Am I just a teck stack to you? How senior employees carry software between firms

As Mr. Mugatu from Zoolander may say, AI vendors are so hot right now. They are all over Ramp’s vendor index. From an economist's standpoint, technology spreading through the economy is the key driver of higher output. Understanding how tech spreads gives us a window into why some economies perform better than others.

There's two ways that tech spreads:

• Word of mouth (Ambient): people hear about tech from friends or the internet and bring the tech into a firm.
• Job hopping (Carriers): people have used tech at a prior company, get hired, and bring the tech with them.

Let’s say all technology diffuses through job hopping. Then economies with more job hopping will experience more tech adoption and thus higher productivity. Thus, understanding the dynamics of diffusion are important for explaining economic differences.

In this post, we’re going to zoom in even further:

• How much technology spread is driven by job hoppers versus word of mouth?
• Which AI vendors are new joiners more likely to bring with them

Dataset

To watch tech move between firms, we need two things: what tech each firm uses, and who moves between firms. We stitch together two datasets that give us exactly that.

• Ramp spend tells us what tech a firm uses, and when. Every time a company first pays an AI vendor, that's our "adoption" event. Across Ramp's vendor index we track 56 AI vendors, from the frontier labs (OpenAI, Anthropic) to niche infra (Together AI, Modal, Hugging Face) to consumer tools (Midjourney, ElevenLabs).
• Revelio Labs tells us who works where, and when they move. It's a labor-market dataset built from public professional profiles.

For any worker who changes jobs, we can see the tech stack of the firm they left and the tech stack of the firm they joined. That lets us tag every move as a carrier move (their old firm was already paying for vendor X, so they could bring it) or not, and then watch whether the new firm starts paying for X within a year.

We have 50,657 matched firms, of which 17,937 adopt at least one AI vendor over our 2022–2026 window. We zoom in on technical hires, Revelio software roles (Software Engineer, Technical Architect, Infrastructure Engineer, Data Analyst, QA Tester, plus Scientist) at seniority ≥ 3 (experienced individual contributors and up). These are the people most likely to carry a stack with them, and their clean job-to-job moves (left the old job before starting the new one, and deduped so an entire team switching firms counts as one event, not fifty). For the Love Index later we widen the lens beyond AI to Ramp's full set of ~330 tracked software vendors, so we can compare how AI tools travel against the broader tech stack.

We only observe a mover’s prior stack when their old firm is also a Ramp customer. So carrier is measured on the subset of moves where both firms show up in our data. We observe some job hoppers that enter a new firm, but do not know where they come from. We do not count those as carriers. Thus, we are under-counting the true carrier effect.

How Tech Spreads?

First, the topline: word-of-mouth does most of the work. Across our data, ambient exposure accounts for roughly 85% of adoption and carrier movement about 15%, a ~5:1 split. Job-hopping is a real but minority channel; what makes it worth studying is where it matters. As we'll see, it punches far above its weight for young, niche tools.

This comes from a regression of adoption on carrier status and the vendor's ambient market share, with vendor×month fixed effects absorbing the ambient trend: carrying a vendor lifts adoption +4.9%, and each +10% of market share lifts it +28.6%. Weighting each by how much it actually varies in the data splits the explained adoption ~14% movement / ~86% ambient.

The above results come from Ramp Rate data. Let's walk through the analysis step-by-step. We measure the market share of a given vendor. We call this the "ambient" exposure of a given technology because it's more likely to be known to a general person. A vendor with higher market share has higher ambient exposure. Below we walk through a specific vendor to explain this phenomenon. Granola's market share has risen a lot.

We can compare the likelihood a firm adopts Granola when we identify a carrier and when we don't. The gap between the two is the lift a carrier brings. We plot the different probabilities by the market share of Granola. What you can see is that the gap noticeably shrinks once Granola has >5% market share.

The carrier odds ratio just condenses those two curves into one number — the odds of adoption given a carrier, divided by the odds without one. As the two lines converge, the ratio falls toward 1:

What it plainly means: when a company hires someone who used a tool at their last job, how much more likely is it to start paying for that tool? The Carrier Odds Ratio captures that boost. A value of 5 means a firm's odds of picking up the tool are 5× higher when it hires a “carrier” of it than when it doesn't; a value of 1 means carriers make no difference.

Technically, for each vendor we take at-risk destinations (firms that hadn't already used the tool), split their hires into carriers (whose prior employer paid for the tool) and non-carriers, and divide the two groups' adoption odds — OR = [pᶜ/(1−pᶜ)] ÷ [pᵃ/(1−pᵃ)], where pᶜ and pᵃ are their 12-month adoption rates.

The index runs high for two reasons: users genuinely bring the tool with them (the signal we want), and the tool's market share is low (a small ambient rate mechanically lifts the ratio). That second force is also the main caveat, a niche tool concentrated in one industry can look “carried” just because workers move within that industry (sector homophily, not love). So we treat a high index as a candidate diamond in the rough and sanity-check the leaders.

This metric tracks growth. For each vendor we compute its carrier odds ratio, then measure how much its Ramp Rate market share grows. Across ~200 tech vendors, a higher odds ratio goes with faster subsequent share growth and, crucially, the relationship survives controlling for the vendor's starting share, so it isn't merely "small tools grow faster." The diamonds in the rough do tend to spread, which reassures us the index captures genuine love rather than mere niche appeal.

(How to read it: each dot is a vendor — x is its carrier odds ratio, y is the log-change in its market share, and the line is the best fit. One honest caveat: the odds ratio is estimated over roughly the same multi-year window in which we observe the growth, so this is a strong association rather than a clean out-of-sample forecast; a strict "odds ratio in year t → growth in year t+1" test is limited by how few carrier moves we see in the earliest years.)

Our general finding: a new joiner is far more powerful at spreading a technology while it's still small, below roughly 1–5% market share. Once a tool crosses that threshold it reaches escape velocity, spreading on its own momentum, and hiring a champion barely moves the needle.

The tldr: a new joiner is most powerful at spreading young technology. We call the carrier odds ratio the Vendor Love Index, the higher it is, the more a tool spreads through the people who use it rather than through the market at large. Modal, for instance, ranks near the top.

Which tech is loved?

We use the carrier odds ratio to find diamonds in the rough, tech that is loved relative to its market share. Below is the list of the top 10 general tech vendors, and top 10 AI vendors that have the highest carrier odds ratio. Note, an AI vendor could appear in the tech vendors list