Georgian Partners, in collaboration with NewtonX and an 11-partner global consortium, has released its AI, Applied Benchmark Report, offering a robust snapshot of how AI is transforming B2B software and enterprise companies worldwide. This expanded second wave draws on a blind survey of 612 executives—split evenly between R&D and go-to-market leaders—across 10 countries and 15 industries, representing companies with annual revenues ranging from $5 million to over $200 million.
What sets this report apart is its global scope and strategic backing. Consortium partners include the Alberta Machine Intelligence Institute, AI Marketers Guild, FirstMark, GTM Partners, Untapped Ventures, the Vector Institute, and Tel Aviv–based Startup Nation Central and Grove Ventures, among others. Their involvement helped broaden participation and ensure sector-diverse, international benchmarks.
More than just a measure of adoption, the report captures the structural barriers, emerging AI use cases like Vibe Coding, and the evolving maturity curve of AI integration. With findings grounded in validated, executive-level input, the report offers companies a practical framework to benchmark where they stand—and what’s holding them back.
AI Becomes a Strategic Imperative
Artificial intelligence is no longer considered optional. The report finds that 83% of B2B and enterprise companies now rank AI among their top five strategic priorities. In fact, three of the top five most selected business priorities are AI-related, showing how embedded it has become across corporate agendas.
The leading motivations for AI adoption continue to be:
-
Improving internal productivity
-
Creating a competitive advantage
-
Enhancing cost efficiency and revenue growth
What’s changed, however, is that competitive differentiation has now overtaken cost savings and revenue as the second most important motivator. This marks a shift in mindset: AI is not just a tool for automation—it’s a weapon for market leadership.
Vibe Coding Enters the Mainstream
A standout insight from the report is the rapid rise of Vibe Coding—a term referring to automated code generation and debugging using AI models. Vibe Coding has become the #3 R&D use case reported in production, used by 37% of companies, while another 40% are actively piloting it.
This trend is not simply about improving developer productivity. It’s also a direct response to an industry-wide challenge: the shortage of AI technical talent, which has now become the #1 barrier to scaling AI. Forty-five percent of R&D leaders cited this talent gap as their top concern—surpassing even the high cost of model development.
Vibe Coding is helping fill that gap by allowing leaner engineering teams to accelerate delivery timelines, debug faster, and produce cleaner, documented code with less overhead. Respondents noted measurable reductions in manual effort across QA, infrastructure, and deployment workflows.
AI Productivity Gains—and Their Limits
The use of AI across development pipelines is showing clear benefits. According to the report, 70% of R&D respondents report faster development velocity, 63% see improved code quality and documentation, and over half have increased deployment frequency.
However, not all metrics have improved. Areas like mean time to restore, cycle time, and change failure rate remain weak spots. This suggests that while AI is accelerating the front end of development, stability and resilience remain human-dependent for now.
Infrastructure Upgrades Power the AI Stack
Supporting these gains is a dramatic shift in infrastructure investment. AI-driven teams are adopting new tooling to move from experimentation to production:
-
LLM observability platforms have been integrated by 53% of companies
-
Data orchestration tools such as Dagster and Airflow are now used by 51%
-
Vector databases, cron jobs, and durable workflow engines are being deployed to support scale and reliability
Meanwhile, companies are sourcing more data than ever to fuel their models. The use of owned data rose 12 percentage points to 94%, while public data use rose to 80%. Synthetic and dark data—once fringe sources—are now being used by over half and a quarter of companies, respectively.
LLM Adoption Diversifies
OpenAI remains the leading provider of large language models, with 85% of respondents using its models in production. However, the landscape is evolving rapidly:
-
Google Gemini saw a 17-point surge, now used by 41%
-
Anthropic Claude rose to 31%
-
Meta’s Llama 3 family is gaining traction with 28% adoption
-
Reasoning-specific models like OpenAI’s o1-mini (35%) and DeepSeek (18%) are also entering production
This shift reflects a move toward multi-model AI stacks, where organizations match models to use cases rather than relying on a single vendor ecosystem.
AI Maturity Gains Are Uneven
Georgian segments companies using its Crawl, Walk, Run AI maturity model. While more organizations are progressing from beginner to intermediate levels, the top tier of maturity remains elusive:
-
“Walkers” dropped to 40%, down from 49%
-
“Joggers” rose to 31%, indicating growing momentum
-
“Runners” remain stagnant at 11%, suggesting a ceiling in scalability
The companies that do reach the “Runner” stage tend to be those who connect AI projects directly to revenue or cost outcomes—a capability still underdeveloped across much of the industry.
ROI Remains Elusive
One of the most persistent challenges identified in the report is the lack of clear ROI measurement. More than half of R&D teams admit they are not connecting AI projects to any concrete KPIs. Only 25% directly link AI initiatives to new revenue, and just 24% report a positive impact on customer acquisition costs.
Still, optimism persists. Over 50% of respondents say AI has improved customer satisfaction and long-term value. But the overall sense is that the financial justification of AI remains fuzzy, particularly at the mid-maturity level.
Cost Management Is Improving
While talent remains the biggest obstacle, costs are slowly becoming more manageable. The report shows:
-
A 9-point shift toward stable or reduced data storage costs
-
Declining costs in software maintenance, labor, and operations
-
Less reliance on cost-cutting measures like project restrictions
Additionally, 68% of companies now rely on third-party AI solutions to manage cost and complexity, especially as AI becomes embedded in GTM software and internal platforms.
A Look Ahead
The implications of this benchmarking data extend far beyond dashboards and boardrooms. As AI becomes central to how software is built, deployed, and maintained, the industry is entering a new phase—one where productivity is no longer just about people, but about how intelligently teams can augment themselves with machine partners.
Vibe Coding represents a turning point. It’s not just a productivity tool; it’s becoming a foundational layer of modern software development. For companies facing persistent talent shortages, it offers a way to unlock throughput, reduce time-to-market, and improve code quality without scaling headcount at the same rate. And for those further along the maturity curve, it creates the backbone for AI-native engineering workflows—ones that can scale with observability, reliability, and measurable business impact.
The broader message is clear: the companies that succeed won’t just use AI—they’ll operationalize it, embed it, and evolve with it. In this new era, automation isn’t about replacing developers. It’s about amplifying them.
Those who treat Vibe Coding and its supporting infrastructure as strategic investments—not experiments—will define the next wave of enterprise innovation.
