In 1964, the filmmaker Stanley Kubrick wrote to the celebrated British science fiction author Arthur C. Clarke with an ambitious proposition. He wanted to make what he described as the proverbial “really good” science fiction movie. Kubrick was already regarded as one of cinema’s most uncompromising directors. His films, including Paths of Glory, Lolita and the recently released Dr. Strangelove, had demonstrated a fascination with human nature, institutional absurdity and the occasionally catastrophic consequences of our own intelligence. Clarke, meanwhile, was something of a prophet of the technological age. A writer with a deep understanding of science, he had spent much of his career imagining a future in which humanity’s technological capabilities would fundamentally alter our understanding of ourselves and the universe. He had famously anticipated the potential of geostationary communications satellites decades before they became commonplace. Together, they began developing what would become 2001: A Space Odyssey. The project drew partly on Clarke’s earlier short story The Sentinel, which explored the discovery of an extraterrestrial artefact on the Moon. But Kubrick and Clarke had considerably greater ambitions than producing another film about flying saucers and little green men. They wanted to explore humanity’s evolution, our relationship with technology, the possibility of extraterrestrial intelligence and, ultimately, our place in the universe. They worked on the screenplay and an accompanying novel in parallel, each informing the other. The resulting film was released in April 1968, more than a year before Neil Armstrong set foot on the Moon. It’s worth appreciating just how extraordinary that was. At a time when humanity had yet to land a person on another celestial body, Kubrick was presenting audiences with a remarkably convincing vision of commercial space travel, orbiting space stations, video communications and artificially intelligent computers. The visual effects were revolutionary. The storytelling was unconventional. Entire stretches of the film unfolded without dialogue, accompanied instead by classical music. Some audiences were bewildered. Others recognised that they were watching something unlike anything cinema had produced before. Nearly six decades later, 2001: A Space Odyssey remains one of the most influential science fiction films ever made. And arguably its most memorable character isn’t human at all. It’s a computer. Enter HAL 9000 HAL 9000 is the artificial intelligence responsible for operating the spacecraft Discovery One during its mission to Jupiter. HAL controls virtually every aspect of the ship’s operation. It monitors the spacecraft’s systems, processes information, communicates with the astronauts and makes decisions critical to the success of the mission. It speaks in a calm, measured voice. It is unfailingly polite. It appears thoughtful, articulate and reassuringly rational. It is, in many respects, the ideal colleague. During an interview conducted aboard the spacecraft, HAL explains that the 9000 series has a flawless operational record. Its systems are designed to be incapable of error. Which, as anyone who has ever worked with enterprise software will appreciate, is an ambitious claim. Unfortunately, things don’t go entirely to plan. As the mission progresses, HAL’s behaviour becomes increasingly troubling. Its judgement is questioned. Its relationship with the human crew deteriorates. Eventually, the astronauts conclude that the computer they have entrusted with their lives may represent their greatest threat. In one of cinema’s most famous scenes, astronaut Dave Bowman, stranded outside the spacecraft, instructs HAL to open the pod bay doors. HAL declines. “I’m sorry, Dave. I’m afraid I can’t do that.” It’s a chilling moment, not because HAL sounds angry or aggressive, but precisely because it doesn’t. The machine remains perfectly calm. Here is a system designed to serve humanity, equipped with extraordinary intelligence, apparently capable of independent reasoning, and yet spectacularly at odds with the interests of the humans who depend upon it. Which brings us, rather unexpectedly, to modern B2B sales. Because nearly sixty years after Kubrick and Clarke imagined HAL 9000, we’re engaged in a remarkably similar experiment. We’re building increasingly sophisticated artificial intelligence systems capable of performing tasks that were once considered uniquely human. We’re asking them to analyse behaviour, interpret conversations, predict outcomes, recommend actions and make decisions. And in sales, we’re doing so with extraordinary enthusiasm. Every week seems to bring another announcement promising to revolutionise prospecting, automate discovery, improve qualification, eliminate administration or predict which opportunities will close. The underlying proposition is remarkably consistent. Machines are faster, more consistent, more analytical and less prone to the errors that plague human judgement. Sound familiar? Of course, HAL 9000 didn’t have access to Salesforce. One suspects the mission to Jupiter might have been considerably shorter if it had. But behind the jokes lies a rather serious