AI is not yet delivering on its productivity promise as token spend surges >10x

FOR IMMEDIATE RELEASE

Plandek releases new research from 2,500+ engineering teams revealing a widening gap between AI investment and real productivity gains, putting business plans built on AI-driven output at risk

LONDON — 5/10/26 — Plandek, the leading Developer Productivity Insight (DPI) platform, today released its Q4 2026 AI Adoption and Impact Benchmarks, a study of quantitative engineering data from more than 2,500 software engineering teams worldwide.

The findings reveal a stark gap between the productivity gains business leaders expect from AI coding tools and what engineering organisations are actually delivering, even as spending on AI tokens has increased by 10 to 13 times since January 2025.

With Goldman Sachs forecasting a further 24x increase in token consumption by 2030, Plandek warns that token spend is accelerating faster than measurable engineering output — a gap that, left untracked, threatens to thwart AI-dependent business plans.

"AI is not yet delivering to the productivity expectations of stakeholders, at a time of rapidly increasing token spend," said Charlie Ponsonby, Co-founder and CEO of Plandek. "The need to really understand and improve tokenomics is acute."

Throughput is up, but quality is down

The benchmarks, drawn from the Plandek DPI platform and measured against Plandek's proprietary 4 Pillars of Engineering Productivity framework, show that the most mature agentic engineering teams have increased throughput (calculated on a normalised basis) by 50% over the past six months.

But that gain has come at a cost: these same teams are spending less time on value-creating work, with their Stories-to-Bugs ratio declining by 27% as quality has declined.

Cycle Time improved by 16%, yet Lead Time to Value, the true idea-to-production measure that matters to businesses, improved by only 4%, held back by bottlenecks across the software delivery lifecycle.

These findings reinforce the need to improve foundational engineering processes in order to deliver real productivity improvements.

A persistent performance gap between teams

Beyond the AI-specific findings, the research shows that fundamental performance gaps between high- and low-performing engineering teams persist and remain large, regardless of AI maturity. Relative to the lowest-performing cohort, top-quartile teams:

• Deliver 4x the throughput

• Deliver 3x fewer defects

• Are twice as focused on value-creating work

• Deliver work in half the time

• Are 55% more predictable

An orderly AI transition needs a plan

Plandek's qualitative research identifies common characteristics among the organisations successfully managing tokenomics and converting AI investment into real business outcomes:

• They follow a clear AI transition strategy instead of surrendering control

• They improve foundational engineering processes as they adopt AI tools

• They prioritise clear visibility of AI tools, focusing on the leading indicators of effective adoption (e.g. harness and context engineering, and model selection).

"The organizations pulling ahead aren't just adopting AI tools faster — they're pairing that adoption with disciplined visibility into what it's actually producing," commented Charlie Ponsonby. "Make every token count isn't a slogan, it's now a mandate for any team whose business plan assumes AI-driven productivity gains."

To help engineering and business leaders close the gap between AI investment and delivered value, Plandek has developed the AI-Transition Playbook, built on its RACER™ metrics framework (Rollout, Approach, Constraints, Engineering Impact, Results). The framework gives leaders a structured way to track AI tool adoption, identify constraints slowing impact, and connect engineering activity to measurable business ROI.

[[ENDS]]

About the research

The Q4 2026 AI Adoption and Impact Benchmarks are based on quantitative data from more than 2,500 engineering teams globally, tracked using the Plandek Developer Productivity Insight (DPI) platform, using Q1 2025 as the pre-AI baseline.

The full benchmark report, including detailed quartile data across Focus, Speed, Predictability and Quality, is available at [www.plandek.com].

About Plandek

Plandek is a leading global Developer Productivity Insight (DPI) platform based in London. Plandek partners with clients worldwide to help them drive software engineering productivity and accelerate the transition to the agentic PSDLC safely and productively.

Plandek is unique in offering market-leading software (the Plandek DPI platform), proven frameworks (including RACER™ and the 4 Pillars of Engineering Productivity), and deep domain expertise to help clients realize rapid business benefits as they adopt AI tooling.

Plandek is particularly active in the private equity space, working with PE value creation teams and portfolio companies across the U.S. and EMEA.

Plandek is used by more than 2,500 engineering teams worldwide across the private and public sectors, and is recognized as the leading Niche Player in the Gartner Magic Quadrant for Developer Productivity Insight Platforms 2026, and top global vendor in DevOps VSM Market Guides (2020, 2022, 2023), SEI Market Guide 2024, and the 2025 DPI Guide.

Media Contact

Ruth Dillon-Mansfield
rdillonmansfield@plandek.com

Ines Mitsou
ines.mitsou@11content.com
+34 689 210 438