Review of Make’s New AI‑Powered Automation Features

Quick Verdict

Make has added five AI‑driven modules that let you generate text, create images, classify data, extract structured information, and run AI‑based routing inside a workflow. They work well for low‑to‑moderate volume tasks, integrate natively with existing Make scenarios, and cost roughly the same as the underlying model usage. For hobbyists or tight budgets, n8n’s open‑source AI nodes or Zapier’s limited AI actions are cheaper, but Make offers the most seamless UI for end‑to‑end automation.

New AI Modules in Make

Make introduced the following modules in its AI Automation suite (released Q2 2024):

Module	Underlying Service	Primary Use	Pricing Model
AI Text Generation	OpenAI GPT‑4o (fallback to GPT‑3.5)	Summarize, draft emails, generate code snippets	Pay‑per‑token (included in OpenAI usage)
AI Image Generation	Stable Diffusion via Hugging Face API	Create placeholders, marketing graphics	Pay‑per‑image (≈$0.02 per 512×512)
AI Classification & Sentiment	Google Vertex AI (AutoML Text)	Tag tickets, sentiment scoring	Pay‑per‑prediction (≈$0.001 per 1k tokens)
AI Data Extraction	Azure Form Recognizer + GPT‑4	Pull fields from invoices, PDFs, emails	Fixed per‑page + token cost
AI Decision Engine	Custom LLM prompt routing	Conditional branching based on free‑form input	Same as Text Generation

1. AI Text Generation

The module accepts a prompt, optional system message, and token limit. It returns raw text or JSON if you include a structured response schema. It supports streaming, which reduces latency for longer outputs. You can chain the result directly into downstream actions (e.g., send email, update Airtable).

2. AI Image Generation

Make calls the Hugging Face Stable Diffusion endpoint. You provide a textual prompt and optional seed. The module returns a base64‑encoded image that can be saved to Google Drive, S3, or sent via Discord. The UI includes a preview pane, which is handy for quick iteration.

3. AI Classification & Sentiment

You feed a piece of text and a list of labels; the module returns the most likely label with confidence. It also supports multi‑label classification. For sentiment, the preset model returns positive, neutral, or negative with a score.

4. AI Data Extraction

Make now bundles OCR (via Azure) with an LLM that parses the raw text into a JSON schema you define. This is useful for extracting line items from receipts or contract clauses without writing custom regex.

5. AI Decision Engine

Instead of hard‑coded routers, you write a prompt like “If the user mentions a refund, route to refunds, otherwise route to support.” The LLM returns a routing key that Make uses to select the next branch.

How to Set Up an AI‑Powered Scenario

1. Create a new scenario and add the desired AI module from the Apps list.
2. Authenticate – most modules require an API key from the underlying provider (OpenAI, Hugging Face, Google Cloud, Azure). Make stores keys securely.
3. Configure the prompt – use Make’s built‑in variables ({{variable}}) to inject dynamic data.
4. Define output handling – map the module’s text, image, or json fields to the next step.
5. Test – the UI offers a “Run once” button that shows the raw response and any parsing errors.
6. Set execution limits – to avoid runaway costs, add a Router that checks token usage or image count before proceeding.

Performance, Limits, and Costs

• Latency: Typical response times are 800 ms for text, 1.5 s for image, and 2 s for extraction. Network overhead adds ~200 ms.
• Rate limits: Make forwards the provider limits. For OpenAI GPT‑4o, the default is 60 rpm; you can request higher limits via the provider.
• Cost tracking: Make’s dashboard now shows per‑module usage in token or image counts, letting you monitor spend in real time.
• Free tier: Make’s free plan includes 1 000 AI module executions per month (shared across all AI modules). Beyond that you pay the provider’s fees plus a $0.01 per execution surcharge.

Comparison With Alternatives

Platform	AI Features	Free Tier	Notable Limits	Ease of Integration
Make	Text, Image, Classification, Extraction, Decision Engine	1 000 executions/mo	Provider‑level rate limits	UI‑driven, visual mapping
Zapier	OpenAI actions, AI‑generated content	100 tasks/mo (incl. AI)	20 calls/min for OpenAI	Simple but fewer AI modules
n8n (self‑hosted)	Community nodes for OpenAI, Hugging Face, Azure OCR	Free, self‑hosted	Limited by your own compute	Requires setup, fully customizable
Pipedream	AI actions via code steps	10 000 events/mo	Pay‑as‑you‑go for providers	Code‑first, good for devs

If you need a zero‑cost solution and are comfortable with a self‑hosted environment, n8n with the community OpenAI node is the cheapest. For the most plug‑and‑play experience with a modest free allowance, Make remains the best choice.

When to Use Make’s AI Modules

• Rapid prototyping: Build a prototype that drafts marketing copy or generates product images without writing code.
• Customer support triage: Classify incoming tickets and auto‑assign them based on sentiment.
• Invoice processing: Extract line items from PDFs and push them into QuickBooks.
• Dynamic routing: Use natural‑language intent detection to decide which downstream workflow to trigger.
• Content generation pipelines: Combine text generation with image creation for automated blog post drafts.

Pros and Cons

Pros	Cons
UI‑first workflow builder; no code needed for most use cases.	Dependent on external provider limits; costs can add up with high volume.
All AI modules live in the same scenario, reducing context switching.	Free tier is modest; heavy users must monitor spend.
Built‑in usage dashboards simplify budgeting.	Some advanced model parameters (e.g., temperature, top‑p) are hidden behind “advanced settings” that may be confusing.
Supports multiple providers, giving flexibility.	Image generation limited to 512×512; larger sizes require external services.

Bottom Line

Make’s AI Automation suite brings the most common LLM‑driven tasks into a visual workflow environment. For solo founders who need quick, low‑volume AI assistance without managing servers, it’s a solid bet. Keep an eye on token usage and consider a self‑hosted n8n instance if you anticipate scaling beyond a few thousand calls per month.

On pricing: plans and rates change, and this post was written on a specific date. Check the vendor's current pricing page before you commit to anything.