Should you hire an AI Townie? 

It's just the latest of questions I never thought I'd ask. I grilled Berky, my debonair, otter-presenting AI alter-ego from Town.com, to help me answer it.

Thanks to the recommendation of my nephew, Ari Berkowitz (the one Berkowitz in this industry I know I'm related to), I had to take Town for a spin. When you get to Town, you create an AI assistant and integrate as many of your tools and data sources as you're comfortable with. Your Townie then gets to work, and you can bother it any time; I talk to mine mostly by emailing [email protected], but there are various ways of engaging.

The Townie pens daily briefings, drafts email responses, gathers research, schedules meetings, and launches custom automations called "routines" that run on their own. 

It's subscription-based, using credits for different tiers; the free trial includes two weeks on the $60/month pro plan; there's a plan for $15 with fewer features and credits, a free one that doesn't seem to do much, and more expensive tiers for power users.

Credits are the bane of any AI user's existence.

Imagine if, for food, you survive on Cheerios, and instead of them coming in a box, you routinely go to this seemingly infinite trough and take what you need; you discover that some days you eat 500, and some days you eat 5,000. Then you're told midway through the month that you've eaten your allotment of 15,000 Cheerios. If you want to survive, you have to either pay for a bigger tranche, or you have to be mindful earlier in the month to not eat more than an average of 500 a day. 

Except Cheerios are tangible. And they make you full. Credits are immaterial, and if you're using an AI service you like, there never seem to be enough credits to satisfy you.

I asked Berky the Townie otter where my credits were going, and it told me my credits per routine. Berky let me know that my biggest credit drain was spent reading emails, auto-tagging them (not sure what that meant), and drafting responses (only occasionally helpful). I turned that off, and also stopped the email briefings covering each meeting. That may be useful for sales meetings that appear on your calendar, but it's less helpful when you have some idea about who you’re talking to.

The most useful routine I created was the one you see the results of below: a briefing to help me scout news to include in this very newsletter. I already created a Custom GPT in ChatGPT and a Gem in Gemini to help with that, and the routine from Town did a better job the others did with less training, so that was the biggest win I’ve had with Town so far.

I asked Berky about Townie power users. What do those in the top 0.1% do?  And what could I do to become such a user? Berky said it came down to the depth of routines. Connect a CRM so it can help with deal flow. Turn calls into action items with pre-drafted messages. Create LinkedIn posts based on projects I'm vibe coding. Maybe a salesperson would view this differently, but as a marketer, little of this felt transformative, or worth paying for.

After interviewing Berky, I asked it to write the column based on the conversation in my voice. I tried to save myself a few hours this week. Since it had access to my inbox, it could read my past columns first to capture the style.

It was terrible. If you don't love this edition, at least know a human wrote it. I also considered doing one of those posts where the writer "interviews" an AI bot, but those started feeling tired in 2023. When I gave Berky a harsh critique of the draft, it admitted it shouldn't have even offered to do it, and it said it will never ghostwrite my columns again. That’s for damn sure.

I may not become a Town power user, but I like the format of being able to connect with the same AI agent across channels. Having one assistant that you use across email, SMS, WhatsApp, Slack, and elsewhere will have to be the norm. Integrations are key and will be table stakes. 

The infrastructure is there. The two biggest questions I have are to what extent the credit system will limit adoption, and whether the major AI engines (especially OpenAI, Google, and Anthropic) will roll out good-enough versions of such agents.

Go ahead, though. Introduce yourself to a Townie. Hire one for a couple weeks. Put it to work. Just don’t hire Berky. Even AI otters need a break.

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: Brand Marketers, Creative Directors, Performance Marketers

What: Cuisinart and agency Global Overview ran a one-month head-to-head test on Amazon: identical spend behind an in-house produced video and an Amazon AI-generated video. The AI-generated creative outperformed. The brand fine-tuned one Amazon-proposed video from a short list and ran it with minimal post-production fixes.

Why it matters: A controlled, same-spend A/B test from a real brand is a more credible data point than a vendor case study. The pace of improvement in AI video production, from unreliable a year ago to competitive today, is the signal for creative teams.

(AdExchanger)

Who: Brand Marketers, CMOs, Content and SEO Teams

What: ADWEEK surveyed marketing leaders from Bayer, Zoom, and a health tech firm on their biggest AI wins. Bayer is actively building AI discoverability across six brands using Profound. Zoom automated product naming with an internal AI agent. A health tech SVP cited using Profound and Writer together to scale LLM visibility.

Why it matters: This is a rare look inside the GEO implementation stack at real enterprise brands.

(ADWEEK)

Who: Brand Marketers, Performance Marketers, Agency Strategists

What: Japan Airlines partnered with agency Jellyfish to deploy its proprietary Share of Model tool, which measures how frequently LLMs mention the airline versus competitors. Based on those insights, JAL introduced 150+ new search themes and shifted budget away from lower-performing markets toward higher-value international routes. Results: 56% increase in conversion rate, 14% improvement in ROAS, 21% reduction in CAC, and 12.5% reduction in overall media spend.

Why it matters: This is what GEO/AEO looks like in practice. A major brand is actively using LLM visibility data to redirect real media dollars and getting measurable performance results from it.

(Aviation A2Z)

Liking AI Brief? The Refresh is a weekly recap of all the most important advertising news, written by AdTechGod.

Thanks for reading this week’s AI BRIEF by Marketecture. 

Marketecture Media is committed to bringing you the latest news, opinions, and community updates. 

If you are interested in advertising on this newsletter, our podcasts, The Advertising Forum, or in our Slack community, you can contact us here.