When to chat with AI (and when to let it work)

AI with ALLIE

The professional’s guide to quick AI bites for your personal life, work life, and beyond.

You can win: the top referrer of this newsletter at the end of the year will
receive $1000 and a special bonus, and it’s anybody’s game.

🎁 see bottom of email for more info 🎁

Here's the wake-up call you didn't know you needed: if you're still thinking AI is all about chat interfaces, you're missing the bigger picture. Let me show you why, starting with a story from the trenches.

I coded my own chatbot in 2016. It was awful.

I remember it like it was yesterday. It was 8 years ago, and I was laying down on a black leather couch in Austin in the summer (because clearly I’m a masochist and was just hoping 8 years later, I would be able to write a story about it for a bunch of strangers on the internet). I cracked open my laptop and accessed IBM Watson Conversation to build my very own chatbot that would solve a real problem I had seen in grad school: onboarding new international students. Not only was there a ton of information for new students to sift through (government processes, school requirements, housing, etc.), there were also a lot of ‘hidden tips’ that only students held that weren’t published anywhere (ex: 4 unrelated students would go to Verizon and buy a family plan together to get cheaper US phone plans).

The process of building a chatbot in 2016 was extremely manual and about as elegant as programming a VCR. You had to predefine all of the intents you thought would come through from users (like “How do I start a bank account?”) as well as map out versioning of those intents, so that a model could recognize the same question regardless of the wording.

Then, ChatGPT and the LLM revolution came and everything changed. Systems became more flexible, chatbots could handle a wider range of questions, and we could understand what people meant without us having to program every possible use case. LLMs brought about a tech shift that was, in a word, more natural.

Because of this breakthrough, combined with years of chatbot familiarity, everyone got excited about chat-based AI. And you probably saw it everywhere:

- Chatbase was letting you drop a chatbot onto your website to answer FAQs

- Perplexity enabled a new kind of search engine

- Microsoft Edge built AI chat directly into your browser

And that's where the majority of buyers and users got stuck. The chat paradigm became a trap, with users going back and forth with these systems just to refine their queries or get follow-ups. People started thinking chat was the answer to everything.

AI is not just chat.

The way we work with computers remains stubbornly orthogonal to millions of years of evolution.

Humans (well, ancestors to humans) started walking on two legs 6 million years ago. We started speaking maybe 70,000 years ago. And yet, our main way of working today, is to sitting on our butts in a stiff computer chair and type a 5-word search query into Google. ARE YOU SERIOUS.

This is why chatbots took off: they let us do what we've been practicing for 70,000 years - have a conversation.

But limiting AI to conversation is like using your smartphone just to make calls. AI's real power lies in ambient intelligence - it's appearing in spreadsheets, running your 10am reports automatically, powering smart doorbells that flag Amazon packages, and enabling gesture controls that finally feel natural. These systems don't need you to chat with them - they just work.

Let me show you an example that breaks free from the chat paradigm.

NotebookLM: Google’s Surprise Hit

NotebookLM is, in my mind, the best AI release from September 2024 - yes, even better than Replit Agent. Here's why: you can upload up to 50 sources (documents, videos, audio files, website links, Wikipedia pages, whatever you've got), and without any chat, it automatically creates six different assets for you. Zero prompting required.

The killer feature that made it go viral is a fully AI-generated podcast with two AI voices discussing your content. So you can upload your brand's recent blog post, and you get what sounds like two people genuinely excited about your latest announcement. I created my own AI podcast course with it here and showed a ton of examples on Instagram Reels. (Quick but important: check their privacy policy before uploading anything sensitive.)

Their AI podcasts stole the show, but when you upload your documents, you also automatically get the following for free:

• A study guide that makes sense

• A table of contents

• An FAQ

• A timeline

• A Q&A bot that knows your documents inside and out

Here are just a few other ways you could use this:

1. Turn your company's annual report into an engaging podcast

2. Create an interactive FAQ for your product manual

3. Generate a timeline of your industry's history from multiple sources

4. Produce a study guide for your online course content

5. Develop a Q&A system for your company's knowledge base

6. Synthesize research papers into digestible summaries

7. Create an executive content briefing from multiple competitor blog posts

8. Generate a podcast discussing the key points of a long-form research paper

Prompting isn’t going away any time soon (and the top AI labs all seem to agree), but also…users are craving AI use cases that don't require prompting. Here's the reality: 99.4% of users aren't developers, and even if they wanted to build it themselves, they don't have the time. Abstraction is what drives the tech industry forward. If you're solving real problems and saving people time while adding features like security, collaboration, flexibility, or access to better compute - that's worth something.

More non-chat: auto code, visualization, games, video gen, agents

My team has done everything from creating games in Claude…

To generating AI Allie videos with HeyGen…

To multi-card trip planning with Cove AI…

To robot love simulations and visualizations (also with Claude)…

To AI agents that do things like prioritize and summarize AI news every morning…

The best enterprises are prioritizing multimodal experiences and integrations. Just like marketing evolved from single-channel to omnichannel, brands must evolve to what I call "omnisense." People are going to interact with your brand in any modality they choose, not just through web, support calls, or social media DMs.

