There is a conversation happening across marketing teams right now, and it goes something like this: “We’ve implemented AI for personalisation. Now what?” It’s a question Yann Gautier, Chief Consulting and Transformation Officer at WoolfHodson, hears frequently – and while it’s reasonable, he’d argue it’s the wrong one.

Personalisation has become table stakes. And yet, despite the sophistication of the tools now available, customer fatigue is accelerating, not receding. Inboxes are fuller. Engagement is declining. And somewhere in the middle of all this optimised outreach, the customer has started to feel less understood, not more.

What I mean and don’t mean by empathy

Predictive empathy isn’t a soft concept dressed up in tech language. It’s the ability of AI to infer where a customer is, emotionally and contextually, before they signal it explicitly. This is distinct from personalisation, which answers the question of what to say. It’s distinct from relevance, which answers when to say it. Predictive empathy is concerned with whether to say anything at all, and why – in a way that genuinely respects the customer’s state of mind in that moment.

It draws on a combination of behavioural signals, timing, lifecycle stage, and external context. Not a single data point but a confluence. And when it works, it doesn’t feel like marketing. It feels like being known.

Why I believe it matters now

What the industry has actually built, in most cases, is hyper-personalised spam. AI-powered outreach that knows a prospect’s name, their industry, their likely pain points – and still manages to arrive at the wrong moment, in the wrong tone, at the wrong frequency.

Volume optimisation through AI is a real thing, and it is making the problem worse. The machinery of modern demand generation is extraordinarily good at sending more. What it hasn’t yet learned to do is send less, better. The marginal cost of one more AI-generated, personalised send is effectively zero. Recipient annoyance is an unpriced externality. When the cost of sending is zero, and the penalty for being unwanted is also zero, the system is not malfunctioning – it’s doing exactly what it has been asked to do.

B2B buyers, and the buying committees around them, decide emotionally and justify rationally, just like every other human being. AI that can read buying committee sentiment, not just engagement scores, changes the game entirely. But most teams are not building toward that capability. They are optimising for outreach efficiency when the real opportunity sits in deeper customer understanding, a direction that remains largely unexplored.

The shift from intent data to emotional context signals is where the real unlocks are. And it exposes a structural problem: organisations are rushing to AI tool selection without first asking whether the AI readiness is genuinely in place.

Leadership mandates an AI agenda, a pilot is set up to demonstrate momentum, but nobody has asked the more fundamental question – does this organisation have the data quality, process maturity, and team capability that AI actually needs in order to deliver? The result is predictable: adoption stalls, ROI doesn’t materialise, and the most likely conclusion is that AI was the problem, not the environment it was deployed into.

What’s holding us back

Most data infrastructure was built for behavioural tracking. It is exceptionally good at recording what someone did: pages visited, emails opened, forms submitted. It is not built for what those behaviours might mean in aggregate, contextually, at a human level. A prospect who opens three emails and then goes silent is not a cold lead. They may be overwhelmed, under internal pressure, or close to a decision and waiting. The system doesn’t distinguish. It sees a drop in engagement and fires the next sequence.

Marketers have become fluent in setting AI systems against conversion goals – click rates, pipeline contribution, form-fill rates. Almost no one is briefing AI against empathy goals. Not because they don’t value empathy, but because they have never been asked to operationalise it. That is a skill gap at the practitioner level. It is also a failure at the industry level. Both are true, and neither has been seriously addressed.

And underneath all of this is a challenge that has persisted for 15 years without resolution: teams are being asked to layer AI innovation onto unresolved data-quality and process gaps. Most organisations are using around 20% of what their existing platforms can do – the rest switched off, misconfigured, or never properly set up. The ambition to add AI on top is there; however, the foundation it needs isn’t.

What good looks like

None of this needs technology that doesn’t exist. It needs teams to use the signals they already collect.

Start with engagement-quality scoring. Weigh how someone engages, not just whether they do: dwell time, repeat visits, depth of content consumption. A prospect who reads to the end and returns twice is not the same as a four-second bounce. Most teams already capture this data and flatten it into a single number. Rebuilding that scoring model is mostly a configuration change, not a platform investment.

Add lifecycle triggers from real-world change. Role changes, restructures, funding rounds, and expansion signals – these are already sitting in the enrichment data most teams pay for. The inference is straightforward. What is missing is the discipline to act on it. Running a static nurture track while a prospect’s world changes around them is not a technology problem. It is a decision about what the programme is actually trying to do.

Finally, be mindful of overload. When engagement drops sharply, the instinct is to fire the next sequence. The better response is to hold back until conditions look right. This is not about inferring what a prospect is feeling. It is about reading an ambiguous signal correctly and resisting the pressure to fill silence with volume.

My take on predictive empathy

The problem with AI-powered marketing is not the AI. It is what the AI has been asked to optimise for. The organisations that get this right will not have done so by purchasing a new platform. They will have done so by deciding, at a strategic level, that empathy is a measurable outcome, not a philosophy, and then building their AI strategy around that conviction. Critically, they will have done the readiness work first: establishing the data infrastructure, the process maturity, and the internal capability that gives AI the conditions it needs to succeed, rather than deploying it into an environment that will quietly defeat it.

Organisations that build empathy into their AI strategy now, and do the unglamorous foundational work to make it viable, will rebuild the trust that years of hyper-personalised outreach quietly eroded – not by pretending to know the customer, but by demonstrating through restraint and relevance, that the signal has been read correctly. The industry has spent a decade learning to speak to customers at scale. The next decade belongs to those who learn to listen.

——

Yann Gautier is Chief Consulting and Transformation Officer at WoolfHodson, a B2B marketing technology consultancy that helps organisations build data-led, technology-enabled marketing programmes. With more than two decades of international experience, he is widely regarded as a leading authority in digital and marketing transformation, having led large-scale programmes across the Americas, Europe, and Asia Pacific for global enterprises including Unilever, The Coca-Cola Company, Colgate-Palmolive, JPMorgan, Liberty Mutual, and JLL.