Review

Review: this week's AI launches for marketing teams, and the one finding that matters more

By Michael Vary, fractional CMO & Chief AI Officer

· 3 min read

Here is how I would sort this week's AI launches for marketing teams, based on what has been announced. I have not tested all of them myself yet, so read these as first assessments, not test results.

The reviews

Figma Weave inside AI agents. Verdict: watch closely. Figma says its Weave video and image models, including Seedance, Kling and Nano Banana, can now be used from external AI agents through the Figma MCP. In plain terms, an assistant you are already talking to can produce an image or a video mid-conversation, without anyone switching tools. You need a Weave account linked to Figma, and every image or video uses Weave credits. For a content team this is a real shift: the brief, the copy and the visual can come out of one conversation. It also means someone has to own the credit budget and the approval step, because generating is now easier than reviewing.

Hallmark. Verdict: worth trying for anyone shipping landing pages. Hallmark is a design skill for coding assistants such as Claude Code, Cursor, Codex and OpenCode, built to avoid the generic look of AI-made pages. It ships with fifty themes and 755 openly licensed photos, it can build pages from a prompt or a real product, and it can redesign and audit existing pages. Each page comes out as self-contained HTML and CSS. The audit feature is the part I would test first. A second opinion on a page that already exists is lower risk than letting a tool design from scratch.

GPT-6.1 Terra. Verdict: test it on your own tasks before switching. OpenAI released GPT-6.1 Terra and describes it as faster and more cost-efficient, with better visual reasoning, computer use and writing. Those are the vendor's claims. The only benchmark that matters for a marketing team is your own: run last month's real briefs through it next to the model you use now, and compare the drafts your editors would actually keep.

Understory's B2B LinkedIn swipe library. Verdict: use it this week. No AI required. It is a library of 42 B2B LinkedIn ads chosen from 13,483 creatives across 251 advertisers, broken down into 14 ad formats with notes on visuals, headlines, copy and calls to action. Before you ask any AI to write ad variations, give it good examples. A curated library like this is exactly the kind of input that makes the output better.

The finding that matters more

The most useful item this week was not a launch. A study found that people working with both AI and a colleague did 10.2% better, compared with 9.6% with AI alone and 6.3% with a colleague alone. They were also roughly three times more likely to rank among the top performers in expert evaluations. The catch: the benefit disappeared when the two people had the same skills.

My read: AI works best as a third member of a pair with different strengths, such as a strategist with a designer, or a copywriter with an analyst. It does much less when it is handed to one person as a shortcut.

A story from monday.com points the same way. Their former VP of AI GTM described putting a voice agent on the first call to every new lead. They started small and gave the agent more leads only once it matched the reps on the ones it already had. Wait time for a callback went from about 25 hours to under two minutes, and the team still reviews call scores for 20 minutes every morning. The people stayed in charge, and the agent earned more work one step at a time.

What I would do

The tools will keep coming every week. The team design is what decides whether any of them pays off.