Apr 1, 2024 · 31m · another-podcast
Looking for AI use-cases
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Benedict Evans and Tony Karen Brown analyze generative AI adoption trends, public trust deficits, and the ongoing search for durable enterprise and consumer use cases. Drawing analogies to historical platform shifts like VisiCalc, they evaluate the transition from open chatbot prompts to structured, ambient software workflows.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 76.1% of the talking time here. How this is scored →
The hosts as informed peer 7.4 Guest teaching 2.6 Guest disagreement 1.3 The hosts pushing back 1.3
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:35 Brown rejecting team coordination dependency
Brown challenges the conventional enterprise adoption parallel by pointing out that unlike collaborative tools like Trello, individual AI use requires zero team buy-in to provide immediate output.
Hardest push from the hosts ▶ 8:46 Evans challenging trust in general knowledge queriesEvans challenges the assumption that LLMs are great for learning unfamiliar domains by advising people to test LLMs on subjects they know thoroughly to see the subtle errors.
Biggest teaching moment ▶ 6:53 Brown detailing multi-tool filmmaking workflowsBrown educates Evans on how modern creators actually use AI in practice, detailing a filmmaker using distinct specialized tools for character design, animation, and world-building.
The host holds their own ▶ 19:39 Evans on enterprise SaaS and structured workflowsEvans demonstrates deep domain analysis by explaining how WPP and enterprise SaaS wrap raw models into bounded UI pull-downs and workflows rather than exposing open prompt boxes.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| ChatGPT Adoption Statistics and Public Trust Metrics | 7 | 1 | 1 | 1 | Evans opens with Pew Research adoption data across age demographics and frames the fundamental problem of finding real utility beyond narrow initial use cases like code and brainstorming. | |
| AI as Platform Shift and the VisiCalc Analogy | 8 | 1 | 1 | 1 | Evans draws historical parallels with Dan Bricklin's VisiCalc and platform shifts, explaining why revolutionary horizontal tools initially seem irrelevant to specialized non-target professions. | |
| Practical Daily Workflows and the Problem of Factual Accuracy | 7 | 4 | 2 | 2 | Brown shares concrete examples of filmmakers and marketing execs adopting multi-tool AI workflows, while Evans dissects the difference between hallucination-tolerant creative work and factual verification. | |
| Preference Articulation, Machine Learning, and Curation Unbundling | 8 | 3 | 1 | 1 | Evans compares the difficulty of prompting subjective travel preferences to machine learning cat classification, while Brown relates this to pre-algorithm lifestyle bloggers and curated human taste. | |
| Enterprise SaaS Proliferation and Structured Workflow Integration | 8 | 3 | 1 | 1 | Evans breaks down enterprise SaaS unbundling of Office/Oracle and WPP's structured LLM workflows, while Brown brings up the real-world consequence of Williams F1 managing components via Excel spreadsheets. | |
| Limits of No-Code and Ambient Workplace AI Deployment | 7 | 4 | 2 | 2 | Evans argues why bottom-up no-code tools fail to penetrate rigid enterprise back-offices, and Brown highlights why AI spreads faster because individuals do not need team consensus to see immediate personal value. | |
| Enterprise Proof of Concepts and the Search for AI Utility | 7 | 2 | 1 | 1 | Evans details Accenture's quarterly generative AI revenues and explains how enterprise spending is currently concentrated in small POC pilot projects rather than holistic transformations. |
Statements from this episode (13)
Evans: Pew data shows 18% of US adults have used ChatGPT
“US adults, 18% have used chat GPT ever, and 33% of 18 to 29, up to 43% have used ever.”
Evans: Roughly one-third of Americans under 30 use ChatGPT for work
“Something like a third of Americans aged under 30 have used it for work for some reason.”
Evans: Only 2% to 3% of Americans trust ChatGPT output
“So a third of people haven't heard of it and two or three percent would trust what it says.”
Evans: Generative AI had four primary use cases in 2023
“It seems to me that there were sort of four use cases. Use case one is coding. Secondly, brainstorming and suggesting ideas. Thirdly, and this is more like a corporate thing, knowledge management pointed at an internal corpus of thousands of documents or PDFs …”
Evans: Ad agencies replace 10-person two-week workflows using ChatGPT and Midjourney
“ChatGPT plus mid-journey means two people in a day instead of a team of 10 people in two weeks.”
Brown: Generative AI Cuts Movie Trailer Production to Weeks
“One of my good friends is a filmmaker, and AI has completely shifted the way that she can create, for example, the trailer. It used to, you know, you'd used to need, what, a million in budget and a couple of months to create a good trailer for a movie? She can…”
Evans: Generative AI requires error tolerance or easily verifiable mistakes
“So you need to have a use case where either it doesn't matter if it's not a hundred percent right, which is, you know, Give me 50 ideas for slogan for a new brand of toothpaste. Like there's not, you're not, there's not like a complete. There's no right answer…”
Evans: Articulating personal taste to AI mirrors classic machine learning problems
“It's hard to tell a computer how I recognize a cat, and that's the same category of problem as it's hard to tell a computer why I like that suitcase, or why I like that hotel, or why that's a bad flight to book.”
Evans: LLMs are rebundling the curation that influencers unbundled from magazines
“Influences unbundling for magazines and TV, but mostly for magazines. And of course, LLM then, you know, ChatGPT goes off and reads the whole web and synthesizes it for you, so it's bundling all the influences back up to the top again”
Evans: Bottom-up software adoption caps at 5-10% TAM before needing enterprise sales
“And there's been like two things in the history of enterprise software. This has worked. I mean, one of them is Slack and I forget what the other one is. And what happens is you get to like five or 10% of your tan, and then you run out of the kind of people wh…”
Evans: ChatGPT traffic dropped over summer 2023 because students went on break
“This is the reason why ChatGPT traffic fell off last summer, because all the kids went home from school.”
Evans: OpenAI's ChatGPT plugins and custom GPTs have clearly failed
“We also incidentally saw like part of the attempt to make this work with first of all the plugins and then GPTs, neither of which have worked clearly.”
Evans: Accenture Booked $1.4B From Generative AI Over Past Year
“So Accenture revenue from generative AI projects in the last four quarters, a hundred million dollars, two hundred million dollars, 450,000,600. Well, over, over 450 and over 600. So they've done 1.4 billion probably in the last 12 months, and 1.1 billion in t…”