Growth Lab Florianópolis: 7 lessons on AI and pipeline 2027 from a founder's notebook

On November 12 I attended Growth Lab 3rd edition at Zoho Brasil's office in Florianópolis, Brazil. The theme was "Pipeline 2027: where AI accelerates and people decide". I came back with seven ideas we are already applying at the agency: AI reads your site but rarely sends clicks, a page that answers well matters more than markup, whoever writes should not approve, Claude can tie a team's tools together, an ad creative can drop from 90 to 15 minutes with a person signing off, the best ad words come from customers, and before raising the budget you need to know where the bottleneck is.

I write this as co-founder of We make it Lab, in charge of operations and paid media for ecommerce brands. So every idea ends with the same question: what changes for an online store that spends on ads.

1. AI reads your site, but rarely sends the click

Michell Lech, CMO at Piperz and founder of Alpaca Labs, shared the case of a blog built with AI. One number stuck with me: in 24 hours (September 21 to 22), Anthropic's bots read the site 481 times, OpenAI's 273, Moonshot's 262, Perplexity's 176 and Googlebot 104. AI reads far more than most of us assume.

The other half of the story is the uncomfortable one: those reads almost never turn into visits. The proof that an AI had recommended the brand did not show up in an analytics dashboard. It came up in a conversation with a customer.

What to do about it in a store: add "how did you hear about us?" at checkout or in the first WhatsApp message, with "ChatGPT or another AI" as an option. Waiting to see it in GA4 will not work.

2. What made an AI cite the brand

In the same case, 77% of appearances in Google's AI answers and 98% of Copilot citations (342 of 350) came from blog URLs. A single page, an image library that was already indexed, earned 27% of appearances and 65% of citations.

The most useful part was how they rated each action:

For a store: each product page should answer on its own what a buyer asks (price, shipping, payment options, sizes, returns). That is the page an AI can cite. Schema helps, but it does not replace a good answer. We cover the steps in cómo aparecer en ChatGPT y las respuestas de IA and in our guía de SEO para ecommerce en Tiendanube.

3. Whoever writes does not approve

Piperz produces content with a team of AI agents, each with one role: research, planning, writing, review and release. All of them read the rules from a single document, a kind of editorial constitution. The principles they kept: the writer never approves, a number only goes in with a source and a date range, and the agent proposes while the decision to publish follows a rule and a review.

We work in a similar way. Before a deliverable reaches our account manager, it goes through an automated quality check (built with Claude skills) for brief compliance, spelling, consistency and signs of AI-written text. Then a person reviews it. Keeping the maker and the checker apart is what stops mistakes from reaching the client.

4. Claude as the glue between systems

Rodrigo Chernhak, partner at Agência Tab, showed how they use Claude as the base that connects Asana, Drive, Gmail, Figma, WordPress, their CRM and their own prospecting and proposal tools. "Everything goes through here", the slide said.

We got to the same place from paid media. Claude is connected to our Meta Ads, Google Ads, GA4, CRM, Canva, ClickUp and Slack through MCP (see how to conectar Claude a tu CRM o ERP and how to implementar Claude en una empresa). Weekly reports come together from connected data, and the team's time goes into reading the numbers and deciding.

5. From 90 to 15 minutes per creative, with a person at the end

Tab cut production time per piece from 90 to 15 minutes. Besides AI, they rely on a design system in Figma with colors, type and components, and every piece is finished by a designer.

In ecommerce we add one more rule: the product is always the real one. AI can change the background, the scene or the format, but never regenerate the product. That is why we push for real product photography. An ad that shows something different from what arrives at the customer's door costs you returns and bad reviews.

6. Customer words beat ours

Chris Cornehl, CEO of MeetRox, analyzed thousands of sales conversations. For marketing, his advice was direct: use the words that trigger customers, taken from real conversations (the same input you need to build useful armar buyer personas que sirvan). He also reminded us that first impressions start before the meeting.

For a store: the best ad copy lives in chat messages, reviews and questions that come in through Instagram. If customers say "it arrived in two days" or "it looks exactly like the photo", that line will outperform any slogan.

7. Before spending more, find the bottleneck

Tháfila Araujo, co-founder of Resultados Contínuos, asked each type of attendee one question. The founders' question hit home: do you know exactly when the bottleneck is marketing budget and when it is strategy?

It is the question we hear most from store owners. Often the answer is not more budget. It can be margin, average order value, the product page or tracking that measures the wrong thing. That is why we judge paid media by profitability (contribution margin, MER and break-even ROAS), and not only by what Meta reports.

Rodrigo closed his talk with a 1930 prediction: Keynes expected that by 2030 we would work 15 hours a week. The average week dropped from 45 to 48 hours to about 38, and the time saved turned into more work. AI is giving us time back again. What we do with it is our call.

What we changed at We make it Lab

• We will add "how did you hear about us?" with an AI option to our account review forms, to measure what GA4 cannot see.
• Every blog article, in Spanish, Portuguese and English, will answer one clear question, as the Piperz case showed.
• We will bring customers' own words, taken from real conversations, into our pages and guides.

Conclusion

Growth Lab confirmed what we see every day in the accounts we manage: AI speeds up production and analysis, but the decisions that move money are still made by people. The brands that win in 2027 will use the time AI gives back to think harder about their numbers.

Thanks to Alpaca Labs, Agência Tab, Zoho and Resultados Contínuos for organizing it, and to everyone who shared real numbers on stage.

If you run an online store and want to know whether your bottleneck is budget or strategy, book a 30-minute account review. In the meantime, see our casos de clientes.

Sources

• Growth Lab 3rd edition, event page on Sympla (accessed November 12, 2026).
• Talks by Michell Lech (Piperz / Alpaca Labs), Rodrigo Chernhak (Agência Tab), Tháfila Araujo (Resultados Contínuos) and Chris Cornehl (MeetRox) at Growth Lab, November 12, 2026. Figures taken from the slides.
• Economic Possibilities for our Grandchildren, John Maynard Keynes, 1930.
• AI features and your website, Google Search Central (accessed November 13, 2026).