The Connection Drift: how AI agents are changing customer acquisition for marketing leaders

AI agents are reshaping how customers discover, evaluate and buy from brands. As tools like ChatGPT, Google AI Mode and B2B procurement AI increasingly mediate the customer journey, the moment a customer actually reaches your platform is becoming rarer and more valuable. This is the connection drift, and at our executive breakfast with Optimizely at The Century Club, senior marketing and digital leaders explored what it means in practice and what to do about it. 

Most organisations are responding to AI by doing what they've always done, only faster. More content, more automation, more optimisation of channels that are quietly becoming less relevant. The real disruption isn't about speed or efficiency. It's about where the customer relationship is now forming, and whether brands are even present at that moment of connection. 

The customer journey has already changed. Most brands haven't noticed. 

Our keynote speaker, Dr Nicola Millard, futurist and Principal Innovation Partner at BT, opened with a provocation: the customer journey isn't starting where brands think it is. 

She described a world where customers now fall into distinct modes: the 'shopper' doing independent research across multiple sources, the 'customer in crisis' who only engages directly when something's gone wrong, and the 'utilitarian' who just wants the transaction done fast. AI now mediates large parts of all three. 

Where AI gets genuinely complex (and this provoked real discussion in the room) is the emerging reality of bots talking to bots. A customer's personal AI agent interacting with a brand's AI agent, negotiating an outcome on the customer's behalf. The internet, as Nicola pointed out, was never architected for this. It was built for humans. And before we get into the full era of agentic commerce, there are unanswered questions that will matter enormously to brands: what happens when something goes wrong between two AI systems? Who's legally responsible? Would you trust your AI with your customer's credit card? Would they trust yours with theirs? These aren't hypothetical questions. They're the ones brands need to start answering now. 

She also drew a sharp line on personalisation that every marketer in the room will have felt. There's a sharp line, she argued, between a butler and a stalker. A butler notices your preferences, anticipates your needs, makes things feel effortless. A stalker collects everything, surfaces it at the wrong moment, and makes you feel watched rather than understood. The data to do both is increasingly available. The judgment about which side of that line you're on, and whether customers trust you enough to give you that latitude, is becoming one of the defining brand questions of the AI era. 

The implication for brands is significant. If AI is sitting in front of the early stages of the customer relationship, shaping perceptions, summarising your brand and filtering options, then the moment a customer actually reaches your platform carries far more weight than it used to. 

Nicola was followed by a panel that brought together Siobhan Cameron from Warner Hotels, John Prior from MSQ DX, and Mårten Bokedal from Optimizely, each bringing a different vantage point on how organisations are navigating this shift in practice. What followed was a candid, wide-ranging conversation that covered hospitality, B2B, content infrastructure, commerce protocols, and the very human question of why organisations keep getting stuck. 

Hospitality has seen this before. The rest of marketing is catching up. 

One of the most grounding perspectives came from Siobhan, Head of Digital Experience at Warner Hotels. The hospitality sector navigated something structurally similar years ago with online travel agents: a third party sitting between the brand and the customer, controlling discovery, shaping perception, and extracting margin. 

"It feels that AI is going to be a lot more impactful because OTAs were chosen by both the brand and the customer," she said. "AI disruption is something that's going to happen whether you've asked for it or not." 

She highlighted a concern that resonated around the room: many customers, particularly older demographics, will encounter AI-aggregated content without realising it. They'll assume what they're reading is either impartial or brand-owned. If it's unfavourable, or simply incomplete, the brand damage is real, quiet, and hard to trace. 

Warner's response has been deliberate: lean into intentional GEO (Generative Engine Optimisation), actively shape what gets surfaced in AI content, and double down on the quality of the direct experience when customers do arrive on owned channels. "If your brand is emotionally resonant enough, it will compel guests to look beyond the AI-aggregated content," Siobhan said. "If you invest in the real, AI becomes more of an amplifier than a threat." 

The moment of direct connection is getting rarer and more valuable 

MSQ DX Consulting Partner John Prior pushed the conversation toward what happens at the moment AI hands the customer over to you. By the time someone arrives on your platform via an AI-mediated journey, they're already qualified, but they've also been shaped by content and recommendations you didn't control. 

"When a customer arrives through an AI on-channel, they are going to be qualified but opinionated," John said. "Some of those opinions are going to be driven by content you didn't create. The temptation is to replicate what AI has given them. You've got the opportunity to be something different and better." 

That means leading with reassurance, delivering on the specific promise that got them to your door, and building a relationship that extends well beyond the transaction. In B2B particularly, John emphasised the value of a strong CRM foundation: knowing everything you legitimately can about that customer before they need to tell you again. 

Content Must Work for Two Audiences Now: Humans and Machines 

Mårten brought the platform perspective. "The behaviour is changing," he observed. "The discovery phase is changing. The most used agent our customers run on our CMS is our GEO optimisation agent, which looks at pages to see if they're actually optimised for AI to read them. So many sites are not even readable today by AI systems." 

