AI Automation DB

What to look for · Chapter 06

AI automation platforms and integration

10 min read

Share

The right tools decide how well an automation fits your business. You don't need to be technical here. But you should know what to ask. This chapter explains the platforms, models and integrations agencies use.

The advance of technology is based on making it fit in so that you don't really even notice it.
Bill GatesBill GatesCo-founder, Microsoft

The AI automation stack: three layers

An automation build sits on three layers. Each does a different job.

Your appsCRM, ERP, email, sheetsAutomation platformn8n, Make, ZapierAI modelOpenAI, Claude, Gemini
An AI automation build stacks three layers.

Your apps hold the data. A platform connects them and runs the flow. An AI model handles language and reasoning. Complex jobs may add custom code on top.

n8n vs Make vs Zapier

These three cover most no-code automation. Here's how they compare.

PlatformStrengthWatch outData control
ZapierSimplest, most appsCostly at scaleCloud only
MakeVisual logic, good valueSteeper learning curveCloud only
n8nOpen-source, flexibleNeeds more setupCan self-host

Pick Zapier

For speed and the widest app support.

Pick Make

For richer logic at a lower price.

Pick n8n

When you need to control your own data.

Which AI models agencies use

Most builds use one of three model families. Each has a sweet spot.

ProviderKnown forGood for
OpenAI (GPT)Broad and capableGeneral tasks, agents
Anthropic (Claude)Long context, carefulDocuments, analysis
Google (Gemini)Multimodal, integratedWorkspace, media

Agencies pick by task, cost and your data rules. Ask which model they'd use for your workflow, and why.

AI automation platforms by category

Different jobs use different platforms. These are the common ones by category.

CategoryCommon platforms
Workflow automationn8n, Make, Zapier, Power Automate
Voice AIRetell, Vapi, ElevenLabs, Synthflow
ChatbotsVoiceflow, Botpress, custom builds
AI modelsOpenAI, Anthropic, Google

How agencies integrate with your stack

Automation is only useful if it connects to your tools. Agencies link up through APIs, webhooks and native connectors. That lets the automation read and write to your CRM and ERP.

Your systemsCRM, ERP, emailAutomation platformn8n, Make, ZapierAI modelOpenAI, Claude, Geminiresults flow back to your systems
Data flows out to the AI, then back to your tools.

Questions to ask about integration

  • Do you work with the tools we already use?
  • How will you connect to our CRM and ERP?
  • Does everything run on our accounts?
  • What happens if an app has no API?

Where it runs and who owns it

Ask where the automation runs and who holds the keys. It should run on your accounts, not the agency's. That keeps you in control of data, billing and access.

Custom code or no-code?

No-code ships fast and handles many workflows. Custom code suits complex logic, heavy volume or performance needs. Many builds mix both. Let the problem decide, not a preference.

Common mistakes with platforms

Key takeaways

  • An AI automation stack has three layers.
  • n8n, Make and Zapier cover most no-code work.
  • Insist the automation runs on your own accounts.
  • Let the problem pick custom code or no-code.

Need a specific platform?

Filter agencies by the platforms and tools they work with.

Browse the directory

Sources & further reading

Frequently asked questions

What platforms do AI automation agencies use?+

Most use no-code platforms like n8n, Make and Zapier to connect apps. For AI, they use models from OpenAI, Anthropic or Google. Complex or high-volume work often needs custom code. A good agency picks the tool for the job, not one favourite.

What's the difference between n8n, Make and Zapier?+

Zapier is the simplest and most polished, but can get costly at scale. Make offers more logic and visual flows for a lower price. n8n is open-source and self-hostable, which helps with data control. All three connect apps without heavy code.

Which AI model is best for automation?+

There's no single best model. OpenAI's GPT is broad and capable. Anthropic's Claude handles long documents and careful analysis. Google's Gemini fits multimodal and Workspace tasks. Agencies pick by task, cost and your data rules. Ask which they'd use, and why.

How do AI automation agencies integrate with my systems?+

They connect through APIs, webhooks and native app connectors. That lets automation read and write to your CRM, ERP and other tools. Confirm your key systems have APIs or exports. If a tool is closed, integration gets harder and slower.

Should the automation run on my accounts or the agency's?+

On your accounts, wherever possible. That keeps you in control of data, billing and access. If it runs only on the agency's platform, leaving later is harder. Ask where it runs and who holds the keys before you start.

Do I need custom code or is no-code enough?+

No-code handles many workflows well and ships fast. Custom code suits complex logic, heavy volume or tight performance needs. Many builds mix both — no-code for the flow, code for the hard parts. A good agency uses whatever fits.

What platforms do voice AI agencies use?+

Voice AI often runs on Retell, Vapi, ElevenLabs or Synthflow. These handle speech, phone calls and real-time responses. They pair with an AI model for the conversation. If you need voice, ask which platform an agency knows best.