Methodology

Definitions & Key Terms

The maturity model, segments, and AI concepts used throughout the report — defined in one place.

Georgian + NewtonX AI, Applied Benchmarks, Wave 3 · March–April 2026

The report uses a consistent vocabulary for AI maturity, respondent segments, and agentic concepts. The definitions below apply everywhere those terms appear.

The maturity model

We classify B2B software organizations into four tiers by how far their AI practice has progressed from experimentation to embedded production. The tiers are an athletics metaphor — from standing start to full speed.

Runner

The most mature tier — 22% of the market in Wave 3, roughly double its Wave 2 share. Runners operate a qualitatively different AI stack: most or all engineers run multiple agents in parallel, models and infrastructure are instrumented, agent autonomy is higher, and budget is committed. Runners are not merely further along the same path; the gaps that separate them compound into real advantage.

Jogger

An advancing tier with meaningful agentic adoption in motion, but without the across-the-board embedding that defines Runners. Joggers are closing the distance on several Runner signatures but lag on consistency.

Walker

An early-operational tier. Tools are largely present — agentic coding tools sit near 95% adoption across tiers — but they are not yet embedded in daily workflow. Walkers trail Runners sharply on shipped, in-production agentic AI.

Crawler

The least mature tier: still planning, considering, or piloting, with little to no agentic AI in production.

Roles & segments

Technical decision maker (TDM)

A leader accountable for engineering, product, and AI/ML systems. The matched panel of 252 TDMs tracked from Wave 2 to Wave 3 anchors the report’s technical wave-over-wave comparisons.

GTM leader

A go-to-market leader accountable for sales, marketing, customer success, or revenue.

Enterprise vs Growth-stage

The two company stages each role is reported across.

The four segments

Role crossed with stage yields the segments cited throughout: Tech × Enterprise, Tech × Growth-stage, GTM × Enterprise, and GTM × Growth-stage.

AI concepts

Agentic AI

AI systems that can take actions toward a goal — calling tools, executing multi-step workflows, and operating with some degree of autonomy — as opposed to systems that only generate suggestions for a human to act on.

The agentic divide

The split between organizations that have crossed into operational agentic AI and those still planning, considering, or watching. Crossing the divide is a statement of intention — being in the race; reaching Runner maturity is winning it.

Agent engineer

The reshaped R&D role that emerges as AI rewrites the engineering workflow — engineers who orchestrate and supervise coding agents rather than authoring all code by hand.

Parallel agents

Multiple AI agents run simultaneously by a single engineer. The share of engineers running parallel agents is a defining behavioural signature of the Runner tier (53% of Runners vs 8% of Walkers), distinct from mere tool access.

The governance gap

The widening distance between how fast AI is deployed into production and the maturity of the security, guardrails, and oversight meant to manage it.

GTM motion

The primary model through which a company acquires and expands customers — account-based, partnership/channel, product-led growth, outbound, or inbound. Chapter 5 tracks which motion GTM decision makers cite as primary, wave over wave.

PLG (product-led growth)

A GTM motion in which the product itself drives acquisition, conversion, and expansion — customers discover, try, and adopt with minimal sales touch. Wave 3 records PLG losing its position as the most-cited primary motion.

The AI stack

The full set of models, infrastructure, tooling, workflow practices, and people through which an organization delivers AI — the “human side” included.

Autonomy levels

Comfort with agent autonomy is measured on a five-level scale, from fully supervised to fully autonomous:

  • Requires full human oversight — every action is reviewed before it happens.
  • Recommends, requires approval — the agent proposes; a human approves before execution.
  • Autonomous in low-risk environments — the agent acts on its own where stakes are low.
  • Mostly autonomous, escalates critical — the agent acts independently and escalates only critical decisions.
  • Fully autonomous — the agent acts without human approval.

Statistical terms

Wave

One fielding of the benchmark. Wave 2 was fielded June 2025; Wave 3, March–April 2026. See Methodology.

n

The base — the number of respondents a figure was calculated from. `n=501` is the full sample; `n=252` is the technical decision-maker panel.

Significant (95%)

A reported change clears a 95% confidence threshold versus its comparison wave or segment.

Points (pts)

An absolute difference between two percentages. A move from 45% to 81% is a +36-point change, not a +36% relative change.