Happy Thursday!

Jacob Posel and I have been Twitter friends for a while, and I've been an HQ user for a couple months now, so it was time to get him on the pod. My favorite take of his: when AI rollouts fail inside a company, it’s an infrastructure problem, not an AI problem.

Vendor of the Week

Paramount Ads Manager — Know what your TV spend is doing for your business.

If you've been holding off on testing TV because you're unsure how you'd measure it, Paramount is worth looking at. Their Pixel connects someone seeing your TV ad to what they do on your site afterward, including whether they buy or not. You can install it on Shopify, and it also lets you retarget site visitors with TV ads.

Midwest Photo used that retargeting capability and saw a 41% higher site visit rate in Paramount's beta testing.

They’ve also added measurement partners, including Haus and Northbeam, so you can look at your TV campaigns alongside the rest of your marketing. If your team already uses one of these tools, you have a way to evaluate the test through something you're familiar with and that’s already a part of your media planning workflows.

You can run across Paramount+ and Pluto TV with CPMs starting at $7. There's also an ad credit available through participating measurement partners toward your first campaign, so check the offer with your partner if you're considering a test.

Check out the Paramount Ads Manager Partner Directory here.

TV Strategy Session

Streaming hours climb 20% in Q4.

Holiday viewership spikes 2-3x around Thanksgiving and Christmas. And TV ads are the single highest-influence format for holiday gift purchases, ahead of social.

Most brands still aren't planning around this.

On September 30th, with Paramount, I'm hosting a free session on how to fix that.

Save your spot.

Why Your Company's AI Chats Are a Mess

Here's the thing Jacob sees inside every company that hands out Claude or ChatGPT/Codex seats: the distribution of who actually uses them is wildly uneven.

Two people go deep, build something great, and have no way to share it. Everyone else duplicates the work or gets scared off. You get more efficient individuals and a business that doesn't move.

HQ sits underneath all of it as shared context. I describe it as Dropbox sync for everything AI across your company. Jacob's version: Dropbox for agents, 1Password for agents, and agents for agents. If I build a workflow, connect a system, or fix a piece of context, you inherit it the next time you open Claude Code, Codex, or ChatGPT.

The slower folks catch up automatically because they're operating on top of what the best people already built.

Listen here: YouTube | Spotify | Apple

What I pulled from this one:

1. Fragmentation kills the compounding. Give 20 people Claude and you'll get 20 different levels of proficiency and zero shared benefit. Efficiency per person goes up. Revenue doesn't. The fix is a shared layer where one person's work becomes everyone's starting point.

2. Do the work first. Turn it into a skill after. Nobody can write the perfect landing page prompt up front. Let people build end to end, iterate with the agent, then say "note down every adjustment I made and turn this into a skill." That's how you get something real. An AI consultant interviewing your team and writing skills from scratch fails every time.

3. One person's taste becomes the whole company's template. Jacob's one-pager had all the right numbers and looked terrible. A teammate rebuilt it beautifully. Now every partnership, sales, and investor deck at HQ runs through her eye without her touching it.

4. AI didn't take the jobs because nobody was at capacity. The people who use it aren't getting fired. They're doing more, with fewer dumb mistakes: no more ad spend on out-of-stock SKUs, no more messy Shopify tags. Young brands with tiny teams are now running LTV-to-CAC by SKU and by subscription cadence.

5. Use it every single day or nothing else matters. Jacob's line: if the average employee used Claude Code daily for two months, they'd be the most valuable person in their company. Start with dumping a CSV in and asking questions. Just do a little more tomorrow.

That's all for today

If your team's best AI work is living in one person's account, you have an infrastructure problem you can fix this month.

It's Thursday, and it's all downhill from here. Get some rest, drink some water, and I'll see you Sunday.

Nik