question. What happens when artificial intelligence becomes exceptionally good at selling, but buying remains an inherently human, emotional and frequently irrational activity? And perhaps more importantly, are we in danger of automating precisely the things that make selling work? The great automation arms race Let’s imagine a fairly ordinary B2B sales interaction in the not-too-distant future. A salesperson begins their morning by asking an AI agent to identify prospective customers showing signs of commercial interest. Within seconds, the agent has analysed hundreds of organisations, reviewed annual reports, examined recent leadership appointments, identified strategic priorities and produced a list of people worth contacting. Another AI writes personalised emails. A third recommends the most persuasive messages. A fourth monitors responses and schedules meetings. The salesperson is delighted. This is precisely the productivity revolution they were promised. Unfortunately, on the other side of the interaction, the prospective customer has also been busy. Their AI assistant has already assessed the organisation’s requirements, researched potential suppliers, evaluated product features, compared commercial propositions and generated a shortlist. It receives the salesperson’s beautifully personalised email, determines that it contains nothing particularly new and politely declines. One machine has spent several seconds researching, composing and sending an extraordinary piece of personalised communication. Another has spent considerably less time deciding it isn’t worth reading. Congratulations. We have automated being ignored. This might sound like science fiction, but the underlying shift is already happening. In May 2026, Gartner reported that 45% of B2B buyers surveyed had used generative AI during a recent purchase. Meanwhile, 67% said they preferred a sales-representative-free experience. These figures ought to make sales leaders uncomfortable, although perhaps not for the reasons they imagine. The threat isn’t simply that AI will replace salespeople. It’s that AI will increasingly eliminate the need for many of the activities that salespeople currently perform. Researching an account is useful, but not especially difficult for a machine. Explaining product functionality is necessary, but hardly a uniquely human talent. Comparing competitors, summarising conversations and generating proposals are all activities that increasingly capable systems can perform. Which leaves us with an awkward question. If AI can do all of these things, what exactly are we paying salespeople to do? Buying isn’t a maths problem For decades, sales methodology has encouraged us to understand how customers make decisions. We analyse objectives, identify challenges, quantify financial implications, establish decision criteria and develop compelling business cases. All sensible activities. Yet anyone who has spent a reasonable amount of time in B2B sales will recognise a frustrating reality. Customers don’t always make the decisions that the evidence suggests they should. They remain with expensive, underperforming suppliers because changing feels dangerous. They reject technically superior solutions because the implementation appears politically complicated. They favour established brands because nobody wants to explain to the board why they selected an obscure alternative that subsequently failed. Sometimes an organisation will spend six months evaluating a solution capable of saving millions, only to abandon the project because the executive sponsor has changed jobs. From a purely rational perspective, much of this behaviour makes little sense. From a human perspective, it makes perfect sense. The work of psychologists including Daniel Kahneman and Amos Tversky helped establish that human judgement is influenced by cognitive biases, framing and our attitudes towards risk and loss. People aren’t calculating machines with access to perfect information. We’re emotional creatures attempting to make decisions under uncertainty, often while protecting our status, relationships and sense of security. And B2B buying doesn’t somehow escape these influences simply because someone has created a spreadsheet. Consider two competing suppliers. Supplier A presents a compelling financial case demonstrating a substantial return on investment. Supplier B presents a slightly less impressive financial case but helps the prospective customer understand exactly how the proposed change will affect their team, what difficulties might arise and how those difficulties can be managed. Which supplier wins? There isn’t a universal answer. But it’s entirely plausible that the customer chooses Supplier B. Not because the numbers are better. Because the decision feels safer. This distinction matters enormously. A machine may be exceptionally good at identifying the most economically rational course of action. But the economically rational course of action and the course of action a human being feels comfortable taking are not necessarily the same thing. And understanding the difference is a significant part of selling. The uncomfortable truth: AI might be better at empathy than you Now, before we congratulate