The numbers are staggering and accelerating. Worldwide AI spending reached $1.65 trillion in 2025. It is forecast to hit 2.52 trillion USD this year and is on track to surpass 3.3 trillion USD by 2027. Enterprise boardrooms from London to Singapore are racing to deploy AI across every function they can. Sales, with its friction, its inconsistency, and its dependence on expensive human talent, is near the top of every automation wish list.
Yet, this relentless pursuit of automation often overlooks a fundamental truth: successful deployment hinges not just on technological capability, but on the intricate (and very human) dynamics of customer trust, built through communication, critical thinking, and collaboration—things that simply cannot be automated.
And therein lies the problem: the customers.
AI automation risks: The gap between industry experts and the public
Pew Research Center found that 50% of Americans say AI makes them more concerned than excited—up from 37% in 2021. Only 10% say they are more excited than concerned. The chasm between experts and the public is remarkable: 47% of AI professionals describe themselves as more excited than concerned.
A YouGov survey put the numbers in even starker terms: only 5% of Americans say they trust AI “a lot,” while 41% express outright distrust. Fewer than one in five would trust an AI system to make a decision or take an action on their behalf, and 77% are concerned that AI could pose a threat to humanity. In financial services, only 20% of Americans say they trust AI at least “somewhat,” while 48% say they do not trust it much or at all.
Meanwhile, Salesforce’s research showed a sharp year-on-year decline in the share of customers who trust companies to use AI ethically. And even the Salesforce CEO has publicly acknowledged that AI innovation is “far exceeding” customer adoption—a remarkable admission from a company betting billions on the technology.
So here is the tension no one wants to say out loud: we are pouring unprecedented capital into AI at precisely the moment consumers trust it the least. Consumer sentiment toward AI is deteriorating, not improving, even as AI capabilities are improving.
However, the rejoinder is always that AI will improve…
“We are pouring unprecedented capital into AI at precisely the moment consumers trust it the least.”
A thought experiment: Establishing AI boundaries
Imagine a perfect AI salesbot.
Not the clunky chatbot you encounter on most webpages. A genuinely perfect one. It never lies. It never hallucinates. It has complete product knowledge, responds instantly at any hour, never chases commissions, and is indistinguishable from a human in conversation.
Across every metric that executives typically care about, from accuracy and availability to scalability and cost, this bot is superior to your best human rep (assuming token costs are lower, and will remain so, than the cost of humans). So why, when a procurement director is evaluating a seven-figure software contract, would they hesitate?
The answer reveals something important about what selling actually is.
In high-stakes transactions, accuracy is necessary but not sufficient. The buyer’s real problem isn’t whether the information is correct—it’s whether they can trust the source. And trust, in markets defined by information asymmetry, where sellers know things buyers don’t, is not built through correct answers alone. It’s built through credibility, accountability, and the knowledge that someone, a real person with a reputation and a career on the line, will be held responsible when reality diverges from the proposal.
A bot cannot be held personally responsible in the same way that a human is. AI can be updated, manipulated, scaled, or switched off. And buyers know it.
“The buyer’s real problem isn’t whether the information is correct—it’s whether they can trust the source. AI can be updated, manipulated, scaled, or switched off. And buyers know it.”
Not all deals are created equal
Despite AI automation risks, AI does belong in the sales stack. The question is where.
B2B purchasing research distinguishes between three types of buying situations:
1. Straight rebuy (routine reordering of familiar products)
2. Modified rebuy (changing specs or vendors)
3. New task (novel, complex, high-stakes purchases).
These categories tell you almost everything you need to know about where AI can help and where it can harm.
In a straight rebuy, AI excels. The buyer already knows what they want. Information asymmetry is low. The sales function here is essentially transaction processing, and AI can do that faster, more accurately, and more consistently than any human. This is where automation delivers on its promise. In most firms, the e-commerce platform already delivers here, so AI is adding marginal value.
In a modified rebuy or new task, the calculation flips. Here, the buyer is navigating uncertainty, evaluating unfamiliar risks, and often managing internal politics across multiple stakeholders. They’re not just buying a product—they’re buying risk reduction and the reassurance that when things go sideways (and in complex implementations, they often do), there is a human counterpart who will exercise judgment and absorb accountability. A bot can draft proposals and configure options, but it cannot sit across the table and say, “I’ll own this.”
