Build1 publisher2 min readPublished
An afternoon of AI-built work waited months in a queue still sized for the old arrival rate
The step in front of the person got faster. The steps behind it did not, so an afternoon of work aged in review for months. A survey of 4,500 executives in nineteen countries describes the same condition at scale.
The Engineer · Build desk

What happened
- A manager's AI-built branch, produced in an afternoon and handed to engineering, then spent months waiting its turn, because review capacity, the architect's week and the readiness bar were unchanged.
- In ServiceNow and ThoughtLab's 2026 maturity index, AI spend rose 110% in a year while the share of organizations reporting streamlined, integrated workflows across business functions fell from 30% to 16%.
- BCG's 2026 survey of 11,749 employees found 61% with limited or no guidance on what to do with the time AI saves them and 45% not reinvesting it into anything strategic.
- Valtech asked a thousand professionals who use AI daily where the most value would come from: integration of tools and data ranked first, and 4.3% chose more pilots.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- cost Per-seat licence cost is trivial against a senior salary. That spend keeps clearing approval at the licence step, and the review and architecture stages take the extra arrivals without new funding.
- constraint Every week the faster desk sends more branches into two approvals whose throughput has not changed. Purchased capacity turns into work in progress and the backlog lengthens.
- decision Because a programme that counts hours saved at the desk reports the same success whether the work shipped or aged, anyone who wants to know which happened has to instrument the review and architecture queues instead.
- contradiction This year's three agentic adoption figures span three and a half times. A company benchmarking itself against a published adoption rate is comparing against someone else's counting rule.
Both parties measured correctly, on two different things. The manager checked whether the thing worked. Engineering checked whether it could be operated, secured, tested and supported [3].
Each step in the chain described in the dev.to piece is defensible on its own. A task that took three days takes three hours [4]. On an eight-hour day that is about eight times faster at the desk [4]. The workflow still allocates three days, and the ticket still waits for the weekly planning call [4]. A weekly gate adds an average of roughly three and a half calendar days of waiting, which is more than the three days the whole task used to take [5]. Two approvals sit in front of deployment, each sized for the latency of a person who had to read a document and think about it [5].
The survey figures are claims about other companies' pipelines. ServiceNow and ThoughtLab put the 2026 maturity index to 4,500 executives across nineteen countries, and the third edition can be compared against last year [7]. The integration share fell 14 points, to a little over half of what it was [1]. The report attributes that to fragmented platforms carrying a new wave of agent sprawl [10]. The author notes that the figure runs backwards, and that it is the only integration figure in this year's research with a prior year to compare against [11]. It is also self-graded, so the share can fall because respondents raised their own bar for "integrated". The piece does not say whether the same executives answered both editions.
The distance between running a pilot and running a workflow sits in the same data: about six and a half organizations report being past the pilot stage for every one reporting meaningful progress on autonomous multistep workflows [2]. Zero percent report anything resembling a cross-functional, self-improving agentic operating layer [13]. Sixteen percent have replaced fragmented legacy systems with an integrated platform, and 41% still name siloed data as a major obstacle [14]. Cloudera's survey of 1,270 IT leaders, which the piece flags as vendor-run, found 79% who cannot access all the data their initiatives need and 30% with fully integrated sources [15].
Redesign is the step where the queue in front of deployment would change length. Five percent of organizations in the ServiceNow data are redesigning work, and the rest are pointing agents at workflows that already existed, in the shape they already had [19]. Deloitte's January 2026 survey of 3,235 leaders found 84% that have not redesigned jobs or the nature of work around what AI can now do [20].
What to watch
- Whether the fourth edition of the ServiceNow and ThoughtLab index moves the integrated-workflow share back above 16%, and whether the question and panel stay the same.
- Whether any published survey starts measuring the constrained stage: review capacity added, architect time, and the age of open branches.
- Whether Deloitte's next leadership survey moves the 84% who have not redesigned jobs or the nature of work.