The AI Tools We Keep Coming Back To

Hiya 👋

Every week I try to make sense of all the new AI announcements, and lately it feels like they’re coming faster.

Most of it is noise and shiny demos that don’t really provide value. But some of it actually changes how we work, right now.

This week had a few of those moments: Gumloop making automation dead simple, Replit showing vibe coding isn’t just a toy anymore, and Notion turning into a full-on AI coworker.

Instead of focusing on “state of the AI union”, I’m using this week to break down with my take on what’s real, what’s hype, and what it means for how teams get stuff done.

Let’s jump in👇

The news: Gumloop announced a natural language workflow builder: just describe the automation you want, and it builds it for you.

Why it matters:

• No learning curve. Their Slack demo shows how a basic Q&A bot can be live in minutes, no technical background needed.

• Real adoption. Shopify rolled Gumloop out company-wide after just weeks of testing. Now 110+ teams have built 6,000 workflows, saving thousands of hours.

• Proof of scale. These aren’t just side experiments, Gumloop is handling millions of automation actions with zero security incidents.

The big picture:
This marks a shift: automation is no longer gated by engineering talent. Anyone at a company can start experimenting. That’s huge for adoption, but it also creates a fork in the road:

• Simple, local wins: Perfect for one team automating repetitive tasks (think Slack bots, notifications, or SEO tweaks).

• Complex, strategic workflows: Cross-department processes (e.g., customer support tied into CRM + data warehouse + RevOps handoffs) demand thoughtful design, compliance, and integration strategy.

Our take: Gumloop shows how accessible AI automation is becoming. But scaling it safely and strategically across departments is where most businesses hit a wall. That’s the gap Switchboard fills, we help teams turn these cool demos into enterprise-ready automation.

Ready to go from “playground” to real business impact? Let’s talk →

Replit’s Agent 3 marks a real step forward in “vibe coding”, giving natural language prompts and watching an app, bot, or automation come to life.

• It runs for up to 200 minutes autonomously, with live monitoring.

• It uses a self-testing loop to catch and fix its own mistakes, up to 3x faster and 10x cheaper than comparable models.

• For the first time, it can even build other agents, which slot into tools like Slack or email.

What’s happening in the wild:

• A developer spun up a payroll ticker app from a single prompt in under 90 minutes.

• An Upwork job description was transformed into a working salon check-in system, customer flows, database, and dashboard included.

• Builders are using it to make Telegram bots and automations that test themselves as they run.

Why it matters:
Agent 3 shows how quickly vibe coding is maturing. It’s not yet reliable for enterprise-scale systems, but for prototypes, early proofs of concept, and internal tools, it’s already viable. This means the cycle from idea → working demo is collapsing from weeks to hours.

🧵 See the full breakdown of Agent 3 in Adam Silverman’s thread →

Notion announced its biggest update yet: Notion 3.0, with AI Agents built directly into the platform.

What’s new:

• Do the work, not just suggest. Agents can now create docs, build databases, search across tools, and complete multi-step workflows, not just edit text.

• Personalization. You can give your Agent custom instructions, memory, even a name and personality, so it learns your working style.

• Custom Agents (coming soon). Imagine a whole team of AI specialists: one Agent compiling customer feedback daily, another posting weekly project updates, another triaging IT tickets.

• Permissions & integrations. Database row permissions and new connectors make it more enterprise-ready.

Why it matters:
Notion is moving from being a workspace where you do the work to a workspace where the work gets done for you. Agents can now run up to 20 minutes of autonomous work across hundreds of pages, things like synthesizing customer feedback from Slack, email, and Notion into a database.

The big picture:
This is Notion’s attempt to become the AI operating system for knowledge work: docs, databases, and now agents that take action. It could redefine how teams think about productivity platforms, less as static canvases, more as autonomous collaborators.

Learn more about Notion 3.0 here →

Know someone who’d enjoy this newsletter?
Forward it along, or send us a note if there’s a topic you’d like us to unpack next.