasn’t an endpoint but “a start of a new chapter,” a “platform in stability” that opens an “open field for innovation.” The launch is not the completion of anything. It is the beginning of what the Encore brand becomes.
The Architecture Question: Why Some Platforms Cannot Leverage AI
Many vendors have simply attached “AI-powered” to their existing feature lists without addressing the underlying structural requirements. Sarkissian was direct about what separates real AI readiness from labeling.
“We shouldn’t be the bottleneck ever. I should never go into a client where our platform is integrated deeply into their stack, and there’s a situation where they say, ‘We want to do X,’ and we say, ‘No, ours doesn’t do that.’”
ReactorCX was built on this principle from day one. Not just point-and-click configuration, but full programmability: the ability to write actual code to define loyalty behavior. Very few clients ever needed that depth. But having it meant there was always an answer to any problem. No ceiling.
The Killer Use Case: AI Excels at What RCX Was Built For
The challenge with programmability has always been complexity. Enterprises don’t want their marketers writing code. But AI has changed the equation.
“The killer use case right now is coding. AI can code really well. It turns out we are programmable. So therefore, now, if we combine the two together, it’s now easier than ever to say, not only is everything possible, but the complexity cost of it has gone very, very low, because you can now say in English what you want, and it’ll just go do it with the platform.”
RCX’s programmability has become a structural asset because it speaks the native language of large language models. Where other platforms require AI to navigate dropdowns and click-through workflows, RCX allows AI to work directly with code and JSON configuration. AI can generate complex rule configurations from natural language, write validation logic automatically, and build multi-step workflows by composing API calls.
MCP: The Integration Layer That Makes AI Practical
Technical architecture only matters if it delivers tangible value. Loyalty Methods has implemented an MCP server — a Model Context Protocol server — that allows large language models to reach deep into ReactorCX configuration, even performing programming of its own when needed.
The practical application is direct. A strategist can now ask in plain English: “Can you take a look at our program’s terms and conditions and examine our configuration and see if there’s a mismatch?” The system responds with a table: earn rates, breakdowns, areas that need legal review. That is operational risk management that would previously require days of manual cross-checking.
What makes this possible is ReactorCX’s machine-readable structure. AI performs best in systems that are explicit rather than implicit, structured rather than ad hoc, observable rather than opaque. RCX enforces these characteristics by design.
The Democratization of Complex Configuration
The sharpest claim Emil made concerned who can now operate these systems. Hanifin asked whether a strategist with no coding background could say “this is what I’m looking for, build me a system.”
Sarkissian was unequivocal: “Absolutely.” He cited a hackathon where a lawyer produced the best application because they understood what they wanted better than the engineers who had coded their entire careers. Knowing what to ask for is now the constraint. Not knowing how to build it.
Critically, the AI doesn’t accept vague instructions blindly. Systems like Claude Code will “do some research and come back and interview you. It’s good at eliciting a requirement.” It asks about segments, location restrictions, product groups — ensuring configuration matches actual business intent, not just the first thing someone typed.
The Economic Shift: From Infeasible to Inevitable
Too many organizations view AI through the lens of headcount reduction. Emil pushed back on that framing directly.
“What they’re failing to see is that, instead of that, you actually now have to reevaluate everything that wasn’t economically sensible before that now might have become economically sensible. Our job is really to reach out and re-engage even harder on those topics, on those strategy topics, to reevaluate whether this new world we’re just entering is suddenly opening up opportunities that have just been there all the time, but economically infeasible.”
The question isn’t “How can we do what we’re doing now with fewer people?” It is “What strategic initiatives have we been avoiding because the execution cost was prohibitive?” For loyalty programs, that means hyper-personalized promotion testing at scale, real-time program optimization on member behavior, and complex multi-partner coalition mechanics that were previously too intensive to manage.
The Non-Negotiables: Determinism, Reliability, and Accountability
Where vendors once competed primarily on features, enterprises now demand operational readiness and accuracy as baseline requirements. This is where the probabilistic nature of AI creates real tension with loyalty systems.
Sarkissian was emphatic: “These are deterministic systems, which means that the system has to work 100% of the time. AI is probabilistic in nature. It kind of works, but sometimes it makes mistakes. That’s not acceptable in the world of loyalty.”
ReactorCX keeps AI in an advisory role. The AI accelerates configuration, surfaces insights, and suggests optimizations. The deterministic engine guarantees accuracy. Those are two separate jobs and they stay separate.
“No one’s going to trade reliability, 100% accuracy and responsibility and accountability. That’s a purely human thing that you need still. AI will not take responsibility for millions and billions of dollars of liability that grow for these systems. If you don’t have the reliable, accurate, performant, scalable system to sit on, then all of this other stuff we just talked about is completely useless.”
Five Things This Conversation Gets Right About AI in Loyalty
The implications run beyond any single platform:
- 01 Architecture determines AI effectiveness because platforms built with explicit, machine-readable rules give AI a structured surface to work against. AI operates by generating and executing code. Platforms that expose a programmable surface give AI something to work with. Platforms that bury logic in proprietary UI do not.
- 02 The bottleneck is shifting from technical to strategic. When AI can translate business intent into technical configuration, the constraint moves from “Can we build this?” to “What should we build?” That elevates the role of strategists and diminishes the role of technical translators.
- 03 Economic feasibility boundaries are collapsing specifically for initiatives that required specialized developer time to configure and maintain. That cost is trending toward zero. The right response is strategic reevaluation, not a headcount reduction plan.
- 04 Determinism remains non-negotiable. AI must accelerate and assist, but it cannot replace guaranteed accuracy in financial systems. The winning approach keeps AI advisory while maintaining deterministic execution.
- 05 Integration capability matters more than ever. AI increases, rather than reduces, the need for platforms to integrate cleanly with existing enterprise systems. The value of AI multiplies when it can access and coordinate across POS, CRM, CDP, and partner systems without reconciliation overhead.
Architecture Is the New Competitive Signal
As Emil noted, “The cost of software development is basically trending to zero.” That doesn’t mean all platforms are equally positioned. The ones that will thrive were already disciplined in their architecture, their APIs, and their data models — well positioned to support the kind of explicit, structured interaction that AI requires.
For enterprises evaluating loyalty technology, the questions have changed. It is no longer sufficient to ask “Does your platform have AI?” The more important questions: Can AI access your system’s full capabilities programmatically? Is your rule structure machine-readable or locked in proprietary UI logic? Can AI operate within your governance framework, or does it require bypassing controls?
The loyalty industry is entering an era where strategic ambition no longer needs to be constrained by implementation complexity. But realizing that potential requires platforms designed for exactly this moment. As Sarkissian put it: “That’s the bread and butter of anyone who’s trying to run a serious program across a big brand. If you don’t have the reliable, accurate, performant, scalable system to sit on, then all of this other stuff is completely useless.”
The foundation still matters. It just unlocks far more value than it used to.