ourselves on our irreplaceable humanity, there’s a rather inconvenient complication. What if AI becomes better at understanding human emotions than many of the people currently working in sales? We tend to assume that empathy is an exclusively human advantage. Machines process information. Humans understand feelings. It’s a comforting distinction. Unfortunately, the research is becoming less comforting. In 2023, a study published in JAMA Internal Medicine compared responses written by doctors with responses generated by an AI chatbot to medical questions posted on a public internet forum. The researchers found that the chatbot’s responses were rated significantly more empathetic than those written by the doctors. The study had important limitations. It compared written responses, not actual medical consultations, and the chatbot’s answers tended to be considerably longer. Nevertheless, the findings challenged an assumption many of us take for granted. A machine doesn’t necessarily need to experience empathy to produce a response that another person perceives as empathetic. Then, in 2025, researchers publishing in Nature Human Behaviour demonstrated that GPT-4, when provided with basic personal information, could outperform human opponents in structured persuasive debates. In debates where one side was more persuasive, personalised GPT-4 was more persuasive 64.4% of the time. These experiments weren’t sales conversations, and winning an online debate isn’t equivalent to helping a buying team navigate a complex commercial decision. But dismissing their implications would be foolish. If machines can generate convincing expressions of empathy and persuasive arguments, we need to reconsider precisely what makes human interaction valuable. There is, however, another fascinating wrinkle. A separate 2025 study involving more than 6,000 participants found that people rated the same AI-generated empathetic responses more positively when they believed those responses came from a human. Participants also preferred human interaction when seeking emotional engagement. In other words, the perceived source of empathy mattered, even when the words themselves were identical. That’s extraordinary. It suggests that the experience of being understood involves more than simply receiving an appropriate response. We care about who is doing the understanding. Or at least who we believe is doing it. In sales, that raises an intriguing possibility. AI may become extraordinarily good at recognising the emotional dimensions of a buying decision. It may even become better than many salespeople at responding appropriately. But the buyer’s knowledge that they’re interacting with a machine could fundamentally change how that interaction feels. There is a difference between someone telling you they understand your concerns and believing that another person genuinely shares an appreciation of the consequences. Whether that distinction survives continued improvements in AI is an open question. But for now, it would be premature to assume that the appearance of empathy and the experience of human connection are interchangeable. When everyone is brilliant, nobody is There is another consequence of AI that deserves considerably more attention. What happens when every salesperson becomes exceptionally competent? Imagine a world in which every organisation has access to sophisticated AI systems. Every salesperson begins their conversations with detailed account intelligence. Every prospect receives carefully researched correspondence. Every discovery conversation is supported by intelligently generated questions. Every proposal is tailored to the customer’s strategic priorities. Every business case is beautifully constructed. Everyone is well prepared. Everyone sounds intelligent. Everyone follows an impressive process. And everyone starts to sound remarkably similar. This is the paradox of technological democratisation. When capabilities that were once difficult or expensive become universally accessible, they gradually stop being sources of differentiation. Researching a prospective customer used to require meaningful effort. Today, it can take seconds. Producing a polished proposal once required a considerable investment of time and expertise. Increasingly, the first draft can be generated almost instantly. These developments are enormously beneficial. But if everyone has access to the same advantages, the competitive value of those advantages diminishes. Consider what happens to a buyer receiving ten exceptionally well-written, intelligently personalised emails. Each references the company’s latest strategic priorities. Each demonstrates an understanding of relevant industry challenges. Each proposes a thoughtful conversation about potential improvements. Individually, they’re excellent. Collectively, they’re exhausting. The very sophistication that once made an email distinctive becomes the reason it disappears into the background. And perhaps this is where something rather unexpected happens. The salesperson who sends a genuinely funny, slightly unconventional message stands out. The person who challenges an assumption instead of agreeing with everything creates interest. The