State of Sales 2026: The Numbers Behind the AI Agent Wave, Read Honestly

Salesforce published its State of Sales 2026 report in September, based on 4,050 sales leaders, reps and operations professionals across 22 countries, Germany among them. Vendor research deserves skeptical reading, Salesforce sells the thing the report celebrates, but the dataset is large, the trendline matches what we see in DACH first calls every week, and several numbers cut against the vendor's own hype. Those are the interesting ones. Here is the honest read.

The adoption numbers: the race is over

54 percent of sellers say they already use AI agents. Nearly nine in ten plan to by 2027. 87 percent of sales organizations use AI in some form, for prospecting research, forecasting, lead scoring or drafting. And 94 percent of leaders who deployed agents call them critical to meeting business demands.

Whatever error bars you put on self-reported vendor-survey numbers, the direction is unambiguous: using AI in sales stopped being a differentiator sometime in the last eighteen months. That matches the market shift we tracked from the AI-tools era through named agents like Salesforce's Hunter. The competitive question moved one level up: everyone has the engine now, so the edge is in what you point it at, which is targeting, data and process.

The number nobody puts in the keynote

Sellers still spend roughly 40 percent of their time actually selling. Gen Z reps: 35 percent, losing about two hours a week to manual data entry alone.

Sit with that for a moment. Three years of AI in every sales tool, 87 percent adoption, and the core dysfunction of the profession, most of a seller's week going to non-selling work, has barely moved. Our read: most deployments so far have been assistive, a drafting box here, a summary there, which speeds up tasks without removing them. The report's own expectation numbers point at the next phase: sellers project around a third less time on prospect research and content creation once agents take over those jobs wholesale rather than assisting with them. That is the difference between AI as a feature and AI as a coworker, and it is why the agent wave is not a rerun of the copilot wave.

The top-performer gap is about the boring stuff

Two findings deserve more attention than the adoption headline. High performers are 1.7 times more likely than underperformers to use AI agents for prospecting. And 79 percent of high performers prioritize data hygiene, versus 54 percent of underperformers.

Read together, these say something we have been repeating for years, now with vendor data behind it: the AI payoff is downstream of the data discipline. An agent standing on a clean CRM with a sharp ICP compounds; the same agent on stale data automates noise. The 25-point hygiene gap between winners and losers is the least glamorous and most actionable number in the report, and it costs nothing to act on. It is also, not coincidentally, the first thing we check in the funnel diagnosis of our DACH benchmarks.

The Gen Z warning

Two more numbers, easy to skip, expensive to ignore: 46 percent of Gen Z reps say they rarely get feedback on their sales conversations, and 47 percent say they lack roleplay practice before facing customers. The generation that will run sales in ten years is being onboarded worse than any before it, precisely while the entry-level tasks that used to be the training ground, research, list work, first drafts, move to agents. If agents do the apprentice work, companies need a deliberate answer for how juniors learn the craft. The teams treating this as a coaching problem now will own the talent market later.

What this means in DACH

The survey includes Germany, but the numbers are global averages, and two local corrections apply. First, agent-assisted research and admin transfer to DACH one to one; agent-written outreach does not, because German register and tone still need a native human gate, the constraint we detailed in AI agents in sales. Second, the legal layer of first contact is stricter here per channel and country, so "the agent handles outreach" translates in DACH to "the agent prepares outreach". Teams that adopt with those two corrections get the report's upside without its failure mode.

The practical sequence stays what it was, and the report's own data now argues for it: fix data hygiene first, point agents at research and admin second, keep humans on message, language and conversations, and measure meetings from real buyers rather than activity. Adoption is table stakes. Discipline is the edge.