A Davos panel with the bosses of Philips, Visa, Saudi Aramco and Accenture converged on the same diagnosis: the hard part of scaling AI is the organization. Those solving it are focused on who owns the work, who trusts the outputs, and who can spot the use cases worth shipping.
Growth is replacing “productivity” as the headline: Accenture’s latest Pulse of Change reporting says 78% of organizations now see AI as more beneficial to revenue growth than cost reduction. As a former consultant, I view this mostly as a shift in what leaders are trying to buy. And that matters.
Data advantage is the real moat (if you already have it): Aramco’s CEO said the firm expects $3–5bn in “technology realized value” in 2025, up sharply versus prior years. The underlying claim, echoed in coverage and company statements, is that the edge is less about compute than deep operational data built over decades.
The talent bottleneck is domain expertise, not ML engineers: Aramco and its executives have repeatedly stressed that scaling depends on training subject-matter experts (engineers, operators, geologists) to generate the use-case pipeline; the company has pointed to training 6,000+ people in AI-related capability.
Visa is positioning “agentic commerce” as a new transaction regime: Visa’s push is infrastructure. They’re building protocols to let merchants distinguish legitimate shopping agents from bots, plus standards for agent–merchant communication during checkout. In parallel, Visa is marketing “Intelligent Commerce” as a toolkit of tokenization, authentication and transaction controls intended to let agents transact safely on a user’s behalf (noting the product is still being deployed).
Clinician time is the sell in healthcare, not “efficiency”: Philips has framed the value as shifting time from admin to care, arguing that nurses can spend 10–15 minutes each hour on administrative work, and that automation should return that time to patients. I’ll note here that Buurtzorg outdoes these numbers by a long ways without AI, but that’s for another day.
Leadership literacy is an adoption constraint: Broad access to tools seems to produce little change until senior leaders build with them directly. After all, governance, risk appetite and prioritization sit at the top in most firms
What to watch
- Whether “AI value” shows up as revenue-linked measures (conversion, retention, pricing power) rather than time-saved.
- Whether firms build a repeatable kill / pilot / scale operating cadence (rather than function-by-function demos).
- In payments, whether “trusted agent” standards become default plumbing for online checkout.
- In industrials, whether incumbents’ historical data + first-mover advantage actually converts into defensible performance gaps.
Sources
- World Economic Forum: Scaling AI: Now Comes the Hard Part
- Accenture: Pulse of Change
- Saudi Aramco: corporate update on Davos / CEO panel participation
- Bloomberg: Aramco expects $3–5B “technology realized value” in 2025
- World Economic Forum: Roy Jakobs (Philips) on freeing clinician time for patients
- Visa: Trusted Agent Protocol announcement (Investor Relations / Business Wire)
- Visa: Trusted Agent Protocol technical documentation (developer portal)
- Visa: Intelligent Commerce product page