AI Projects - October 23, 2026

Week of: October 23, 2026

Overview

This week’s AI Projects coverage focused on practical deployment: a recruiting firm using Claude with Zapier MCP, a three-person hiring startup building its go-to-market engine, and guides for building agents and improving brand visibility in AI answers. Vertical workflow examples and agent-cloud infrastructure economics also surfaced.

Stories

1. Recruiting agency uses Claude and Zapier MCP to find candidates who aren’t looking

Source: Zapier Link: https://zapier.com/blog/blue-castle-agency-zapier-mcp

Zapier’s case study profiles Blue Castle Agency, a recruitment firm for real estate and property management founded by Curtis Key, and describes how it uses Claude with Zapier MCP to find candidates who are not actively looking. The article notes Key’s roughly 20-year background building automation, first with code and scripts and later in Salesforce.

The example is a concrete workflow pattern for AI builders: pairing an LLM with MCP-connected business tools to support a high-touch professional service. It shows automation applied to sourcing and research rather than only content generation.

Impact Analysis: For agentic workflow builders, the useful pattern is connecting models to existing operational systems through MCP rather than replacing the recruiter’s process.

2. Three-person hiring startup Vetano built its go-to-market engine before the rush

Source: Zapier Link: https://zapier.com/blog/vetano-next-gen-zaps

Zapier profiles Vetano, a hiring platform founded by Chris Fairley in early 2025, which aims to replace the traditional resume with ID-verified skills and discusses candidates showing their skills on video instead. The article focuses on how the three-person startup built its go-to-market engine before the rush.

For founders, it is an example of a very small team using automation to run go-to-market functions that would otherwise require more headcount. The hiring context also highlights pressure to differentiate candidates as AI screening becomes more common.

Impact Analysis: Lean startups can use automation platforms to stand up go-to-market operations quickly, though the case is still early-stage.

3. HubSpot publishes a step-by-step guide to building and deploying an AI agent

Source: HubSpot Link: https://blog.hubspot.com/marketing/build-ai-agent

HubSpot’s guide describes building an AI agent as a multi-step process: define the agent’s task, connect approved data and tools, and set limits on its actions. It frames agent-building as accessible to marketers, sales reps, and customer teams once the basics are understood.

That emphasis on scoped data access and action limits is useful for teams moving from demos to production. It also reflects the shift toward non-engineering teams owning narrow automation projects.

Impact Analysis: Start with a narrowly defined task and explicit permissions before expanding an agent’s autonomy.

4. HubSpot compares Profound and Scrunch for answer-engine optimization

Source: HubSpot Link: https://blog.hubspot.com/marketing/profound-vs-scrunch

HubSpot compares Profound and Scrunch, two tools that measure how AI answers cite a brand. It says Profound leans toward market analysis and content creation, while Scrunch leans toward technical optimization and task handoff.

As AI answer engines become a discovery surface, marketing teams need workflows for tracking citations and fixing gaps. The comparison is framed around where an AEO workflow stalls rather than which tool tracks more engines.

Impact Analysis: Choose AEO tooling based on the team’s bottleneck—analysis and content or technical implementation.

5. Moshi’s Anekha Sokhal on fitting AI tools into a trader’s workflow

Source: Latent Space Link: https://latent.space/p/moshi/

Latent Space interviews Anekha Sokhal of Moshi about how the company turns market data into actionable insights. The discussion centers on why AI tools have to fit the trader’s existing workflow.

For applied-AI builders, the takeaway is that vertical tools succeed when they adapt to domain processes rather than asking users to change how they work. Trading is a high-stakes example where trust and workflow fit matter.

Impact Analysis: Domain-specific AI products should be designed around existing professional workflows, not just model capabilities.

6. Modal CEO Erik Bernhardsson discusses elastic GPUs and the agent cloud

Source: Latent Space Link: https://latent.space/p/modal-d/

Latent Space talks with Modal CEO Erik Bernhardsson about elastic GPUs, sandbox growth, and the economics of the agent cloud. The episode is framed around why AI compute comes in rectangles.

For teams building agents and AI products, infrastructure choices around elasticity and sandboxing affect cost and reliability. The discussion connects those operational concerns to the business model of agent cloud platforms.

Impact Analysis: Agent builders should treat elastic compute and sandbox economics as part of product design, not just backend plumbing.

Source Links

• Zapier - Recruiting agency uses Claude and Zapier MCP to find candidates who aren’t looking
• Zapier - Three-person hiring startup Vetano built its go-to-market engine before the rush
• HubSpot - HubSpot publishes a step-by-step guide to building and deploying an AI agent
• HubSpot - HubSpot compares Profound and Scrunch for answer-engine optimization
• Latent Space - Moshi’s Anekha Sokhal on fitting AI tools into a trader’s workflow
• Latent Space - Modal CEO Erik Bernhardsson discusses elastic GPUs and the agent cloud