Every week another AI tool launches, and every week the same shortlist keeps showing up in real small-business stacks. This isn't a comprehensive directory — it's the 10 tools I actually reach for when I build automations for small businesses, grouped by the job they do. Pick the two or three that map to your bottleneck and ignore the rest.
The stack, in one sentence
You need a place to store data, a way to move data between apps, an AI model that reads and writes text, and a way to talk to customers. Everything below fits into one of those four buckets.
1. n8n — the automation backbone
n8n is an open-source workflow builder that connects almost anything to almost anything else. Unlike Zapier or Make, you can self-host it for free, which matters when you're moving customer data around and don't want a per-task bill. It's the tool I use as the glue between every other item on this list — CRM to email, form to Slack, invoice to spreadsheet.
Best for: any small business past the 'a few forms and a spreadsheet' stage.
2. OpenAI (GPT-5, GPT-4o) — the default LLM
OpenAI's API is still the safest default for generation, summarization, and classification. GPT-4o is fast and cheap enough for real-time customer replies; GPT-5 is where you go when quality matters more than latency. Pair it with a good system prompt and it handles 80% of the 'AI feature' requests small businesses ask for.
Best for: chat replies, email drafting, document summarization, classifying incoming leads.
3. Claude (Anthropic) — the long-context specialist
Claude 3.5 and 4 shine when the input is long — a 30-page contract, a full call transcript, a year of support tickets. It's also the model I trust most for tone-sensitive customer writing. Use it as a second option alongside OpenAI so you're not locked into one provider.
Best for: contract review, meeting-notes summarization, on-brand customer replies.
4. Supabase — the database that ships in a weekend
Supabase gives you a Postgres database, auth, file storage, and edge functions behind one dashboard. For a small business, that means you can stand up a real backend — customer accounts, order history, admin panel — in days, not months. Row-level security is the killer feature: your data stays yours even when the AI code around it is messy.
Best for: any AI tool that needs to remember users, orders, or history.
5. Groq — fast, cheap inference
Groq runs open-source models (Llama 3.3, Mixtral) at speeds that make real-time chat feel instant, and their free tier is generous enough to run a live customer chatbot on. Use it as the primary model when latency matters and OpenAI/Claude as the fallback when quality does.
Best for: on-site chatbots, real-time suggestions, high-volume classification.
6. Cal.com — the booking layer
Not strictly AI, but every AI tool that ends in 'book a call' needs it. Cal.com is open-source, self-hostable, and its API means an AI chatbot can offer real slots and confirm bookings without a human in the loop. This is the last-mile piece most AI chat tools skip.
Best for: any workflow where the AI hands off to a human meeting.
7. Resend — transactional email that doesn't hate developers
AI generates the email, Resend delivers it. Clean API, React email templates, and deliverability that matches SendGrid at a fraction of the setup pain. When an AI workflow sends an invoice, a summary, or a lead notification, this is the pipe it goes through.
Best for: lead alerts, AI-drafted follow-ups, transactional receipts.
8. Whisper (via OpenAI or self-hosted) — voice to text
Speech-to-text is the unlock for any small business that runs sales or discovery calls. Feed the Whisper transcript into GPT-4o and you get a summary, action items, and a CRM-ready note in under a minute. Cheap, accurate, and works in dozens of languages.
Best for: post-call summaries, meeting notes, voice-note capture.
9. Perplexity API — grounded answers with sources
When you need the AI to answer from live web data — pricing pages, competitor updates, news — Perplexity is the shortcut. It handles the retrieval and citation step that would otherwise take a full RAG pipeline. Useful for internal research assistants and market-watch tools.
Best for: research bots, competitive monitoring, up-to-date Q&A.
10. Cursor / Windsurf — build the AI tool, faster
This is the meta pick. Every tool above is easier to wire up when your IDE has an AI pair working with you. If you have any technical person on the team, an AI-first editor cuts build time by half. It's not a customer-facing tool, but it decides how many of these you can actually ship.
Best for: the person actually building the automations.
How to pick two, not ten
You don't need all ten. Start with the bottleneck:
- Drowning in repetitive admin? n8n + OpenAI.
- Losing leads at first contact? Groq chatbot + Cal.com + Resend.
- Sales calls not turning into follow-ups? Whisper + GPT-4o + your CRM via n8n.
- Need customer accounts and history? Supabase first, then layer AI on top.
The honest caveat
AI tools change monthly. Prices drop, models get replaced, and today's default becomes tomorrow's fallback. The stack above is what I'd deploy this quarter for a small business — the specific model names will shift, but the categories (glue, LLM, database, delivery) stay the same. Build for that shape, not for the current version numbers.
The best AI stack for a small business is the smallest one that removes the bottleneck this month.