Welcome back to another “Essays” episode.

It’s only two weeks since we started our live experiment of automating VC with AI and we already have more than 110 members in our free Slack group + various tools and ideas being explored.

Last week, I created a ChatGPT workflow automation to access my inbox, filter newsletters, summarize them, and create a daily digest. Simple but powerful.

Next on my task list, I wanted to automate a flow where if someone tells me about a hot startup (or founder), I can quickly check via WhatsApp or Slack to see whether there’s an entry in our CRM, what the status is, what the company does based on all available internal and external data, and receive a brief assessment hereof.

Yes, this flow is indeed more complex than the email newsletter summarization. And unfortunately, it’s already too complex for ChatGPT. While ChatGPT recently added a few dozens of “apps” and is great for simple automations, it lacks connectors to most of my preferred apps and fails to translate more complex workflow descriptions into robust sequences.

In short, ChatGPT is super easy to use and has low/no technical requirements, but it quickly reaches its limits - at least today. So we need an alternative…

First off, let’s take one step back. On the highest level, there are two distinct groups of automation tools:

Due to the complexity and multi-app nature of our VC workflows, I’ve spent most of my time working with API-/MCP-based solutions. Starting in 2017, I mostly used Zapier as it was one of the early providers with sufficient liquidity of connectors. It’s relatively easy to use but lacks some customization and becomes VERY expensive, very fast.

In 2020, I met a Berlin-based founder called Jan Oberhauser who just launched a new project called n8n. BTW the name’s a mix of “node” and “automation”, that initially was “nodemation” which then got shortened to n8n. An amazing product where - unfortunately - I lost the lead for the Seed round against Sequoia. You win some, you lose some. In any case, I was immediately hooked by the flexibility and community angle, that not only drove connector liquidity but also the ability to share workflow templates and build together.

Digging deeper into the automation hole over the years, I found and tested more and more tools such as Make, Bardeen, Gumloop, or Lindy AI. To set the scene for the next experiments and following a discussion in our Slack group on “Which is the best?”, I’ll dedicate today’s post to an automation tool comparison focused on ease-of-use, technical requirements, connector liqudity, customization, and costs, among others.

At the end of this article, you’ll know the pros and cons, and where to start automating your workflows.

Sorted from easy to more difficult to use & from simple to more complex workflows..

This group represents a newer wave of AI-native, agent-leaning automation tools that sit somewhere between classic workflow builders and autonomous assistants. Their core strength is that they treat unstructured work as first-class: reading emails or Slack messages, understanding intent, extracting entities, making lightweight decisions, and taking follow-up actions across tools without requiring rigid schemas upfront.

This makes them particularly attractive for VC workflows like intro triage, founder research, note summarization, follow-ups, and early signal detection, areas where “good enough, fast” often beats perfect. The trade-off is control and determinism: once LLMs are embedded deeply into the execution path, outputs become probabilistic, harder to debug, and more difficult to audit or version.

As a result, these tools shine as augmentation layers (assistants that prepare, suggest, draft, or pre-structure work), but they are still risky as fully autonomous backbones for systems of record like CRMs or deal databases. My current mental model is that Lindy, Gumloop, Bardeen, or Relay are best used to compress human effort at the edges of workflows, while more deterministic tools (like n8n, Make or Zapier) remain the foundation.

Best for:

Strengths:

Weaknesses

Verdict: Easy to get started, great for high volume & low complexity workflows.

Zapier is the default choice for a reason: it has massive connector liquidity, a smooth onboarding experience, and it’s easy to get to value for standard SaaS workflows (“when X happens in Gmail → create Y in CRM → notify Slack”). For VC teams, it’s great for early-stage ops: routing inbound deal flow, syncing basic fields, pushing reminders, lightweight enrichment, and “glue” between common tools.

The pain points appear with sophistication and volume: multi-branch logic and data transformations can become awkward; workflows become hard to reason about when they sprawl; and pricing can spike fast as tasks scale. I see Zapier as ideal for simple, high-leverage automations, but if automation becomes core infrastructure (or you run high volume), you’ll either hit cost ceilings or logic ceilings.

Best for:

Strengths:

Weaknesses:

Verdict: Zapier is great until it isn’t. Many teams outgrow it the moment workflows become core infrastructure.

Make sits in a sweet spot between Zapier and n8n: more powerful and expressive than Zapier (especially for transformations and branching), but often easier to approach than n8n for non-engineers because the scenario builder is very visual and intuitive. For VC ops, it’s strong for multi-step processes—enrichment, routing, syncing, automation across CRM/Notion/Airtable/Slack—without immediately needing custom code.

The main downside is maintainability at scale: complex scenarios can become visually dense, and debugging/observability can get annoying once you have many moving parts. My view: Make is excellent for medium complexity automation where you want power without committing to the full “automation engineering” mindset.

Best for:

Strengths

Weaknesses

Verdict: Make is a strong middle ground. More power than Zapier, less overhead than n8n, reasonable costs.

n8n is the most compelling “automation OS” for power users because it combines a visual builder with real engineering primitives: strong branching, loops, error handling, custom code, reusable sub-workflows, and (optionally) self-hosting. For VC workflows, especially anything like “Slack → CRM lookup → enrichment → scoring → create/update records → notify”, n8n is where you can build something robust instead of a fragile chain of triggers.

The trade-off is that it requires you to think like a systems builder: no matter if you manually design workflows or rely on the AI text-to-workflow builder - you need to think it through, from handling edge cases, and maintaining workflows, but the result is dramatically more controllable and scalable. If your goal is a dependable pipeline that can evolve (and not just a quick demo), n8n is often the best long-term bet.

Best for:

Why I like it:

Trade-offs

Verdict: n8n is the closest thing to a workflow operating system for serious automation. If automation becomes a competitive advantage, this is where you end up.

Automation for VC is no longer about finding a single “best” tool, but about choosing the right abstraction layer for the job. ChatGPT and browser-based tools like PhantomBuster excel at fast, low-friction experiments and filling API gaps, but they break down once workflows become complex or mission-critical.

API-based platforms like n8n, Make, and Zapier offer far more robustness and control, with clear trade-offs between ease of use, flexibility, and cost. Meanwhile, a new generation of AI-native tools such as Lindy, Gumloop, Bardeen, and Relay is emerging to handle unstructured, ambiguous work, augmenting humans rather than replacing deterministic systems.

The key takeaway: durable leverage in VC automation comes from combining tools deliberately, using AI and browser automation at the edges, and system-level automation at the core.

For me personally, aiming to automate my VC job end-to-end, from simple to complex tasks, I’ll mostly use n8n for the foundation of this experiment. Yes, it’s probably the most technical and difficult to get started solution (it’s actually not that difficult…) but its new AI text-to-workflow builder, the breadth of guides and community resources, and the ability to share templatized workflows with you makes it my preferred choice.

Until next week, where I’ll share the step-by-step automation that allows me to check WhatsApp or Slack for any startup or founder, reads all internal information across GDrive, CRM, inbox, and other sources, summarizes them, and provides a company brief for a quick assessment.

Stay driven,
Andre