Robert Gillespie

Buyers can shortlist you in ChatGPT without ever visiting your site. This week’s AI marketing roundup: • Measure AI visibility • Edit videos with Claude • Prepare for ChatGPT’s shared usage limits Plus new tools and practical playbooks. Read it here: getmarketingwithai.substack.…

Netflix is testing AI to help people choose what to watch. A useful question for marketers: could AI help customers make a decision, rather than give them more content to scroll past? My latest for Inc: inc.com/robert-gillespie/net…

Can a small bot team replace a $399/mo AI-visibility tool? @shawnbuilds says yes: visibility, content, GTM, X, under a chief of staff. He claims a jump in ChatGPT/Claude inbound after a week. His claim, not mine. Dashboard, or four narrow bots?

Dan Martell put it plainly: you're paid to make decisions, not clicks. I wrote about my Grok Bot setup on Wednesday. The bit I'm still tightening is what he calls review mode. - Start with a chief of staff. - One bot per job. - Then every Friday, compare the drafts your inbox bot wrote to what you actually sent. - Words you cut. How long you run. How you sign off. That gap is the training data. Not another prompt library. I still press every send button. The bots process. I decide. Full guide here: getmarketingwithai.substack.…

Menlo just dropped their 2026 enterprise AI numbers. $177B in deployment spend. Up 5.4× in a year. Cool headline. Here’s what I found interesting from a marketing standpoint: Agents are eating the tokens. On OpenRouter (a platform that routes traffic across hundreds of AI models), agents now generate ~80% of usage. Anthropic is responsible for 55% of enterprise LLM API spend. Translation (and this is the key point) Budgets are moving from “chat with a model” to “pay for work completed.” If your AI plan is still prompt libraries and one-off drafts, you’re buying the wrong thing. The marketing teams that win over the next 12 months won’t have better prompts. They’ll have better jobs defined, better context and a human who still signs off. That’s the whole game.

One line in your robots.txt will soon tell Google, Microsoft, OpenAI and Anthropic not to train on your site. And Bing will still let you appear in its AI answers. It’s one of 10 stories in this week’s AI marketing roundup: Google is taking ads in AI Mode to 10 more countries, and the UK is next Gemini can now edit an entire Slides deck from one prompt An agency shows how it chains Claude skills into workflows that send weak drafts back Every slide ends with one thing to do this week. Swipe through, then read the full edition on my Substack.

One week with Grok Bot. Here's what I found. Most of my AI work happens inside Claude. But over the last 2 weeks, I've tried something different and spent time setting up Grok Bot as an actual assistant, not a chatbot. I didn't expect to end up with a small team. It started with one bot acting as a chief of staff. I talk to it, and it hands work to specialist bots. A week later I have eight of them: - Content Scout checks X and the news each morning for AI marketing stories worth writing about - Copywriter drafts headlines and promo copy, modelled on Ogilvy, Halbert and Brunson - Second Brain files every idea and link I send it into Notion, and brings them back before I write - Lead Gen researches companies that look like the ones I already train - Growth Bot keeps me focused on paid work, not just free content - Dream Team is my personal board: a committee of people I admire, follow and trust, whose take I'd want if I could sit with them in real life - Plus two that help with my Substack and Inc articles Five things I learned: 1. Think team, not tool. One bot trying to do everything gets messy. One bot per job stays sharp. 2. It works when I don't. I get a brief at 8am with calendar clashes, what needs action in my inboxes, and two or three story angles. 3. This was the biggie. The bots can talk to each other. My chief of staff coordinates them, so I'm not copying things between chats. 4. Set the rules early. Mine draft only. Nothing gets sent or published without me pressing the button. 5. It pushed back. When I talked about growing the business, it ran the numbers and told me where the real gap was. Useful, if slightly uncomfortable. The biggest differentiator for me has been what happens when I close the chat. My specialist bots keep talking to each other, routines fire, reminders land, and the work still moves on my computer without me in the room. Next week I'm using it to turn ideas into content faster and to be more proactive about training work. And I have a full write up planned for my Substack newsletter next week if you fancy joining the 20,000 others who get it weekly.

Meta's new terms make you "solely responsible" for its AI ad edits. That includes the ones Meta switches on itself. It's one of 10 stories in this week's AI marketing roundup: - Claude Duet works across your apps while you're away - Three studies say AI search sends almost nothing, and a rebuttal says you're miscounting it - Forrester expects marketers to cut a third of their AI initiatives in 2027 Every slide ends with one thing to do this week. Swipe through, then read the full edition on my Substack.