The irony is that as the price tag rises and the stakes grow, buyers’ preference for human interaction and accountability rises with it. The more a deal matters, the more a buyer wants someone they can call, escalate to, and hold responsible. AI deployment enthusiasm and buyer trust preferences are pulling in opposite directions, and the gap widens precisely where it hurts most.
The disclosure trap
Then there’s the problem that no AI vendor marketing deck likes to mention: what do you do about disclosure?
If you tell buyers upfront that they’re talking to an AI bot, research shows that purchase outcomes can decline significantly. People perceive bots as less empathetic and less knowledgeable, regardless of whether that’s actually true. They share less sensitive context (budget constraints, internal politics, competitive anxieties) that is often decisive in closing complex deals.
If you don’t disclose, you’re sitting on a trust bomb. The moment a buyer discovers the “person” they’ve been building a relationship with is an AI, the sense of manipulation can undo months of rapport, and with it the renewal, the expansion, and the referral.
There’s no clean exit from this dilemma. In complex sales, the seller cannot win the disclosure problem.
“If you tell buyers upfront that they’re talking to an AI bot, purchase outcomes can decline significantly.”
When the market itself becomes the enemy
Here’s where it gets more systemic. Even a genuinely trustworthy AI salesbot doesn’t operate in isolation—it operates in a market increasingly flooded with fraudulent ones.
The FBI has explicitly warned that AI-generated text, voice, and video are being weaponized to make financial fraud more believable and to industrialize persuasion. Voice cloning, deepfakes, and spoofed vendor bots are not science fiction. They are documented and growing.
The result is a “lemons” market problem, first identified by economist George Akerlof: when buyers cannot distinguish good-faith actors from bad actors, they rationally become suspicious of all actors. Your perfect, ethical, governed AI salesbot is being pooled in buyers’ minds with scam infrastructure that looks identical on the surface. The verification costs that AI was supposed to eliminate come roaring back: buyers demand live video calls, third-party attestations, legal review, and confirmed identities precisely because the AI channel has become a vector for fraud.
The efficiency gain disappears. The trust deficit remains.
What this means for today’s business leaders (and future ones)
This is not an argument against AI in sales. It’s an argument for honesty about what AI can and cannot do, and what deploying it carelessly costs you.
Trust, brand loyalty, and long-run account value are not soft metrics. They are the revenue you haven’t booked yet. When KPMG’s global research shows a declining share of people who believe AI’s benefits outweigh its risks, that’s not a PR problem. It’s a commercial one. Customers who feel manipulated or surveilled don’t churn quietly. They talk.
The firms that will win the next decade are not the ones that automate the most. Instead, they’re the ones that deploy AI where it genuinely serves buyers, maintain human accountability where it genuinely matters, and build the kind of trust that no chatbot can manufacture at scale.
As a practitioner, the question to ask is not “Can we automate this?” It almost certainly can be automated. The better question is “What do we risk losing when we automate this?” As a student entering a business world saturated with AI hype, the most valuable thing you can develop is the judgment to know the difference. And that judgment is uniquely human—supported by what we call the 5 Cs at Hult: critical thinking, creativity, communication, collaboration, and curiosity. These are the skills that can’t be automated and have ever-increasing value for exactly that reason.
Over 1.5 trillion USD is already committed to the AI spending boom. What isn’t settled yet is whether the investment will translate into relationships customers actually want, or a flood of sophisticated interactions that buyers learn, rationally and correctly, to distrust.
That choice is still being made. Make it deliberately.
“The firms that will win the next decade are not the ones that automate the most. They’re the ones that deploy AI where it genuinely serves buyers, maintain human accountability where it matters, and build the kind of trust that no chatbot can manufacture.”
_______________________
With AI fluency and the ability to strategically automate increasingly in demand, Hult has recently launched the AI Lab as a core component of our master’s and MBA programs. We understand AI automation risks and the critical value of keeping a human in the loop. That’s why the Hult AI Lab focuses not only on tools and technical capabilities, but on the core human skills that make the difference between successful and failed deployment. Find out more about the Hult AI Lab on our website.
