The 'Ghost Brand' Crisis: What I Learned Interviewing Sana Remekie
We are witnessing a shift in retail that is as exciting as it is existential. We all understand the potential of Composable Commerce and AI, but if I’m honest, the practical reality hasn’t quite matched the hype yet.
I recently sat down with Sana Remekie, CEO & Founder of Conscia.ai, to discuss this gap. My 'penny drop' moment came recently while analyzing AI styling assistants from major retailers like Zalando and Mango. At first glance, they look magical. But when I asked for a 'beach wedding guest outfit,' one recommended maternity wear. Another couldn't distinguish between specific color nuances, offering brown instead of green.
It made me wonder: Is this just a growing pain, or is something fundamentally missing?
Sana’s answer was a wake-up call. We aren’t just facing a tech problem; we are facing a 'Brand DNA' crisis.
Watch the full conversation here:
Why Your AI Is 'Hallucinating' Your Strategy
Sana pinpointed the root cause immediately: Organizational Semantics.
She explained that Large Language Models (LLMs) are trained on general internet data, not our specific product catalogs or brand voices. When an LLM answers a customer, it is essentially 'guessing' based on probability.
It doesn't inherently know that your brand defines 'orange' differently than 'gold'. More dangerously, it doesn't know your rules.
As I mentioned to Sana, a luxury brand might have a strict aesthetic rule: 'We never style shorts with flip-flops for an evening event.' An LLM doesn't know that. It just sees 'shorts' + 'summer' and makes a recommendation that violates the brand's prestige.
If we let LLMs just 'figure it out,' Sana warned that they will 'confidently hallucinate' an experience that damages the brand.
The Fix: The Orchestration Layer
So, how do we stop the hallucinations? Sana argued that you can’t just 'prompt engineer' your way out of this by stuffing your entire brand history into a chat window.
Instead, she introduced the concept of the Orchestration Layer.
Sana describes this as a 'rules engine' that sits between the AI and the customer. It balances the creative AI reasoning with deterministic control. You can program governance into this layer—literally coding your brand’s DNA into logic like: 'If context is Evening Event, then Flip-Flops = False'.
This layer ensures that the AI adheres to the decisions a Creative Director would make, rather than just guessing based on internet trends.
Google vs. The 'Black Box'
Perhaps the most strategic insight Sana shared was about where this data should live. We are moving toward a world where customers might not even visit our websites—they’ll just ask Google.
Sana observed a massive divergence in the market:
- Google is becoming 'Pro-Merchant': They are building tools (like Merchant Center and UCP) that allow us to explicitly define our data and rules. Sana’s advice was clear: 'Follow what Google is doing'.
- OpenAI is a 'Black Box': Currently, they offer less control for merchants over how data is ingested and presented.
If we don't structure our data for these agents, we risk becoming what I call 'Ghost Brands'—commoditized warehouses that just ship boxes while the AI owns the relationship.
My Takeaway: Intelligence + Infrastructure
This conversation clarified the path forward for me. We need two things to survive the 'Agentic' future:
- Deep Fashion Intelligence: This is what my team at Mapp Fashion has spent 10 years building—a taxonomy that decodes garments into specific styles, aesthetics, and occasions.
- Infrastructure: As Sana noted, we need the technical ability to expose these capabilities to third-party agents.
We have to stop locking our brand guidelines in PDF documents and start exposing them as APIs. It is time to teach the machines our Brand DNA, before they define it for us.
