Guide
AI Tools Stack That Actually Works
The tools I recommend and the ones I tell clients to skip. No affiliate links.
There are thousands of AI tools and most of them are not worth your time or your money. This is the actual stack I use across client work: what each tool does, what it costs, and when I recommend skipping it entirely. No affiliate links. No sponsorships. Just what I actually run.
What's inside
- Claude, the primary AI layer for reasoning, drafting, and system logic
- n8n or Make, the automation backbone and no-code workflow builder
- Voiceflow, the chatbot builder for inbound handling and lead capture
- Airtable, the structured data layer for client CRM and operations tracking
- Loom, for client delivery and async communication
- Notion, for internal documentation and SOPs
- What I skip and why: tools that over-promise for most small and medium businesses
Why it works
The goal is not to have the most tools, it is to have the minimum stack that handles 80% of your operational bottlenecks. Every tool I removed from this list made the operations cleaner and the results more predictable. Most businesses need 3 to 4 tools wired together well, not 20 tools used poorly.
Full Resource
The rule before the stack
- The goal is not to have more tools. The goal is fewer tools, used correctly.
- Most businesses need 3 to 5 tools wired together well, not 15 tools used poorly.
- Every tool on this list was chosen because it handles a specific job better than the alternatives at its price point.
- What I skip is on this list too. Skipping the wrong tool saves more money than adding the right one.
Claude (Anthropic)
Primary AI layer -- reasoning, drafting, system logic
- What I use it for: writing client-facing content, building and refining prompts, auditing workflows, structured analysis tasks, long-document processing.
- Recommended tier: Pro at $20/month. The context window on Pro handles long documents and complex multi-step tasks. Free tier is too limited for daily operational work.
- When to use Claude over ChatGPT: complex reasoning, longer documents, anything where accuracy matters more than speed or creativity.
- Claude Code for developers: adds browser, web search, file access, and custom tool connections. Changes what is possible with a single AI session.
n8n vs Make.com
Automation backbone -- pick one and stay with it
- Choose n8n if: you want to self-host, you have a developer or are comfortable with JSON, you need complex branching logic, you want full control over your data.
- Choose Make.com if: you want something visual and fast to configure, you are non-technical, you need basic automations running within an hour.
- For most small service businesses: Make.com free tier handles 1,000 operations per month. That covers most follow-up sequences, form-to-CRM connections, and notification systems.
- Cost at scale: n8n is cheaper if self-hosted. Make.com charges per operation at higher volumes.
Voiceflow
Chatbot builder for inbound handling and lead capture
- What it does well: conversation design, FAQ handling, lead capture with variable storage, human handoff logic.
- What it does not do: it is not a CRM. Do not store client data in Voiceflow long-term.
- Recommended tier: Free tier for one agent and basic flows. Upgrade when you need analytics, multiple agents, or advanced integrations.
- Why not a simpler tool: most drag-and-drop chatbot builders (ManyChat, Tidio) have limited conversation logic. Voiceflow handles complex branching that those tools cannot.
Airtable
Structured data layer and lightweight CRM
- What I use it for: tracking leads and their status, client project milestones, linking data between automations, lightweight CRM for under 500 active records.
- Recommended tier: Free tier is enough to start. Upgrade when you exceed 1,000 records or need automation triggers and views.
- When to use something else: if you have 5,000 or more records or need a proper sales pipeline with built-in email, use HubSpot free tier instead.
- The key habit: Airtable works only if data goes in consistently. Wire your intake forms and bots to populate it automatically. Manual data entry breaks it.
Loom
Client delivery and async communication
- What it replaces: long email explanations that take 20 minutes to write and 2 minutes to watch.
- How I use it: walkthroughs of work delivered, onboarding explanations, audit findings, quick responses to complex questions.
- Recommended tier: Free tier covers most use cases at small scale.
- Rule of thumb: if explaining something would take more than 3 sentences in an email, record a Loom instead.
Notion
Internal documentation and SOPs
- What I use it for: SOPs, internal playbooks, onboarding docs, client-facing reference documents, project wikis.
- When to use it: when information needs to be found and updated by multiple people over time.
- When not to use it: for live data or anything that changes daily. That belongs in Airtable or your CRM.
- Recommended tier: Free tier covers most small team needs.
What I skip and why
- Zapier: Make.com does 95% of the same things at a lower cost per operation. Not worth the price difference at small scale.
- Monday.com and Asana used as CRMs: project management tools make poor CRMs. Data gets messy, nothing integrates cleanly. Use a real CRM or Airtable.
- Generic AI chatbot builders (ManyChat, Tidio, Intercom chatbot): limited conversation logic. Good for basic FAQs. Insufficient for lead capture or complex routing.
- AI writing tools like Jasper and Copy.ai: Claude and ChatGPT do the same job. No reason to pay for a wrapper when you can use the original models directly.
- HubSpot paid tiers for small businesses: the free tier is excellent. The paid tiers are priced for enterprise. Most small businesses do not need features that justify the jump.
All resources are free. No sign-up required. Back to all resources →