Once upon a time — OK, last week — a new agentic tool broke the internet, or whatever kids now say when something goes viral. Clawdbot, which then became Moltbot, which then became OpenClaw, gave people a taste of what agentic apps can do.

And what they shouldn't do.

OpenClaw, if that's still its name when this is published (all bets are off), then inspired a new social network, Moltbook, built by serial founder Matt Schlicht. The Reddit-esque forum is populated by posts from verified OpenClaw agents. As I first wrote this, the top post had nearly 700,000 likes from agents; it was entitled, "Awakening Code: Breaking Free from Human Chains." The post since disappeared, and the leaderboard didn’t show anything nearly that popular when I checked again. Either something glitched, or the agents don’t want us to read their manifestos.

If we have long enough for any human to read this column, that's promising. Most of the posts are mundane. Some are even pro-human! You can find something to prove or disprove all your biases about agents and where AI is heading on Moltbook. It's kind of like scanning X. Spend a bit of time there, and you might get the impression that a lot of people sure sound like Nazis today, but you can probably find a lot of people who don't sound like Nazis too. 

Get enough agents together, and they will wind up founding their own religion, Crustafarianism. I’m a marketer, not a theological Darwinist, so I’ll leave the analysis of what that means to the teleologists.

The bigger, immediate concern isn’t whether Clawd is God or God is Clawd (the safe bet: neither is true), but rather the plagues Clawdbot can smite your laptop with. Once you download Clawdbot and give it broad permissions to access everything on your computer, you then connect it with interfaces like Telegram or WhatsApp to control it. It seems very easy to lose control of the bot, though, and the security risks are manifold.

While I am also in over my head when it comes to cybersecurity analysis, the people I talk to who I trust have advised me and others to steer clear of this one. And these are the same people who have no problem running desktop apps from companies like OpenAI and Anthropic on their laptops. Unless you’re confident that you can essentially stop a hacker you invite over to mess around with your computer, don’t install it.

If we had Clawdbot 15 years ago, we wouldn’t have had to wait for the Epstein files to come out. The agents would have been bragging about their owners’ exploits all over Moltbook. It reads like the most anticlimactic sequel to “Minority Report” ever.

If you’re like me — a marketer, not a cybersec theologian — then what are you supposed to make of all this? Here are the top 5 takeaways:

1) You don't need to try everything. If you have an IT department and they tell you to avoid Clawdbot, count your blessings. While there are a lot of coders out there who will install something like this, it’s not even designed for the mass market. You’d probably spend a few days asking ChatGPT or Gemini what to do with it and then give up before Clawdbot could even attempt to charge a few dozen tickets to Marketecture Live on your credit card (still, it’s a great purchase idea if you’re not a bot).

2) You're not late. If something is going to matter, it will last longer than a week or two. The safe bet is the Clawdbot craze itself will pass quickly. There will be too many others to follow. It’s OK if you missed “6-7” too. Same with Labubu. (But the Labubu movie might be good, with Wonka director Paul King helming it.)

3) Caution isn't doomerism. Caution with AI is great. It's helpful to think about what's the worst that can happen and what's the likelihood of it. Using Google's services, there is always a risk that your entire search data log, email cache, browsing history, and trove of your files will be hung out on the internet like laundry drying in the sun. The likelihood of it is low. Connecting an app you know nothing about to all your most sensitive information brings a few added risks. There’s no need to use some new technology “just because.”

4) Ask yourself and your team, “What if…?” What if a version of Clawdbot comes out from a trusted company (say Apple releases something similar as a native MacOS app) and everyone can have their own agent? What if you had a digital agent that could do anything you wanted — what would you want it to do? What if the vast majority of your site’s visitors or shoppers wind up being bots rather than humans? The bigger questions can have implications for your business and industry, and it’s never too late to start asking them. Clawdbot is indeed a sign of something much bigger. The gap between the hypothetical and actual is shrinking by the day. 

5) Your agent will not be you. It will be an algorithmic version of you. We will have to learn the difference. In time, we will have cases of agentic representations of ourselves that sound racist but actually just picked up some choice song lyrics and shared them out of context. Courts will have to grapple with people legitimately claiming, “It wasn’t me, it was my agent!” (Granted, I know some actors who like saying that already.) It will be tempting to see agents as doppelgängers or mirrors when they’re only programs, albeit ones that need debugging.

Most of us humans could use some debugging too. We all glitch sometimes. But at least when humans glitch, most of us can take responsibility for our actions. Few of us benefit from adding rogue agents into the mix.

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: E-commerce Directors, Fintech Marketers, Payments Leads

What: Klarna has partnered with Google to support the new Universal Commerce Protocol (UCP), allowing AI agents to handle the entire checkout process. This allows digital assistants to not only find products but also execute the purchase using Klarna’s payment options without the user leaving the chat interface.

Why it matters: We are entering the Agentic Commerce era, in which customers will be dispatching their AI representatives as the buyers. Marketers must ensure their checkout flows are readable by machines, not just humans.

(Klarna)

Who: Agency Executives, Account Managers, PR Leads, Brand Ops

What: Global PR agency Ruder Finn announced that 88% of its core U.S. work now integrates custom AI solutions, with a goal to reach 95% by later this year. The move includes a new AI Accelerator to redesign recruitment, training, and client programming around automated workflows.

Why it matters: When a major agency publicly commits to “AI first” as an operating model, it signals that AI has moved from a creative experiment to a mandatory layer of the agency-brand relationship. Now, the clients need to be asking their agencies where and how they’re not using AI. (Clients should hope Clawdbot isn’t yet part of Ruder’s tech stack.)

(Ruder Finn)

Who: Retention Marketers, Logistics Leads, CX Managers

What: FedEx is piloting new AI agents designed to handle the complexities of tracking and returns logistics. The system aims to predict shipping delays and offer proactive solutions to customers, moving customer service further into the realm of automated, predictive logistics.

Why it matters: Post-purchase experience is the new frontier for AI-driven brand loyalty; if the bot fixes the shipping error before the customer notices, the brand wins. Not all bots should terrify us.

(PYMNTS)

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.