What Makes a Runner

The Runner tier — the most AI-mature tier in Georgian’s Crawl, Walk, Run framework — is now 22% of the market: nearly 1 in 4 B2B software companies surveyed, and roughly double what it was nine months ago in Wave 2.

The Runner tier isn’t just further along the same path as Walkers and Crawlers; it looks like it is operating a qualitatively different AI stack. Six specific, measurable gaps separate Runners from the rest of their surveyed peers — across parallel agent deployment, AI model adoption, AI infrastructure instrumentation, AI operational complexity, AI agent autonomy, and budget commitment. None of them work in isolation. Together, we believe these gaps define a compounding advantage for Runners.

Understanding the Runner segment helps define the difference between intention to cross the agentic divide, as discussed in Chapter 2, and actual production outcomes. In our opinion, it gives the 75% of the B2B world lagging on agentic adoption a blueprint on how to catch up.

This is a key behavioral signature of the Runner tier. It is not a measure of tool access — agentic coding tools sit at ~95% adoption across both tiers. It is a measure of how deeply those tools are embedded in the daily workflow. A Runner engineering floor looks fundamentally different from a Walker floor on any given day.

Runners are more likely than Walkers to experiment with or use a variety of model types beyond basic LLM use — large reasoning models (LRMs), multimodal, open-source, and voice. LRMs in particular are important to Runners and distinguish them from other maturity tiers: they unlock multi-step reasoning, complex planning, and long-horizon agentic execution that standard LLMs alone cannot reach. Runners appear to treat LRM adoption as foundational infrastructure, not a capability experiment for the future.

Runners treat AI as a production system requiring production instrumentation. In contrast, Walkers appear more likely to deploy AI and manage it reactively. LLM observability — prompt tracing, output monitoring, evaluation pipelines, drift detection — helps organizations know when their AI breaks. 55% of all technical decision makers experienced an AI-related incident last year; thanks to their infrastructure implementation, Runners likely knew about these incidents faster.

LLM observability is only one example of where Runner production deployment is well ahead of Walkers. Vector databases (+16 pts), data orchestration pipelines (+7 pts), and queues/background jobs (+19 pts) are also significantly more widely adopted by Runners.

Runners are 12× more likely to run fully autonomous AI agents in production, well ahead of any other segment. This deployment potentially entails agent-to-agent coordination, failure detection without human review, and autonomous rollback of agent decisions. These are not routine tooling problems — they are organizational and infrastructure problems that take time and structured experimentation to develop. The 12× gap is a measure of accumulated operational knowledge, not just accumulated technology.

Still, it begs the question: is there such a thing as too fast? Chapter 7 explores the governance gap — in it we see evidence that AI production deployment gains may be outstripping IT’s ability to keep up.

Runners are ahead on many individual aspects of operational AI, so it is not surprising they are using AI for qualitatively harder problems — multi-step processes spanning multiple tools and multiple teams. 42% of Runner technical decision makers describe their AI workflows as mostly or very complex, versus 11% of Walkers.

The ability of Runners to drive complex B2B workflows with AI is a potential turning point for AI investment. Individual components of the AI stack take significant time and money to implement, so the question is: when will B2B companies see the compounding effect that drives hard ROI? We go deeper on this question in Chapter 6: AI Meet P&L.

The 30% median IT-spend level indicates AI is a key organizational priority among Runners. This level of spend appears to financially sustain the AI stack as used by Runners — a hefty figure to observe this early in the AI adoption lifecycle, and one that shows how committed Runners are to gaining advantage quickly. Interestingly, company size is not the predictor: enterprise and growth-stage Runners spend at the same share. The data indicates that AI maturity is a better predictor of IT spend on AI.

This last finding sets up Chapter 6: AI Meet P&L, where we discuss the growing cost of AI and how CFOs will need to navigate the ROI and forecasting uncertainty created by it.