012 AI For Operators

In this episode, I share three simple AI workflows that I use to run multiple companies, along with tactics to increase adoption that I’ve found useful:

Three Workflows

• Custom candidate researcher: A Custom GPT compiles bios, work history, posts, references, and suggested questions, so I can go into hiring and sales conversations with confidence and context.

• Centralized meeting note pipeline: A Granola/Zapier automation that captures all of my meeting notes (table stakes) and, more importantly, centralizes all of them in a Google Drive formula so I can easily query them with ChatGPT.

• Automated daily check-in: Context is king, and the best way that I’ve found to build context is to just tell ChatGPT everything that I’m doing each day. Every morning at 8 a.m., a workflow starts that asks for my outcomes from yesterday and priorities for today. Another benefit? It means I start the day by opening ChatGPT and then have no excuse not to use AI constantly.

Adoption Tactics

• Dedicated communications practices: A Slack channel where we share our learnings and tips, a weekly requirement to share what you’re trying out at team standups, and more formal learning during monthly all hands meetings.

• Improve output with more context: Enable ChatGPT (or whatever you use) to access your calendar, email, Drive, CRM, and more - it’ll expand your capabilities massively with access to tailored info.

• Daily habits: Embedding small, repeatable AI workflows builds the mindset you need to identify bigger opportunities.

The Bottom Line

Integrating AI has massively expanded our margins, allowed us to ship faster, and taken busy work off our plates. The business impact is undeniable.

What It Does

A Chromium‑based, AI‑native browser that can act on the web (clicks, types, submits forms), reference what’s on your screen, and connect to Gmail/Calendar to execute scheduling, email, research, and shopping workflows.

Why Ops Leaders Should Care

Time is our most precious asset. If Comet can save time and give you AI tools in the browser (where we all probably do most of our work already), it could help both you and your team level up.

Key Features (Pros & Cons)

Pros

• Agentic workflow automation where you already work: browser-based initiation of workflows means you don’t have to switch between tools to combine AI with your own actions.

• Awareness: sees what’s happening on screen (including in non-highlighted tabs), which means you’re generating context and can ask more useful questions.

• Enterprise readiness: with the standard SOC 2 certification, SSO, and easy ways to opt out of the browser using your data for training purposes, Perplexity has made Comet enterprise-friendly

• Citations: Perplexity is still the best consumer-facing tool if you want to double check where your information is coming from.

Cons

• Platform maturity: Comet rolled out on Windows only a few days ago, so it’s likely still going through some testing and optimization - expect some bumps in the road.

• Requires full access to your data (and tracking your browsing, etc.) to maximize how useful it is.

• Reputational/legal exposure: Perplexity seems like the shakiest of the big AI players, with real legal exposure (Cloudflare’s allegations, etc.), a smaller war chest, and a reliance on the foundational model companies.

An Operator’s Perspective

I’m ChatGPT-first, so while I’ve tried Perplexity in the past, it’s been a few months. Lots of companies are saying that agentic browsers are the next competitive battlefield, and Perplexity is the only scaled player with one in the wild today, so I thought I’d give it a whirl.

Onboarding was relatively simple (as a PC user at work, Comet wasn’t available to me until very recently), but obviously required me to download the system to my desktop. Perplexity is very invested in getting you to switch Comet into your default browser spot - I’m sure that’s how you get the most out of it, but porting over my settings from Chrome was sufficient.

One thing to note up front: the place to adjust your privacy settings (whether Perplexity is allowed to train on your data or not, for instance) wasn’t hidden, but they didn’t exactly push it front and center, so make sure you visit Account > Preferences > Artificial Intelligence to de-select AI data retention if you’re working on anything that you don’t want them to train on (you probably are).

Using Comet was pretty intuitive: the form factor of typing into a chat box has become second nature at this point. I tried a simple research project and the results came back very quickly (there have been complaints of slowness from users on Reddit, but that certainly wasn’t my experience) and they were fine, not great. They improved with a bit of additional prompting, but the trade-off for speed is, logically, depth: the variety and quality of sources didn’t impress me.

Next, I connected my email and calendar (head’s up: they ask for a ton of access!) and asked for a daily briefing. Again, the results were fine, not great. With some additional prompting, I managed to make it more tailored and useful, and with some dedicated time I’m sure I could turn it into something that provided substantially more value.

Other Options

Bottom Line

It’s worth checking out, if only to start acclimating yourself to a future where browsers will have agents enabled by default (whether or not the quixotic Chrome bid proves successful), but make sure you tune the privacy settings to fit you before you give it access to your data and browsing history. It hasn’t earned a place yet as my default browser, but I’ll keep trying it out.

Practical

• Beyond the em dash: A practical guide on how to tell if something is AI written (along with a more nuanced look into how AI systems ‘think’).

• How to build your AI operating system: Ben Tossell describes the interlocking systems that he uses to run his business and life.

• Best practices for building agentic systems: How to break down task, manage context, handle errors, and optimize performance.

• Compound engineering - the future? How Every changed their development practices to make sure their systems get better with every change.

Perspectives

• Job loss may not come arrive you think it will: Just because a job type is exposed to automation doesn’t mean that job loss automatically follows. New research suggests that information-processing roles suffer while customer-facing and coordination roles will actually gain an economic advantage. A helpful perspective for your next debate on the potential impact of AI.

• An AI vibe shift? Are things getting a little more pessimistic/realistic out there?

• GPT-5 isn’t for you: If you’re reading this, you’re an AI super user, at least relatively speaking. This piece by SemiAnalysis argues that the actual innovation of GPT-5 is the model router, not the improvement in model capabilities, because it sets the stage for much higher monetization via ads.

News

• Interview with Nick Turley: The Head of ChatGPT talks about what it’s like to build products for 700 million active users, how they’re responding to the backlash to the 4o deprecation, and more.

• Billions Isn’t Cool: Sam Altman is back on his hype game, saying he’s set to spend “trillions” on AI infrastructure.

• Perplexity to buy Chrome? There’s disagreement over whether the $34.5B, all-cash offer is a real bid or a gambit to tie up Alphabet in antitrust proceedings. Either way, Perplexity’s best product may be its PR.

Thanks for reading,

Tom Guthrie

If you liked this edition of AI for Operators, share it with another operator you know!