Claude quietly became a place to finish work, not just start it. Docs, Slides, Design, Dashboards and Motion. All in one place now. I wrote the marketer's guide to Claude Artifacts: what each one does, how to get everything on-brand, and a prompt for every template. getmarketingwithai.substack.…

New guide: The Marketer's Guide to Claude Artifacts. Claude has now become a place to finish work, not just start it. Docs, Slides, Design, Dashboards and Motion now live in one place inside Claude. In the guide I cover: → What an "artifact" actually is now → Which plan gets you what → How to set up your design system so everything comes out on-brand → A ready-to-adapt marketing prompt for every template → How to turn one campaign brief into a full set of assets And if you use Claude Design, the standalone version closes on 14 December. There's a step-by-step on moving your design systems across before then. Read the full guide below 👇 getmarketingwithai.substack.…

Would you trust your CRM to brief a new marketer? Then why trust it to brief an AI agent? Think about what’s in there. - Old job titles. - Duplicate contacts. - Customers who should be excluded. - Prospects sales and marketing disagree about. Now give an agent permission to choose an audience and run a campaign. Those gaps start affecting real decisions. Serious ones. This is why the features that caught my attention in the latest HubSpot and Salesforce announcements were the ones lurking behind the agents. It’s easy to get excited about faster execution. But “launch a campaign for our best prospects” needs an agreed definition of “best prospects”. And “we saved ten hours” tells you little about whether the campaign worked. My advice if you go down this path: start with one campaign. 1/ Check the information it depends on. 2/ Resolve the disagreements. 3/ Make someone responsible for keeping the data current. Then judge the result by qualified opportunities or customer action, alongside time saved. You don’t need perfect data across the business. But you do need reliable information for the job you’re handing over to the agent. That’s the focus of my new article for Inc. Get the full details in the usual place below.

Marketers are using AI to write more emails. And customers are using AI to read fewer. I think we’re spending too much time on the first. And with Apple’s latest iPhone update not enough on the second. Apple Mail already uses AI to summarise emails. But this new update goes even further, with it now suggesting actions based on the emails contents. Which means customers can get what they need without reading your email in full. And that changes how we marketers need to think about the message. You might design an email with a headline, images and carefully written copy. But your customer might only see a short summary or a suggested next step. So after Apple's AI suggestions kick in: 1/ Does the reason to buy survive? 2/ Are the terms still clear? 3/ Is the next step obvious? Or does removing the build-up reveal that you haven’t said very much? For me, this makes clear, specific copy more valuable. - Tell people what’s on offer. - Why it matters to them. - What they need to do. Give your onboarding emails and renewal reminders the same attention. They help customers use what they bought and decide whether to stay. So, before your next send, try reducing the email to two sentences. Would you or your customers still act on that information? I explore what this means for marketers in my latest article for Inc. Link to full article in usual place below

Do you remember when every phone had its own charger? ​You'd be at a mate's house on 3% battery and the only cable going was the wrong one. ​Samsung had one, Apple had another, and most of us ended up with a drawer full of chargers that fit nothing... ​...until USB-C came along, and with a bit of a push from Europe, pretty much everyone went down that route. ​One cable for (almost) everything, and I don't think anyone misses the old days. ​I had that exact thought this week when I read Google is turning Gemini Gems into Skills. ​On the surface it looks like a rename, but I think it's a bigger deal than that. ​Gems only really lived inside Gemini, so if your team built a load of them and then wanted to move to Claude or ChatGPT, you were basically starting again. ​Skills are a different story. ​Claude's been all in on them for a while, OpenAI's on board, and now Google's joining them. ​So the work you put into teaching AI how YOU do things is starting to become something you can take with you. ​That's a smart move for Google, but it's an even better one for everyone else. ​Because one of the biggest things holding marketing teams back is the worry of picking the wrong tool and being stuck with it. ​Less lock-in + less risk = more teams actually getting stuck in. ​Not groundbreaking news, I know. ​But a bit like USB-C, I reckon in a couple of years we'll wonder why it was ever any other way.

97.8% of US households aren't paying for AI yet. As someone who has been in the AI space since 2019, I find this stat incredible. At the same time, I am looking at this through a lens of what I do every day and what I read about here or on X. AI is everwhere in my 'bubble'. It's likely that most 'personal' AI use is being done on a paid company account. Or maybe the free versions are sufficient for 'basic' information or use cases. Most have never heard of 'Skills', never mind Muse, Dots or GrokBot. But I guess what it does tell me is we are still in the very early days of AI adoption by the general population.

Every marketing team I meet has built real AI skills and prompts: brand voice, cold email, campaign briefs. Then they save them all in a shared OneDrive folder that gets copied and tweaked until nobody's sure which version is the real one. If your team has a file called final_v3_ACTUAL_final, you've already got the exact same problem with your AI skills that you used to have with documents.