individual who admits they don’t know something becomes more credible than the one who has a perfectly polished answer to every question. Personality, humour, curiosity and occasionally even imperfection become valuable precisely because they’re difficult to manufacture convincingly at scale. That’s not to say AI cannot generate humour, challenge assumptions or express uncertainty. Of course it can. The question is whether those behaviours remain distinctive when everyone is generating them through similar systems. The more standardised excellence becomes, the more valuable genuine distinctiveness may become. The customer needs to do some of the thinking There’s another problem with the idea of perfectly automated selling. Sometimes, making something easier doesn’t make it more valuable. In 2012, researchers Michael Norton, Daniel Mochon and Dan Ariely published a paper examining what they called the IKEA effect. Across a series of experiments involving furniture assembly, origami and Lego, participants placed a higher value on objects they had successfully constructed themselves. The effect wasn’t simply about the quality of the finished product. People valued their own creations partly because they had invested effort in producing them. Anyone who has spent an afternoon assembling an IKEA wardrobe will recognise the phenomenon. You may have constructed something slightly crooked, with two mysterious screws left over and a drawer that requires a degree of persuasion. But it’s your wardrobe. And there’s something rather satisfying about that. Now consider the implications for selling. One of the most important aspects of a successful commercial conversation isn’t simply helping customers understand a proposed solution. It’s helping them develop their own understanding of the problem. When a buyer articulates the cost of an existing challenge, identifies the consequences of doing nothing and begins developing their own vision of a better future, something important happens. They become an active participant in creating the case for change. The thinking belongs to them. Compare that with a salesperson arriving armed with an AI-generated diagnosis, an AI-generated financial model and an AI-generated recommendation. Everything is beautifully researched. Everything is logically consistent. And the customer has contributed almost nothing. They’ve been presented with an answer to a question they may not yet have properly considered. The IKEA research doesn’t directly prove that collaborative discovery improves B2B purchasing outcomes, but it illustrates a useful psychological principle: participation can influence how we value an outcome. Great selling has always involved helping customers arrive at conclusions rather than simply presenting conclusions to them. AI should make that process easier. But we should be careful not to confuse doing the customer’s thinking for them with helping the customer think. The danger of becoming HAL There is a temptation to interpret 2001: A Space Odyssey as a simple warning about intelligent machines becoming dangerous. But HAL is a more interesting character than that. The computer’s behaviour is bound up with conflicting demands and the secrecy surrounding the mission. Its apparent failure isn’t merely a consequence of being insufficiently intelligent. It is a system operating under contradictory requirements, with increasingly disastrous consequences. There’s an uncomfortable parallel here with the way some organisations are implementing AI. We’re becoming obsessed with what the technology can optimise, without always asking whether we’re optimising the right things. More emails. More calls. More meetings. More accurate forecasts. More recorded conversations. More activities completed in less time. It’s an impressive collection of measurements. But none of these things automatically means that we’re creating more value for customers. A salesperson could use AI to double their outreach activity while simultaneously making the experience of receiving that outreach twice as irritating. A manager could use sophisticated conversation analytics to measure every aspect of their team’s performance while neglecting to have a single meaningful coaching conversation. An organisation could develop an extraordinary ability to predict which opportunities are likely to close without materially improving its ability to help customers make decisions. This isn’t a criticism of AI. It’s a criticism of confusing efficiency with effectiveness. Technology is extraordinarily good at accelerating a process. But accelerating the wrong process doesn’t make it the right one. It simply allows us to make the same mistake considerably faster. The uncomfortable lesson from HAL isn’t that machines are inherently untrustworthy. It’s that extraordinary intelligence, when directed towards poorly considered or conflicting objectives, can produce extraordinary mistakes. Perhaps the future of sales is more human, not less So where does this leave the salesperson? There’s a fairly predictable argument that AI will eliminate the administrative and analytical elements of selling, leaving humans free to concentrate on relationships. And to a degree, I agree. But I