You say spam, I say data set
Across all four professional email addresses I used in 2025, I got a lot of random cold email: 5,122 messages in all from 229 distinct vendors trying to sell me stuff. In fact, I get so much cold email that I wrote about it: they're not all about AI, but after ChatGPT launched, the percentage of total pitches trying to sell me AI went through the roof.
You probably get these messages too. If you're like most people, you hit the unsubscribe button, block the senders, or just press delete. But personally, I save them all so I can analyze them later. Scrappy data sets FTW!
A whole lot of my cold mail in 2025 was pitching me on AI. But the promises being marketed changed hugely across the year.
Jargon is as jargon does
Jargon changes quickly as a field evolves, and this is evident in how people talk about AI. For instance, I recently wrote about how job titles in "AI transformation" were rapidly giving way to more specific roles in "AI enablement" and "AI optimization." Job titles offer a lens in to the changing landscape. So does the stuff that vendors are selling.
I identified the most common AI jargon embedded in the 5,122 cold outreach mails I received across 2025. The list of highly saturated terms is relatively small; these are the terms occur often enough that you're probably aware of them as clichés. There's also a long tail of other terms that show up often enough to be interesting. For this analysis, I looked at 100 AI jargon terms like the below:
- 10x productivity
- Agentic AI
- AI-native
- AI-powered
- Domain-specific AI
- Drive business outcomes
- Drive efficiency
- Efficiency at scale
- Enterprise-grade [security/data protection/guardrails]
- Future-proof your organization
- Human in the loop
- Outcome-based
- Reduce friction
- Reduce headcount
- Single pane of glass
- Transform the way you work
- Unlock value from data
Ring a bell? That's because you're hearing this jargon everywhere. But you're hearing some terms a lot more often than others, and their relative popularity has changed over the last year.
You'll be popular, just not quite as popular as me
For the sake of this story, we're going to focus on the relative popularity of these five examples:
- "AI-native / AI-powered"
- "agentic AI"
- "enterprise-grade"
- "drive efficiency"
- "drive outcomes / outcome-based"
I picked these five because 1) they are the most common by a large margin, and 2) they all represent the kinds of generic (dare I say empty?) promises that people typically use to talk about AI.
Methodology: I took every instance of one of the terms above and added them all together into three big data sets: 1) mail I got from January-April 2025, 2) mail from May-August, and 3) mail from September-December.. Then I looked at the relative prevalence of each term compared to the others to understand each term's "market share" during each time period.
tl;dr: AI promises are on the move, and the shift reveals what people are expecting from AI in 2026.
2025 kickoff: Look! We have AI!
In mail from the first third of 2025, mails that promise "AI-powered" or "AI-native" solutions dominate. Taken together, these terms have 31% of total market share. Most messages don't explain what it means to be AI-native, but no matter. The emails promise it anyway.
Everything agentic is also hot: Among the five most prevalent terms, "agentic AI" has 22% of total market share.
In other words, in the first part of 2025, vendors were trying to make sales just by saying, "Look! We have AI!"
Summer 2025: Security is sexy. Also, maybe you'll get some results!
By the time summer ends, "AI-powered" and "AI-native" have cooled off significantly, and "agentic AI" is starting to fade too. On the other hand, we're starting to see more mail promising "enterprise-grade" solutions that "drive efficiency." And mail that promises to "drive outcomes" has risen sharply.
Across the board, market share among our five terms is reasonably balanced in the middle third of 2025. In other words, the summer market is recalibrating.
Holiday 2025: Um just give me outcomes already
The picture painted by mail from the final third of the year is quite different from what we saw when 2025 began. Only 9% of messages are making "AI-powered" or "AI-native" promises. By contrast, nearly a third of all messages claim their tools are "outcome-based."
In other words, by the end of 2025, it seems that the market no longer thinks it's enough to merely include AI in your tools; buyers are actually expecting the tools to do something valuable.
You can see the relative trend lines across all five terms below. Their relative popularity almost completely flips as the year does on. By the end of the year, sellers are betting that buyers want to see outcomes, efficiency, and enterprise-grade data protections.
The bottom line: When a field develops as rapidly as AI is, the language we use to describe it shifts in a matter of months, weeks, or even days. The patterns in cold mail shows what sellers believe buyers will respond to. In 2026, buyers want real results. Vendors who make (and keep) specific, quantitative outcome promises will win.
What AI promises are you seeing show up in the market?
Kieran
If you liked this story, why not subscribe to nerd processor and get the back issues? Also, why not learn to tell data stories of your own?
My latest data stories | Build like a founder | nerdprocessor.com