Enterprises are not short on AI budget. Gartner puts global AI spend at $2.59 trillion in 2026, up 47% year on year, one of the fastest spending accelerations in the history of enterprise technology. This is not hesitant experimentation. It is real capital, moving fast, with boards expecting results. I have seen this first hand over the last decade, including as the executive accountable for two Salesforce transformations. One went well and helped transform a company. The other gave too much control to Finance and IT, ignored the customer, added process for the sales team, made us less agile and left sellers disengaged. It became the antithesis of agile transformation.
So why do so few sales organisations feel transformed? And if technology is only part of the equation, what else has to be true for AI to create real commercial impact?
Spend is real, but it's pointed at the wrong target
MIT’s NANDA research, one of the most rigorous studies published on enterprise AI outcomes, found that more than half of 2025’s AI budgets went into sales and marketing pilots, precisely because they are visible to leadership. The actual returns showed up elsewhere: in unglamorous back-office automation nobody puts on a slide. Enterprises are not underinvesting in AI. They are investing in the parts of the business that make the best demo, not the parts that move the P&L.
Adoption is the bottleneck, not the model
Even Gartner’s own sales analysts concede this. Productivity gains from generative AI depend on one unglamorous precondition: frontline sellers actually using and trusting the tool. Without that, a company cannot tell whether AI is changing headcount, territory design or comp plans, because nobody is using it consistently enough to measure. MIT’s data makes the same point with a number attached: AI pilots built by blended internal and external teams succeed 67% of the time, versus just 22% for IT-only builds. The difference is not the technology. It is whether anyone did the unsexy work of getting people to change how they work. I have seen this first hand too many times: a tech vendor sells into IT, IT buys a system or tool without full business or customer knowledge, and the business is then expected to make it work. It is amazing how often this still happens. When an idea is not sponsored by the business leader who has to own the outcome, and is instead gifted to them by IT or a vendor, it is rarely as successful as one they have chosen and championed themselves.
Process is the multiplier — for better or worse
This is the part most vendors will not tell you: AI does not fix a sales process, it amplifies whatever process already exists. Feed it a CRM full of duplicate records, undefined stages and deals that sit untouched for months, and AI will simply document that dysfunction faster and with more confidence. Feed it a sales motion with clear stages, clean data and disciplined pipeline hygiene, and AI compounds that advantage just as fast.
McKinsey’s State of AI research backs this with hard numbers. The small cohort of “high performers”, the roughly 6% of companies seeing 5%+ EBIT impact from AI, are not distinguished by better models or bigger budgets. They are nearly three times more likely to have redesigned the underlying workflow before layering AI on top of it. Everyone else bolted AI onto a process nobody had rethought since before the tool existed, and got faster versions of the same broken outcomes.
Outcomes lag — but that's a measurement problem, not proof of failure
Only 39% of organisations report AI impact at the enterprise EBIT level, which sounds damning until you notice where the value actually shows up first: at the individual workflow level, well before it rolls up into a P&L line. The deeper issue, according to Harvard Business Review’s recent research on why innovations fail to scale, is rarely the idea itself. It is that sales, RevOps and IT do not trust each other enough to collaborate across the handoffs an AI rollout requires. That is not an AI problem wearing a technology costume. It is an organisational trust problem, and no model update solves it.
In an AI world, how ideas surface, earn sponsorship, attract a business case and move through ownership, promotion, adoption and execution is an art and a science in itself.
What to watch for next
Three signals will tell you whether your organisation is on the right side of this divide. First, watch whether budget shifts from seller-facing pilots towards the data and process foundations that produce real ROI. Most companies have not made that shift yet. Second, watch whether someone is explicitly accountable for the cross-functional trust-building this requires, what HBR calls a “bridger”, rather than treating AI adoption as an IT rollout with sales as a passive recipient. Third, and most simply: before evaluating any new AI tool, ask whether you can describe how the underlying process changes. If you cannot answer that question, the tool will not answer it for you. It will just do the broken version faster.
The data from Gartner, MIT, McKinsey and Harvard Business Review all converges on the same uncomfortable point: AI is not a substitute for a commercial system that works. It is a magnifying glass held over the one you already have. This is precisely why I co-founded Caidence. Decades of experience have left me with a hardened belief that companies obsessed with the commercial system, and with the alignment that underpins it across the business, are the ones that build long-term sustainable success. It also explains why Caidence, with AI embedded in the work, is part of driving rhythm and movement in a business. It is not the only enabler of success, but it helps turn commercial intent into consistent action.