The challenge he described is a dual-audience content problem. Search queries are becoming more specific and conversational. Customers are asking AI systems detailed, context-rich questions rather than typing two-word searches. Content designed to rank in traditional search, or simply to persuade a human reader, may be failing entirely to inform the AI intermediaries that are now shaping initial consideration. 

This isn't about choosing between human-first and machine-first content. It's about building infrastructure that serves both simultaneously, and at scale. That's a harder problem than most content teams are currently set up to solve. 

The Commerce Protocol Chaos: Don't Bet on One Winner 

One of the sharpest exchanges came around OpenAI's Agentic Commerce Protocol and Google's Universal Commerce Protocol: two competing frameworks that could fundamentally change how purchases happen online by routing them entirely through AI platforms. 

John's advice was pragmatic: don't try to pick the winner. "It's illustrative of the broader point: it's not about betting on Google or OpenAI, but always betting on what has the ability to adapt." He noted the pace of change is unlike anything marketing has faced before. "We used to have years to adapt. Now we have the length of a sales cycle, which can be weeks rather than months." 

The People Problem: Why Organisations Get Stuck 

Nicola offered a blunt diagnosis of why so many AI initiatives stall. It's not technology. It's data, people, and processes. In that order. 

She pointed to the untapped intelligence sitting in contact centres: customer conversations that reveal product and service issues in real time, but that marketing teams rarely access. Sentiment analysis, language patterns, complaint clustering. These are rich leading indicators that most organisations simply aren't treating as marketing intelligence. 

Mårten echoed this from the platform side. Organisations are still buying and evaluating technology the way they did five years ago, far too slowly for a landscape that shifts in months. And without a clear picture of how time is actually spent, it's nearly impossible to make a credible case for what AI is saving. 

Mårten's broader challenge to the room was to stop trying to apply AI to existing processes and instead ask a more fundamental question: what could we now do that we've never been able to do before? 

The Four Challenges Every Marketing Leader Is Wrestling With 

What followed were candid roundtable discussions on the connection drift. Attendees split into intimate groups, facilitated by the MSQ DX team, to share what they're working through in their own organisations. Four themes surfaced consistently across every table. 

The first was data foundations. Before any meaningful AI strategy can take shape, organisations need clarity on what data they actually have and whether it's in good enough shape to act on. AI amplifies good data infrastructure and exposes weak foundations fast. 

The second was the investment tension. Senior leaders are actively wrestling with what to prioritise now versus what to build position for later. With real uncertainty about where the AI landscape will settle, the risk of over-committing to the wrong thing feels as live as the risk of under-investing. 

The third was customer context. As AI mediates more of the discovery journey, customers arriving on owned platforms will increasingly be more qualified, more opinionated, and less patient. Organisations that aren't ready to meet that version of the customer will lose them quickly. 

The fourth was trust. It surfaced at every table and from every angle: customer trust in AI-generated brand content, organisational trust in AI outputs, and the question of whether brands are showing up authentically enough in AI-mediated environments for that trust to hold. 

One observation from the closing synthesis stood out: brands may actually be becoming more important again. In a world where AI smooths over functional differentiation, emotional resonance and brand clarity may be what tips an AI recommendation in your favour. 

Will Brands Have More or Less Control in Two Years? 

We put this to the panel as a final question. Here's where each of them landed. 

Siobhan saw more opportunity but less control. There will be more touchpoints and more ability to personalise at scale, but the beginning of the relationship, where trust is established, will be increasingly outside a brand's hands. 

John predicted the discovery space will remain fragmented and fast-changing for the foreseeable future. Even small version changes in AI models alter how brand content is surfaced and interpreted. Post-connection, though, brands with strong foundations and deep customer understanding will have far more opportunity than they do today. 

Mårten believed we're heading back to genuine customer-centricity. The organisations that will win, he argued, are those building cultures that can adapt, not just those deploying the best technology. 

Nicola offered perhaps the most grounding perspective: "I don't think companies have been in control of their brand for about 15 years. Customers talking to other customers: that's your brand. AI just adds an extra layer of complication." Her call to action: stop asking how we take control back, and start asking how we help customers become our biggest advocates. 

The Connection Drift Is Already Happening 

What struck the MSQ DX team most about this conversation was the absence of panic. Yes, the landscape is shifting faster than traditional planning cycles can accommodate. Yes, the old assumptions are eroding: that customers will visit your website, that you control the first impression, that your funnel is yours. But the leaders in that room aren't frozen. They're adapting. 

The consensus that emerged was clear. Invest in your foundations: data, content structure, brand clarity. Design for the moment when a customer finally reaches you directly, because that moment is now doing more work than it ever has. And build the organisational capability to adapt quickly, rather than betting everything on predicting which AI protocol wins. 

The connection is drifting. The question isn't whether to acknowledge it. It's how to thrive despite it. 

Ready to Navigate the Connection Drift? 

MSQ DX does one thing: translate digital experience into commercial impact. We partner with complex organisations to design revenue-generating digital products, create experiences that convert and retain, and accelerate transformation through senior-led, AI-enabled delivery. 

If the themes from this event resonate with your organisation, we'd love to keep the conversation going. MSQ DX are offering a free AI Enablement Workshop to help you map your own connection drift: identifying where customer relationships are shifting, and where your most valuable moments of direct connection actually lie. 

Frequently Asked Questions 

Who is MSQ DX and what do they specialise in? 

MSQ DX does one thing: translate digital experience into commercial impact. Operating across the UK, Europe, and the US, MSQ DX partners with complex organisations to design revenue-generating digital products, create experiences that convert and retain, and accelerate transformation through senior-led, AI-enabled delivery. Digital defines businesses today. Success is measured by the experiences you create, the precision of your execution, and the speed at which you adapt. MSQ DX exists to help organisations translate digital opportunity into measurable outcomes. With over 500 experts globally, MSQ DX delivers measurable growth, sustained efficiency, and real-world outcomes. Technology agnostic by design, MSQ DX works with all leading platforms globally including Optimizely, Umbraco, Storyblok, Contentful, Kentico, Sitecore, and Shopify Plus. Trusted by Bosch, BCG, Caesars Entertainment, CFA Institute, Porsche, Vaillant, and Vodafone. Part of MSQ, one of the fastest-growing global creative, media and technology groups, with 1,850 people across 24 offices in Europe, North America, and Asia. 

What is the 'connection drift' in digital marketing? 

The connection drift refers to the structural shift in where meaningful customer connections happen. For two decades, brands assumed customers would visit their website during the purchase journey. AI agents, from ChatGPT to Google AI Mode to B2B procurement tools, are increasingly mediating discovery, consideration, and even purchase before a customer ever reaches a brand's own channels. The point of connection is drifting away from owned and controlled touchpoints. 

What is Generative Engine Optimisation (GEO) and why does it matter? 

GEO is the practice of optimising your content so that AI systems, such as the large language models powering search and discovery tools, accurately represent and preferentially recommend your brand. Unlike traditional SEO, which focused on ranking in search results, GEO focuses on how AI agents describe you, the sentiment they carry about your brand, and whether they actively recommend you over competitors. As AI search grows, GEO is becoming a foundational marketing capability. 

How are AI agents changing B2B marketing and procurement? 

B2B procurement AI is already scanning vendor websites, building capability profiles, and generating shortlists before human buyers see an RFP. This means a brand's first meaningful touchpoint with a buyer can now happen after an AI has decided whether they're worth considering. Content designed to persuade human readers may be failing entirely at informing AI intermediaries, making structured, machine-readable content increasingly important in B2B strategy. 

What is OpenAI's Agentic Commerce Protocol and how does it affect brands? 

OpenAI's Agentic Commerce Protocol (and Google's Universal Commerce Protocol) are frameworks that enable end-to-end purchasing to happen inside AI platforms, without a customer visiting a merchant's website. This means a customer could research, compare, and buy entirely within ChatGPT or Google AI Mode. Brands that aren't structured to participate in these protocols risk being invisible at the moment of transaction. However, the protocols are still evolving rapidly, and experts advise building adaptive capability rather than betting on any single winner. 

How should brands respond to AI-mediated customer discovery? 

The most consistent advice from senior leaders at our event was threefold: actively shape what AI surfaces about your brand (intentional GEO), invest in the quality of the direct experience when customers do arrive on your own platform, and treat the moment of direct connection as more valuable than it has ever been. Brands that deliver exceptional experiences and generate genuinely positive customer sentiment will find AI amplifying rather than threatening their reputation. 

What does 'dual-audience content' mean in AI marketing strategy? 

Dual-audience content is content designed to serve both human readers and AI systems simultaneously. Human-optimised content focuses on emotional resonance, storytelling, and conversion. Machine-optimised content focuses on structured information, semantic clarity, and factual accuracy that AI agents can accurately interpret and relay. As AI intermediaries play a larger role in customer discovery, content that only serves one audience is increasingly insufficient. 

What is the role of Optimizely Opal in navigating the connection drift? 

Optimizely Opal is an AI orchestration platform designed to help brands deliver intelligent, personalised experiences across fragmented digital touchpoints. In the context of the connection drift, Opal enables organisations to maximise the moments of direct customer connection they still control, adapting content, workflows, and experiences in real time, even with limited data about how a customer arrived. Its composable architecture also allows brands to adapt quickly as AI protocols and commerce patterns continue to evolve. 

Will brands have more or less control over customer relationships as AI grows? 

The honest answer from our panel: more opportunity, less control. Brands will have unprecedented ability to personalise and serve customers at scale, but the top of the funnel, where trust and first impressions are formed, will increasingly be shaped by AI systems rather than brands themselves. The leaders most likely to succeed are those building customer-centric cultures, strong data foundations, and the organisational agility to adapt as the landscape continues to shift.