Enterprise leaders spent an estimated $30 to $40 billion on generative AI deployments over the past two years, and MIT's Project NANDA found that 95 percent of those pilots produced no measurable impact on profit and loss. Only 5 percent of integrated systems generated significant value, according to the team's GenAI Divide report. The finding has become the year's most quoted statistic in boardrooms weighing where AI spending actually pays off, and Forbes has since described sales organizations as entering their own trough of disillusionment with the technology.
Most executive suites read that number as a verdict on the technology. The same research points somewhere more specific. Sales and marketing functions captured roughly 70 percent of AI budget allocation across the surveyed organizations, even though back-office functions like procurement, finance, and operations delivered the more reliable returns. Executives poured the largest share of investment into the one business function where the product being sold has always depended on trust between two people, then measured its failure to produce revenue as if trust were a workflow that could be automated out of the equation.
Where AI spending actually pays off
That allocation pattern is the more useful data point than the headline failure rate. A dollar spent automating document processing or compliance review compounds, because the work being replaced was mechanical from the outset. A dollar spent automating the long social process of building trust with a client behaves differently because it's anything but mechanical. Capital deployed on the assumption that you can automate trust will keep reporting the same disappointing figures regardless of which vendor sold the tool.
Sales runs on relationship sequencing more than on data processing. Generative tools are exceptionally good at the latter and mostly irrelevant to the former. A prospect moves from cold to committed through a series of trust-building moments. Strip that step out and replace it with automated outreach, and the deployment stops accelerating the sales cycle it was built to shorten. It breaks the cycle instead.
The structural fix for AI in sales
The best solution is a structural separation between the work AI can own and the work a human relationship has to own, built into how a sales organization operates day to day rather than left to individual reps to sort out informally. The strongest version looks like a protected, non-negotiable block of time in the operating rhythm of every seller and every leader, reserved exclusively for relationship-based outreach and the small number of income-producing activities that move a deal forward. Everything else - research, scheduling, data entry, first-draft follow-up copy - gets delegated to AI so the protected hour stays protected.
A useful test for whether this is a real operating system rather than a personal habit is whether the practice survives if the leader who built it steps away for a quarter. A discipline that lives only in one executive's calendar is a productivity tip, useful to that person and irrelevant to the organization's revenue line the moment they are unavailable. A discipline written into how every seller's day is structured, measured the same way across the team, and reinforced regardless of who is running the desk that week is infrastructure. Only the second version scales past a single founder or a single star performer.
Organizations that made this investment pay off kept relationship-driven selling as the core revenue driver and used AI for scale and efficiency around it. That ordering got decided by leadership before any tool was purchased. Once AI efficiency starts creeping into the relationship-building step itself, the trust that actually closes deals goes with it. For sales professionals looking to build this discipline, AI for Sales training can help clarify which tasks belong to the machine and which belong to the human. Sales managers specifically can benefit from an AI Learning Path for Sales Managers that covers CRM optimization and revenue growth without displacing the relationship work.
Re-allocating the investment
Re-allocating AI investment toward back-office processes produces the bigger gain. Shifting expenditure out of the relationship-dependent front end of sales and into functions like procurement and finance, where automation already delivers a more dependable return, closes most of the gap the failure rate describes.
That failure rate keeps getting cited as proof that AI overpromised. It is more accurately read as a map of where automation tried to replace trust, the one variable no dataset can generate on its own.
Why this matters for sales professionals
Certain parts of a revenue function run on trust. Throughput can't substitute for it, and the discipline of protecting exactly those parts is what separates stalled deployments from ones that pay off. The real question for the C-suite is whether the relationship infrastructure underneath an AI deployment was built to survive the automation layered on top of it. That test matters more than the volume of automation any sales organization decides to deploy.
Your membership also unlocks:
