TMT Online ObServer "CINT/GLOB/TOTVS: AI speeds coding, but decision lat..."

研报英文原文证据摘录

TMT Online ObServer "CINT/GLOB/TOTVS: AI speeds coding, but decision lat..."

Global Research

21 May 2026ab

TMT Online ObServer Equities

Emerging MarketsCINT/GLOB/TOTVS: AI speeds coding, but

decision latency still slows delivery; AMX, TEF- Telecommunications

BZ and TIM all register for Oi Soluções auction Leonardo Olmos,AnalystCFA

leonardo.olmos@ubs.com

+55-11-9658 7533

TMT calendar next main events: Gustavo Farias

Associate Analyst

(TODAY) Endava earnings release gustavo.farias@ubs.com

(TODAY) Senior Sistemas Investor Day +55-11-2767 6154

(May-27) AMX Investor Day

CINT/GLOB/TOTVS: AI accelerates coding, but decision latency still slows

delivery

What happened: Yesterday, CI&T released a research paper challenging the

assumption that faster code generation automatically leads to faster software

delivery. Instead, the company argues that the real constraint is now decision

latency — the time work spends waiting for human review and approval before

moving forward. Key takeaways were:

AI productivity gains have not translated into faster delivery. Despite heavy

enterprise investment in generative AI, copilots, and coding automation,

delivery metrics (lead time, deployment frequency, and predictability) have

largely remained flat. The reason is that AI dramatically increases execution

capacity, but most enterprise delivery models still rely on human validation,

governance, and approval flows designed for a pre-AI era.

The bottleneck has shifted from code creation to decision orchestration.

With AI pushing the marginal cost of code production close to zero, invisible

queues compound across the Software Development Life Cycle (SDLC), with

decision latency emerging as the deeper structural constraint inside

enterprise operating models.