The AI value gap, the distance between what companies expect from generative and agentic AI and what actually shows up in EBIT, has mostly been a numbers story so far: BCG, Gartner, and McKinsey statistics showing that individual productivity gains rarely translate into demonstrable enterprise results. Two recent practical cases show what that gap actually feels like when companies try to close it, and what it means for evaluating international AI vendors in DACH sales cycles.
AWS learns from its own rewiring
McKinsey recently described what Amazon Web Services learned from its own internal shift to agentic AI systems (“Lessons from our alliances: What AWS’s agentic journey can teach CEOs about rewiring for AI”). What stands out is not that AWS is adopting agents, that’s expected from a cloud provider of this size, but what the case reveals about the real challenge: even a company with the deepest technical expertise and unrestricted access to its own models had to fundamentally reorganize internal processes, ownership, and approval paths before agents delivered real value. Technology was never the limiting factor. Rewiring the organization was.
That’s the practical confirmation of a thesis that often stays abstract in consulting statistics: AI value comes from restructuring, not from adoption. Layer a tool over existing processes and you get, at best, the individual productivity that the McKinsey numbers already document. Change the processes themselves, as AWS did internally, and you get something else.
Rippling: the cheapest model turned out to be the most expensive
A second, more concrete piece of evidence comes from Rippling’s practice, discussed publicly via SaaStr. Rippling benchmarked 2,100 agent runs across different models for a specific task: processing payroll data with AI agents. The result contradicts the intuitive assumption that per-token model cost is the relevant metric: the nominally cheapest model turned out to be the most expensive once total cost was accounted for.
For DACH sales conversations, that’s a directly usable argument. International vendors who sell their pricing model primarily on low token or usage costs don’t necessarily address the question a buying center actually asks: what does the task really cost in the end, including error rates, rework, and oversight effort? The Rippling case shows that these two questions, unit price and true total cost, can diverge systematically. A vendor who answers only the first question leaves the second to the buyer, and risks exactly the kind of trust erosion that later surfaces as a compliance or pricing objection in the sales cycle.
The numbers behind it: 37 percent, 40 percent, 80 percent
McKinsey’s “State of AI 2026” report provides the quantitative frame around both cases. EBIT attribution to AI initiatives sits at 37 percent, flat versus the prior year, despite noticeably broader adoption: 40 percent of companies with more than $1 billion in revenue are now scaling agents, up from 27 percent the year before. At the same time, 80 percent of companies report individual productivity gains without consistent organization-wide ROI evidence.
That’s the AI value gap in a single set of numbers: more adoption, flat enterprise value, high individual satisfaction. It’s exactly the pattern AWS tried to break through organizational rewiring, and Rippling through genuine total-cost measurement, each in its own way.
The decision dividend, not the headcount dividend
McKinsey’s follow-up piece, “The decision dividend: How AI creates economic value,” provides the methodological explanation for why pure headcount-cost arguments fall short here. According to that analysis, the biggest value contributions from AI rarely come from labor savings alone. They come from faster decisions, better use of existing assets, and opportunities that would otherwise have been missed entirely.
For international SaaS and AI vendors building trust in the German market, that’s an important positioning point. A value argument built primarily on headcount reduction doesn’t address the metric a DACH buying center will actually use to judge the business case later. A value argument built on decision speed, asset utilization, and previously missed opportunities lands closer to what McKinsey’s own data identifies as the real value driver.
What this means for DACH market entry
Both cases point to the same advice from two different directions. AWS shows that the business case for AI is built through organizational rewiring, not through the tool itself, an argument that belongs in change-management language in sales conversations, not pure product language. Rippling shows that a pricing conversation focused only on unit cost leaves the total-cost question to the buyer, and in DACH sales cycles that question tends to surface earlier than elsewhere.
Vendors who build both lessons into their DACH GTM story address the question behind the 37 percent before the buyer has to raise it themselves.
If you’d like to pressure-test your own DACH GTM story against these two cases: in a 30-minute strategic diagnostic call, we look at this directly against your product and your target market. Book a strategic diagnostic call
