What Foxconn CEO Sidney Lu's Three-Year Surprise Rule Reveals About AI-Era Luxury Brand Strategy
The consulting industry's agentic commerce playbook has a foundational assumption baked in that most brand teams haven't examined: that optimizing for AI-mediated discovery works the same way optimizing for human discovery does. Personalize the experience. Predict the preference. Deliver the relevant signal at the right moment. Conversion follows.
That assumption is almost certainly wrong, and the evidence for why comes from an unlikely source: a four-decade manufacturing executive at Foxconn Interconnect Technology who has never managed a luxury brand in his life.
McKinsey's recent interview with Sidney Lu, Chairman and CEO of Foxconn Interconnect Technology, surfaces a management philosophy that Lu has applied across four decades of building components most consumers will never see. Every three years, he says, deliver something that genuinely surprises, or risk becoming history. He didn't frame this as a marketing posture. He framed it as survival infrastructure.
That framing deserves more attention than it will get in the typical brand strategy conversation. Because McKinsey's separate analysis of AI and luxury experiences in the agentic age identifies exactly the problem Lu's rule solves, without recognizing it. The report's diagnosis is sharp: "the strategic question for merchants is no longer whether agents will mediate the journey, but who controls how brands are perceived." Its prescription is where things go sideways.
The Luxury AI Personalization Trap
McKinsey's agentic luxury analysis lands on a familiar recommendation: invest in AI personalization infrastructure, create human-centered experiences within the AI-mediated layer, and protect the emotional resonance that makes luxury legible. The underlying logic is intuitive. Luxury has always been about making people feel seen. AI personalization promises to make that feeling scalable.
The problem is that AI agents don't learn brand signals the way humans do. A luxury consumer builds a brand impression over years of exposure, social observation, and occasional purchase. They update that impression slowly, integrating new information against an accumulated baseline. The result is a relatively stable brand impression that can withstand significant turbulence before it degrades.
AI agents don't accumulate impressions. They optimize for behavioral signals in near real-time. What matters to an AI-mediated discovery system isn't the brand story the CMO believes in. It's the pattern of queries the brand attracts, the purchase triggers it generates, and the behavioral variance in how different consumer segments interact with it. Brands that generate predictable, consistent behavioral signals are easier for AI systems to categorize. And brands that are easy to categorize are, by definition, brands that AI systems can substitute for alternatives in the same category.
This is the structural problem that personalizing the experience doesn't solve. You can deliver a beautifully tailored AI-assisted luxury shopping moment and still be algorithmically commoditized, because the behavioral patterns your brand generates look like every other brand in your tier. The experience feels premium to the consumer. The signal looks undifferentiated to the system routing them.
What Sidney Lu Actually Understood
Lu's three-year surprise mandate wasn't designed for consumers. Foxconn Interconnect Technology's customers are procurement teams, engineering groups, and supply chain managers. These are not emotionally reactive buyers. They are systematic evaluators who care about specifications, reliability, and cost. The emotional levers luxury brands rely on simply don't apply.
What Lu understood is that in a supplier relationship, becoming predictable is becoming replaceable. If a buyer can model your capabilities with high confidence, they can also model what switching to a comparable supplier would cost. The decision to replace you becomes easier the more predictable you are. Delivering genuine surprise every three years, something the buyer couldn't have anticipated, resets that calculation. It forces the buyer to update their model of what you're capable of. That update is expensive to substitute.
This is a different kind of brand differentiation than most brand strategy frameworks describe. It isn't about distinctiveness in perception. It is about unpredictability in capability. Lu isn't trying to make buyers feel something particular about Foxconn. He's trying to make it structurally difficult for buyers to model Foxconn as a known quantity.
The application to AI-mediated brand discovery is direct. When an AI agent is routing a luxury consumer toward a purchase decision, the agent's model of what your brand represents is built from accumulated behavioral data. A brand that has generated consistent, predictable behavioral signals over time is a brand the agent can categorize with high confidence. High confidence categorization leads to high-confidence substitution recommendations. The agent knows what you are, which means it knows what else is like you.
A brand that has generated behavioral surprise, whose purchase triggers don't fit the standard pattern for its category, whose query co-occurrence shifts in ways that resist simple classification, is a brand the agent's model can't fully resolve. Algorithmic uncertainty isn't always a liability. In a competitive marketplace where AI systems are routinely recommending alternatives, being difficult to categorize cleanly can be a structural advantage.
Behavioral Signal Variance: The Metric Nobody Is Tracking
Here is the original framing that McKinsey's agentic luxury report doesn't surface: the relevant metric for AI-era brand strategy isn't brand coherence. It's signal variance.
Brand coherence, the consistency of positioning across touchpoints, has been the operational north star for brand management for decades. The logic is straightforward. Consumers who receive a consistent brand signal can form stable expectations. Stable expectations reduce purchase friction. Reduced friction increases conversion.
That logic holds in human-mediated discovery. It breaks in AI-mediated discovery. AI systems routing consumers toward purchases aren't optimizing for consumer coherence. They're optimizing for match quality between consumer behavioral patterns and brand behavioral patterns. A brand with extremely high signal coherence looks identical to every other brand with high signal coherence in the same category. The AI system's job is easier, which means the brand's job is harder.
