Destreza

Destreza / Thinking

AI at work

What changes when software can carry out an assignment, and who owns the result.

Newest first. Monthly analysis is ordered by its editorial “as of” date.

14 pieces

A group of friends around a dining table in a room with Sonos speakers.
Sonos 27 launch photography, released on 1 September. Its connection to external AI assistants enters Early Access on 8 September; this scene is launch imagery, not evidence of customer satisfaction. Source: Sonos.

Analysis 005

A better model. A better customer experience?

Analysis 0055 min read

September brings new AI controls and fresh evidence of customers’ reluctance to retry a failed chatbot. A launch needs to prove what happens after the conversation.

Read analysis
Two people and a dog sit on a sofa in a room with a Sonos soundbar and rear speakers; one listener wears headphones.
Inside The Weekly Cut 018: Sonos’s September 1 launch photography shows Beam Ultra, headphones and rear speakers in one home system. The products were announced alongside Sonos 27, the software update discussed inside this edition. Source: Sonos.

The Weekly Cut 018

ChatGPT ads expand. New models raise the bar.

The Weekly Cut 01810 min read

A wider advertising channel, more capable models and a staged rollout of image creation inside Google Workspace change the next marketing brief.

Read edition
A smiling Goldfish snack wearing sunglasses poses before a camera against a backdrop made to look like a slice of cheese.
Inside The Weekly Cut 017: Goldfish’s “Photoshoot”, from the campaign discussed in System1’s 27 August review. This is campaign context; the review’s 5.1-Star score relates to the execution it tests, rather than proof about this particular frame. Source: Goldfish / Mischief / Blinkink; still supplied by Marcus Lim SO.

The Weekly Cut 017

Campaign agents, holiday shoppers and creative control

The Weekly Cut 0178 min read

Media-buying assistance, preparation for returns and a customer-search experiment make AI’s commercial use more concrete.

Read edition
A can of Madrí Limón beside a glass of beer and cut lemons.
Inside The Weekly Cut 014: Madrí Limón, pictured in System1’s 6 August fruit-beer research. The study tests the appeal of product concepts; its 4.4-Star result for Madrí describes predicted potential, rather than observed sales. Source: System1 / Madrí Excepcional.

The Weekly Cut 014

AI moves closer to buying and making

The Weekly Cut 0148 min read

Snap opens its advertising data, ChatGPT adds product carousels and Adobe brings creative tools into the conversation.

Read edition
Nick Louisson presenting ISBA’s 2026 Gen AI survey, with the original subtitle “Nothing scales better than mediocrity”.
Nick Louisson discusses ISBA’s 2026 Gen AI survey in its 9 July preview. The results, released on 22 July, distinguish widespread use from reported business impact. Source: ISBA.

Analysis 003

Where should the time saved by AI go?

Analysis 0035 min read

July’s ISBA findings expose a gap between using AI and getting a business result. The next decision is what to do with capacity you can prove you have released.

Read analysis
Mohamed Salah and Vini Jr. sit on BMX bicycles beneath a concrete flyover in Pepsi’s Football Nation campaign.
Inside The Weekly Cut 012: Mohamed Salah and Vini Jr. in Pepsi’s April 2026 Football Nation campaign. Pepsi features in this week’s YouGov brand study; the research does not establish that this campaign caused the movement. Source: Pepsi.

The Weekly Cut 012

AI adoption is widespread. Business impact is uneven.

The Weekly Cut 0128 min read

UK advertisers report near-universal AI engagement, Google improves its working models, and Semrush tests who owns a topic inside ChatGPT.

Read edition
Anthropic’s OSWorld-Verified chart plots computer-use accuracy against average cost per task for Sonnet 5, Sonnet 4.6 and Opus 4.8.
Anthropic’s Sonnet 5 launch chart compares computer-use benchmark performance and cost across effort levels. Its cost axis uses the scheduled standard token rates, rather than the introductory rates available through 31 August; it is a vendor benchmark, not a marketing-workflow result. Source: Anthropic.

The Weekly Cut 009

AI attribution meets the brand-search question

The Weekly Cut 0098 min read

Two measurement developments ask what follows attention, while a model update promises better follow-through on work.

Read edition
The Optimizely Opal interface at the start of Conductor’s June10 AI Brand Visibility Report Agent demonstration, asking what the user wants to work on.
The opening screen of Conductor’s 10 June demonstration of its AI Brand Visibility Report Agent in Optimizely Opal. A supplier demonstration is the starting point; a paid trial should follow the work through to a usable result. Source: Conductor / Optimizely.

Analysis 002

What to ask an AI marketing supplier

Analysis 0025 min read

A paid trial on a real brief can reveal the work, costs and responsibilities behind a persuasive AI pitch. June’s spending evidence makes that scrutiny timely.

Read analysis
Cameron Adams and Bob Safian seated on a brightly lit Canva stage at Cannes 2026.
Canva co-founder Cameron Adams talks with Bob Safian at the Canva Cabana on 24 June 2026, during the Cannes week in which Canva announced Grow 2.0. Source: Canva.

The Weekly Cut 008

Canva connects creation to performance

The Weekly Cut 0088 min read

Creative tools move closer to the media buy, while new retail and conversational formats offer places to test the work.

Read edition