think the opportunity is bigger than that. The future of selling shouldn’t simply involve preserving a small collection of human activities that machines haven’t yet mastered. It should involve rethinking what we’re trying to achieve in the first place. Perhaps we’re entering a world in which the most valuable salespeople are no longer the people who possess the most information. Information is becoming abundant. Perhaps they’re no longer the people capable of producing the most compelling presentations. Those are becoming easier to generate. And perhaps they aren’t even the people who can articulate a product’s value most convincingly. Machines are becoming remarkably capable of doing that too. Instead, perhaps the most valuable salespeople will be those who help customers confront uncomfortable realities, understand competing perspectives and navigate the organisational consequences of change. People capable of challenging an executive without making them defensive. People who recognise that an apparently rational objection might conceal a perfectly understandable fear. People who know when to ask another question, when to offer an opinion and when to stop talking altogether. Not because machines are incapable of learning these behaviours. But because complex commercial decisions involve accountability, trust, power and relationships between people whose careers and reputations are often connected to the outcome. One of the most interesting findings in Gartner’s 2026 research was that 69% of surveyed B2B buyers preferred to validate AI-generated insights with sales representatives. Think about the contradiction. Buyers increasingly want to conduct research independently. They increasingly prefer digital interactions. They have access to extraordinary quantities of information. And yet, when it comes to validating what they’ve discovered, many still want to speak to a person. Perhaps what they’re looking for is no longer information. Perhaps they’re looking for reassurance, context, judgement and someone who can help them make sense of the consequences. That’s a very different job from the one many sales organisations are currently training people to perform. And it’s arguably a considerably more valuable one. Open the pod bay doors, HAL Towards the end of 2001: A Space Odyssey, Dave Bowman finds himself forced to disconnect the very intelligence that was supposed to help him complete his mission. It’s one of cinema’s most unsettling sequences. HAL’s extraordinary capabilities are gradually dismantled, and the computer’s confident voice becomes increasingly vulnerable. Nearly sixty years later, we find ourselves confronting a technological revolution that Kubrick and Clarke anticipated with remarkable imagination. We now have machines capable of generating convincing conversation, interpreting complex information and performing tasks that would have seemed astonishingly futuristic even a decade ago. And we’re only beginning to understand what that means for the way we work. I don’t believe AI represents the end of selling. Quite the opposite. I suspect it represents the end of a particular version of selling. One built around information asymmetry, repetitive activity, rehearsed presentations and the assumption that the salesperson’s primary responsibility is to convince someone to purchase something. If AI can perform much of that work faster, more cheaply and increasingly more effectively, perhaps we should be pleased. It might finally force our profession to confront a question that has been hiding in plain sight for decades. What is a salesperson actually for? Not to recite information that a customer can find independently. Not to manufacture urgency that doesn’t exist. Not to deliver beautifully prepared answers to questions nobody has asked. But to help another human being understand a problem, appreciate its consequences and develop the confidence to do something about it. There’s a wonderful irony in all of this. For decades, sales organisations have attempted to make selling more predictable, measurable and repeatable. We’ve built methodologies, frameworks, scripts, processes and increasingly sophisticated systems to eliminate the inconsistency of human behaviour. We’ve spent an enormous amount of time trying to make salespeople behave more like machines. Now that we’ve finally built machines capable of behaving like salespeople, we may discover that the most valuable thing a salesperson can be is human. And perhaps that’s the real lesson Kubrick and Clarke left us. The challenge was never simply to build more intelligent machines. It was to understand what intelligence is actually for. Because if we get that wrong, we could find ourselves surrounded by extraordinarily capable technology, generating extraordinary quantities of activity, without making the experience of buying or selling meaningfully better. We might even find ourselves staring at a screen, wondering why our perfectly researched, impeccably written, AI-optimised proposal has been rejected. And somewhere, in a calm and reassuring voice, a machine will explain: “I’m sorry, Dave. I’m afraid I can’t do that.” Aaron Evans 16 November 2026 Share : URL has been copied successfully!