We're moving into a multimodal paradigm where large language models are opening up alternative ways of prompting systems and getting what you want. Prompting doesn't mean typing perfectly curated paragraphs anymore - it can look like many different things.

The bottom line? If you're still thinking about AI in terms of chat interfaces, you're stuck in 2023. The future is multimodal, it's ambient, and it's already here. The only question is whether you'll be ready for it.

As always: stay curious, stay informed,

Allie

P.S. NotebookLM is free right now. Just saying.

Looking for an AI course? I got you.

I’m an AI instructor on MasterClass, LinkedIn Learning, and Maven. Whether you’re an exec, an AI champion at work, an AI-curious parent, or just someone who’s played around a few times with ChatGPT, you’ll find something that helps you level up.

These courses teach you:

• How to leverage AI to boost productivity

• Ways to amplify your creativity

• Techniques to automate routine tasks to stay ahead

• How I do AI-first work with an AI-first team

Join me and learn to make AI work for you.

Tools, courses, and blogs that caught my eye

Over the last few weeks, I’ve pulled together some of the top releases, and my take on each one. Check it out.

1. OpenAI turns their focus to series of reasoning models — A few weeks back, OpenAI launched the o1-preview series, introducing AI models that excel at complex reasoning tasks, outperforming previous models in science, coding, and math; access is available for ChatGPT Pro, Team, Enterprise, and EDU users. In their recent Reddit AMA, any questions about OpenAI Sora or Advanced Voice Mode were punted and redirected to o1 and reasoning—agents are coming (read it) (demo 1) (demo 2) (demo 3)

2. Artifacts is available on mobile for all Claude users — Artifacts were made available to all Claude users, enabling teams to collaboratively create and share work products like code snippets, prototypes, and dashboards across web and mobile platforms. I’ve had 4 mega wow moments in A the last 2 years; Artifacts is one of them and code gen is stealing a lot of AI lab attention right now - next time you’re in line for food or waiting for the subway, challenge yourself to code a game in under a minute (read it) (demo 1) (demo 2) (demo 3)

3. AI spending 6x’d from 2023 to 2024 — A recent Menlo Ventures report shared AI spending surged to $13.8 billion this year (more than 6x the $2.3 billion spent in 2023). The report also showed that the top two criteria used to pick generative AI projects are easily measurable ROI (30%) and the ability to customize the tool (26%). They rank code generation as the top use case in the enterprise, but that doesn’t match my experience (where customer support is number one) (read it)

4. You're still typing to AI? That's so last year. I am constantly dictating to AI because speaking is faster and more natural. It makes throwing a ton of context into my prompts incredibly easy, which makes my outputs honestly quite exquisite if I do say so. If you’re getting medium quality responses, test yourself by dictating a ton of context to an AI first (my demo)

5. A custom AI chatbot got conspiracy theorists to back off — 50% of Americans put stock in at least one conspiracy theory (does that seem high to anyone else?). Researchers set up a custom AI chatbot using GPT-4 Turbo to chat with 1000 conspiracy theories believers—confidence in their discussed theory dropped by 21% on average, the effect lasted at least 2 months. AI as a mediator and educator, yes please (read it)

6. Netflix bets on AI to transform gaming — Netflix is changing its gaming strategy by pivoting to AI-powered development and a growing library of free games for subscribers, promising faster production while emphasizing human creativity will remain central to the process (read it) (my thoughts)

7. Turn yourself into a video cartoon in minutes — Runway unveiled Act-One, a new tool for creating expressive character animations using simple video inputs. This allows artists (or just fun-havers like me) to generate high-quality, nuanced performances without the need for traditional motion capture tools or complex facial rigging (read it) (my demo)

8. Anthropic released their first AI agent to navigate and act on your computer — "Computer Use" was released, a new AI agent that can interact directly with your computer and take action. This feature allows the AI to perform tasks such as running coding tests, adding to shopping lists, or navigating the web by simulating actions like clicking, typing, and scrolling (my thoughts) (my demo) (my tutorial so you can use it)

9. Using Google NotebookLM to create a 15-episode podcast course — I built a 15-episode podcast course on AI in under a day using AI tools like Google NotebookLM, ChatGPT, and Claude. It covers everything from AI basics and machine learning to generative AI, ethics, multimodal AI, and superintelligence (read it) (hear it)

10. AI startups hit revenue milestones considerably faster — Milestones like $30M annual revenue were hit five times quicker by AI startups than traditional SaaS companies. Stripe’s data shows top AI companies hitting $1M in revenue in just 11 months on average, driven by global demand and rapid product iteration. However, high compute costs make profitability a challenge (read it) (my thoughts)

11. ChatGPT Enterprise may replace $1K/hr tasks, price hike hinted — OpenAI's CFO, Sarah Friar, revealed in a BloombergTV interview that the $60/month ChatGPT Enterprise subscription is being used to tackle tasks that allegedly cost $1,000–$2,000 per hour with a lawyer (though many people disagreed with her pricing estimates and said it was a few hundred per hour). She also hinted that pricing for ChatGPT could increase in future updates (watch it)

Feedback is a Gift

I would love to know what you thought of this newsletter and any feedback you have for me. Do you have a favorite part? Wish I would add something? Felt confused? Please reply and share your thoughts or just take the poll below so I can continue to improve and deliver value for you all.