"I don't think AI will make your processes faster."
When this title appeared on the Hacker News front page, the comments section immediately exploded. 34 upvotes, 11 comments — not a lot on HN, but the sentiment in the comments was interesting: almost everyone was bringing their own examples to prove the author right or wrong.
The core argument is counterintuitive: AI isn't an accelerator, it's a restructuring tool. If you stuff AI into an already inefficient process, you don't get an "accelerated inefficient process" — you get something more expensive, more complex, and slower.
Why Most Companies' AI Deployments Are Slowing Down
Here are real scenarios.
An e-commerce company's customer service flow: user sends email → agent reads email → checks order system → replies. Now they add AI: user sends email → AI reads first and generates a draft → agent reviews draft → checks order system → modifies draft → sends.
Two extra steps. The AI-generated draft isn't the final answer — it must be reviewed by a human. And reviewing AI output often takes longer than writing a reply from scratch, because you need to judge whether what the AI says is correct.
A consulting firm's report flow: analyst collects data → writes analysis → manager reviews → revisions. With AI: analyst uses AI to collect data → uses AI to write analysis → manager reviews (takes longer because they need to check for AI hallucinations) → revisions (harder because you can't tell what the AI wrote vs what the human wrote).
The problem isn't that AI is slow. The problem is that "adding AI to an old process" inherently adds complexity.
Real AI Efficiency Comes from Process Restructuring
Companies that actually benefit from AI aren't "accelerating existing processes" — they're "redesigning processes."
A classic example is Stripe. Their documentation team doesn't use AI to "write documentation faster." They use AI to analyze user behavior data in real-time when people use their API, and automatically generate "you might need to see this" recommendations in the docs. The docs are still written by humans, but the process of figuring out "what docs users need" has been replaced by AI.
That's how AI should be used: not to make existing steps faster, but to eliminate unnecessary steps.
Why Nobody Wants to Listen
Because "restructuring processes" sounds much more expensive than "accelerating processes."
Accelerating processes = buy an AI tool, $20/month per person, boss is happy. Restructuring processes = redesign business logic, might involve organizational changes, takes months, and might fail.
So most companies choose the former. Then discover things didn't get faster. Then complain that AI is overhyped.
The HN author is right, but he only told half the story. The complete version should be: AI won't make your processes faster — unless you're willing to first admit your processes need to be torn down and rebuilt.
Primary sources:
- Hacker News: "I don't think AI will make your processes go faster" — Frederick Vanbrabant
- Stripe Documentation System Case Study
- Gartner: ROI Comparison of Process Restructuring vs Process Acceleration in AI Deployments