In a controlled field experiment, startups that redesigned end-to-end workflows around AI generated 90% more revenue than equally equipped peers that used AI mainly to speed up individual tasks.
Your research-backed argument about workflow redesign vs. task optimization is crucial. In LATAM, the challenge is amplified: many organizations still haven't mapped their fundamental processes, so when they introduce AI, they're automatíng waste at scale. The 90% revenue lift in your study hinges on organizational readiness to redesign. How are you seeing organizations that lack mature process documentation approach this? Should companies build process intelligence first before deploying agents, or can workflow redesign happen in tandem with AI adoption?
Agree the shift is structural, not technological. But framing it around 'process' and 'agents' may be the next bottleneck.
The goal is absorbing more variation at lower cost. That points toward organising around problems rather than processes—human-AI assemblages where automation handles patterned work and humans handle exceptions, resolving the problem via interactions with other entities.
This would mean that the biggest opportunities are not within the firm at all. They're at the boundary between firms. Think Australian Super: $500M a week flowing through a small coordinating entity, with nearly all work outsourced. The 'human glue' there bridges legal entities, not internal systems. Nobody owns the seam, so nobody redesigns it. That's where the value is.
The task-speed versus throughput distinction is valuable. In research-heavy work, though, some handoffs also preserve provenance: why a source survived review, which claim it supports, and where uncertainty remains. Eliminating that handoff can improve cycle time while making the final judgment harder to audit or revise. How would you balance throughput with the information value of a handoff when designing an AI-assisted research-to-writing workflow?
Absolutely! The 90% revenue increase wasn't from faster tasks, it was from redesigning entire workflows.
At https://www.designstudiouiux.com, we see this constantly: companies optimize the email while missing the chance to eliminate the whole Excel → QuickBooks → email chain.
The shift from "How can AI help this task?" to "How should we design this from scratch?" is everything
This is the part many teams still want to skip. They bolt AI onto yesterday’s workflow and call it transformation. Usually it is just legacy process running with better caffeine. The real shift starts when the sequence of work changes, not when the tool menu gets longer.
Mapping and after that getting people to adopt systems. Many pre-Ai automation attempts failed because the recording of data (the machine’s only access to the universe) relied on humans being open, disciplined and consistent.
By the way, I think I caught a typo…
‘[…]and that 40% of respondents (i.e., 4/5 of those that tried) […]’.
Love it!
Your research-backed argument about workflow redesign vs. task optimization is crucial. In LATAM, the challenge is amplified: many organizations still haven't mapped their fundamental processes, so when they introduce AI, they're automatíng waste at scale. The 90% revenue lift in your study hinges on organizational readiness to redesign. How are you seeing organizations that lack mature process documentation approach this? Should companies build process intelligence first before deploying agents, or can workflow redesign happen in tandem with AI adoption?
Agree the shift is structural, not technological. But framing it around 'process' and 'agents' may be the next bottleneck.
The goal is absorbing more variation at lower cost. That points toward organising around problems rather than processes—human-AI assemblages where automation handles patterned work and humans handle exceptions, resolving the problem via interactions with other entities.
This would mean that the biggest opportunities are not within the firm at all. They're at the boundary between firms. Think Australian Super: $500M a week flowing through a small coordinating entity, with nearly all work outsourced. The 'human glue' there bridges legal entities, not internal systems. Nobody owns the seam, so nobody redesigns it. That's where the value is.
The task-speed versus throughput distinction is valuable. In research-heavy work, though, some handoffs also preserve provenance: why a source survived review, which claim it supports, and where uncertainty remains. Eliminating that handoff can improve cycle time while making the final judgment harder to audit or revise. How would you balance throughput with the information value of a handoff when designing an AI-assisted research-to-writing workflow?
Absolutely! The 90% revenue increase wasn't from faster tasks, it was from redesigning entire workflows.
At https://www.designstudiouiux.com, we see this constantly: companies optimize the email while missing the chance to eliminate the whole Excel → QuickBooks → email chain.
The shift from "How can AI help this task?" to "How should we design this from scratch?" is everything
This is the part many teams still want to skip. They bolt AI onto yesterday’s workflow and call it transformation. Usually it is just legacy process running with better caffeine. The real shift starts when the sequence of work changes, not when the tool menu gets longer.
Fantastic!
This really resonated with my experience building and consulting, especially the end to end org & workflow redesign and parallel experimentation.
Mapping and after that getting people to adopt systems. Many pre-Ai automation attempts failed because the recording of data (the machine’s only access to the universe) relied on humans being open, disciplined and consistent.
By the way, I think I caught a typo…
‘[…]and that 40% of respondents (i.e., 4/5 of those that tried) […]’.
I make 4/5 80% ? …proves you’re human :))