Signal variance, the degree to which a brand generates behavioral patterns that resist clean categorization, is the mechanism that creates algorithmic friction in the substitution calculation. Brands with high signal variance attract queries from unexpected segments, convert through non-linear purchase paths, and generate behavioral data that is genuinely difficult to model. That difficulty is worth something. Not to consumers, who still benefit from coherent brand experiences, but to the AI systems that are increasingly deciding which brands to surface and which to substitute.
What Investment in Signal Variance Actually Looks Like
This distinction changes what brand investment looks like. The brand coherence playbook invests in consistency: unified messaging, disciplined channel management, controlled creative variation. The signal variance playbook invests in surprise: new category adjacencies that aren't obvious from existing brand positioning, partnership structures that generate unexpected behavioral overlap, product introductions that attract segments the brand's existing model doesn't predict.
Lu's three-year mandate is a structural commitment to signal variance generation at the manufacturing level. The timeline is shorter in AI-mediated brand discovery, because AI systems recalibrate faster than human memory does. But the underlying logic is identical. You have to generate behavioral data that forces a model update before the existing model calcifies into a substitution recommendation.
The behavioral science of intervention timing is relevant here. Research on stress management interventions in high-stakes exam settings demonstrates that an intervention at precisely the right moment before evaluation can change outcomes dramatically, not because it changed the underlying capability but because it changed the behavioral signal the evaluator encounters. Brands face a structurally similar dynamic. The moment an AI system is actively building or updating its category model is the highest-leverage moment for signal variance investment.
Dogs Trust and the Brand Refresh as Behavioral Data Event
Dogs Trust's recent bet on a significant brand refresh, building on a 2020 rebrand and framing the current moment as a "tipping point" toward household name recognition, is typically analyzed as a consumer awareness investment. The MarketingWeek coverage frames it in terms of audience-led pivot and growth trajectory. Those framings are accurate but incomplete.
What a significant brand refresh also generates is a burst of behavioral signal that forces AI systems to update their model of what the brand represents. A charity that has maintained consistent positioning for years has a clean, legible behavioral signature. Queries, clickthrough patterns, and conversion signals all fit the existing category model. That legibility is efficient for consumers finding the brand. It is also efficient for AI systems substituting alternatives.
A brand refresh that attracts new audience segments, generates coverage in contexts where the brand hasn't appeared before, and produces behavioral patterns that diverge from the established baseline forces recalibration across every discovery system that has a model of the brand. That recalibration is disruptive in the short term. It is also an opportunity to establish a new behavioral signature before the replacement model hardens.
Dogs Trust's investment in an "audience-led pivot" is significant in this context. If the pivot genuinely attracts segments that the brand's previous positioning didn't reach, it generates the kind of behavioral variance that resists clean categorization. The charity isn't only trying to grow awareness. It's forcing AI systems to update their models before those models become a ceiling.
This is the underappreciated strategic benefit of brand refresh investment in the current environment. It isn't only about consumer perception. It is about forcing AI discovery systems to recalibrate before their existing model of your brand becomes the default substitution reference.
The Decision Framework Emerging from This
The brand collapse dynamics visible in cases like Under Armour and Allbirds share a structural feature: both brands had highly coherent positioning in their peak moment, and both failed to generate the behavioral variance needed to resist algorithmic substitution as their category models hardened. Under Armour's performance identity was legible and consistent. It was also easy to route around once the AI systems mediating sporting goods discovery had a stable model of what Under Armour represented.
The decision framework that emerges from combining McKinsey's agentic analysis with Lu's surprise mandate isn't complicated, but it requires a vocabulary that most brand planning processes don't use.
The first question isn't "how do we make our positioning more consistent?" It's "what is the current variance in our behavioral signal, and is it sufficient to resist clean categorization by AI discovery systems?" Brands that have maintained tight positioning discipline for multiple years are the most vulnerable to algorithmic commoditization, not because their brand is weak but because their behavioral signature is too legible.
The second question is "what is our surprise mandate?" Lu set three years. For luxury brands competing in AI-mediated environments where models recalibrate continuously, the interval is probably shorter. The specific period matters less than the institutional commitment. Surprise needs to be scheduled, not aspirational. An organization that plans to deliver behavioral surprise every 18 to 24 months has to invest very differently than an organization optimizing for consistent positioning.
The third question is "where should we invest in signal variance without destroying brand coherence?" These aren't the same investment. Signal variance doesn't require incoherent messaging. It requires behavioral patterns that don't fit the existing category model cleanly. New adjacencies, unexpected partnerships, and product introductions that attract unfamiliar segments generate variance without requiring the brand to abandon its core positioning.
The broader agentic commerce analysis most holding companies are producing focuses almost entirely on the personalization infrastructure layer. That layer matters. But the brands that will sustain differentiation through the agentic transition are the ones that treat AI systems as a discovery environment requiring a specific strategic response, not just a channel requiring personalization tooling.
McKinsey is right that the question is no longer whether agents will mediate the luxury journey. The next question is harder: how do you remain difficult to substitute in a system that is very good at finding substitutes?
Sidney Lu has been answering a version of that question for four decades. The answer is the same whether you're building interconnect components or selling handbags. You have to deliver genuine surprise before the system finishes modeling you.
If your brand strategy team is navigating the shift to AI-mediated discovery, the STI decision intelligence framework offers a structured approach to signal variance analysis and agentic positioning. Details at smarttechinvest.com/research.