Everyone talks about AI tools in isolation. "Use this for writing." "Use that for video." "Try this for meetings." But nobody talks about how to make them all work together as a cohesive system.
I've spent the past year building what I call my "AI assistant stack" — a collection of AI tools that work together to handle about 60% of the repetitive work in my day. This isn't theoretical. This is my actual, daily-use setup as a product manager and content creator.
Here's the full breakdown.
The Philosophy: Layers, Not Tools
Before I list the tools, let me explain the thinking behind the stack. I organize my AI tools into four layers:
- Capture Layer: Tools that collect information (meetings, voice notes, ideas)
- Processing Layer: Tools that transform raw input into usable output (transcription, summarization, formatting)
- Creation Layer: Tools that produce finished content (videos, audio, written content)
- Distribution Layer: Tools that get content where it needs to go (scheduling, publishing, sharing)
Each tool in my stack serves one or two layers. The magic is in how they connect.
Layer 1: Capture — Never Lose an Idea Again
Fireflies.ai — The Meeting Brain
Role in stack: Captures everything said in meetings
Cost: $18/month (Pro plan)
Daily use: 3-5 meetings/day
Fireflies.ai is the foundation of my capture layer. It joins every meeting automatically and records everything — not just the words, but the context.
What makes it essential to the stack:
- Automatic joining: I never have to remember to start recording. If it's on my calendar with a meeting link, Fireflies is there.
- Speaker identification: It knows who said what, which is critical when I'm reviewing notes from a meeting with 6+ people.
- Topic detection: Fireflies automatically segments conversations by topic, so I can jump directly to "the part where we discussed the Q2 roadmap" without scrubbing through an hour of audio.
- API access: This is the key for stack integration. Fireflies' API lets me automatically push meeting data to other tools in my workflow.
My Fireflies workflow:
- Meeting happens → Fireflies records and transcribes
- Within 5 minutes → AI summary with action items lands in my Slack
- Within 15 minutes → Full transcript is searchable in Fireflies dashboard
- Weekly → I review all action items across meetings for my planning session
Stack integration: Fireflies feeds into my processing layer via webhooks. Meeting summaries automatically go to Notion, action items go to Linear, and customer feedback goes to a dedicated Airtable base.
Typeless — The Voice-First Notepad
Role in stack: Captures thoughts, ideas, and quick documentation via voice
Cost: $12/month
Daily use: 15-20 minutes of dictation
Typeless fills the gap between meetings. All those moments when you have an idea in the shower, a thought while walking to lunch, or need to quickly document something — Typeless captures them.
Why it's in my stack (and not just Apple's built-in dictation):
- Context awareness: Typeless understands what I'm talking about. If I say "add to the sprint retro notes," it knows I'm referring to a specific document and formats accordingly.
- Formatting intelligence: It doesn't just transcribe words — it creates structured text. Bullet points, headers, paragraphs — all from natural speech.
- Speed: I speak at ~150 WPM and type at ~70 WPM. That's a 2x productivity gain for any text-heavy task.
My Typeless workflow:
- Morning (5 min): Dictate daily priorities and any overnight thoughts
- Between meetings (2-3 min each): Capture personal takeaways that Fireflies might miss (political dynamics, gut feelings, ideas sparked by the conversation)
- Evening (5 min): Quick review and brain dump of the day
Stack integration: Typeless output goes to Notion via a simple copy-paste workflow. I'm working on automating this with their API, but honestly, the manual step takes 30 seconds and gives me a chance to review.
How Capture Layer Tools Work Together
Here's the key insight: Fireflies captures what others say, Typeless captures what I think. Together, they create a complete record of both the external conversations and my internal processing.
Example from last week:
- Fireflies captured: Client meeting where they mentioned struggling with onboarding new team members
- Typeless captured (my post-meeting dictation): "Client's onboarding pain is exactly what our new template feature solves. Draft a personalized demo for them focusing on the team onboarding workflow. Also, this is a common pain point — could be a good blog post topic."
Neither tool alone gives me the full picture. Together, they're comprehensive.
Layer 2: Processing — Making Sense of the Noise
The processing layer is where raw captures become actionable information. This is mostly handled by the AI features built into my capture tools, plus some automation glue:
Fireflies AI Summaries → Structured meeting notes with action items
Typeless Formatting → Clean, organized text from voice input
Notion AI → Synthesizes information across multiple sources
Zapier → Connects everything and routes information to the right place
My Processing Automations
Here are the actual Zaps running in my workflow:
- Meeting → Tasks: When Fireflies identifies an action item assigned to me → Create a Linear ticket
- Meeting → CRM: When a sales call ends → Update the deal record in HubSpot with meeting summary
- Meeting → Content: When a customer mentions a pain point → Add to my "Content Ideas" Airtable base
- Daily Digest: Every evening at 6 PM → Compile all meeting summaries and action items into a single Notion page
Layer 3: Creation — Producing Content at Scale
This is where the stack really shines. With all my ideas, meeting insights, and documentation captured and processed, creating content becomes almost effortless.
ElevenLabs — The Voice Layer
Role in stack: Generates professional voiceovers and audio content
Cost: $22/month (Pro plan)
Weekly use: 3-5 voiceovers, 1 podcast episode
ElevenLabs is my go-to for anything audio. I cloned my own voice, and now I can generate voiceovers that sound exactly like me — without actually recording anything.
How it fits the stack:
- Video voiceovers: Every HeyGen video gets an ElevenLabs voiceover for maximum quality
- Podcast production: I script episodes based on meeting insights (from Fireflies) and generate consistent intros/outros
- Audio versions of blog posts: Every article I publish gets an audio version. This takes about 2 minutes per article and significantly increases engagement.
- Multi-language content: My cloned voice speaking Spanish, German, and French for international marketing
Stack integration: I use ElevenLabs' API to automatically generate audio versions of new blog posts. When a post is published in Notion (my CMS), a Zapier workflow sends the text to ElevenLabs and attaches the audio file to the post.
HeyGen — The Video Factory
Role in stack: Produces professional videos without cameras or studios
Cost: $48/month (Creator plan)
Weekly use: 3-4 videos
HeyGen transformed me from "person who avoids video" to "person who publishes 3-4 videos per week." The AI avatars are convincing, the editor is intuitive, and the output quality is genuinely professional.
How it fits the stack:
- Product updates: Weekly video announcing new features (scripted from sprint notes captured via Typeless)
- Tutorial videos: Step-by-step guides with screen recordings + AI presenter
- Social media clips: Short-form content for LinkedIn and Twitter
- Customer-facing content: Onboarding videos, FAQ videos, feature explainers
Stack integration: Scripts come from my Notion content calendar (populated by insights from Fireflies meetings and Typeless brain dumps). Voiceovers are generated by ElevenLabs for maximum quality, then synced with HeyGen avatars.
The Creation Workflow in Practice
Here's how a typical piece of content flows through my stack:
- Spark (Capture): Customer asks a great question on a call (caught by Fireflies) → I dictate "this would make a great tutorial" (caught by Typeless)
- Plan (Processing): The idea lands in my content calendar in Notion with context from both sources
- Script (Creation): I write the script (or dictate it via Typeless), referencing the original customer question for authenticity
-
Produce (Creation):
- Blog post → Written and published
- Audio version → Generated by ElevenLabs (2 minutes)
- Video version → Created in HeyGen with ElevenLabs voiceover (30 minutes)
- Distribute (Distribution): Published across all channels
Total time for one piece of content across all formats: About 2 hours
Before the AI stack: 8-10 hours for the same output
Layer 4: Distribution
This layer is less about AI tools and more about automation:
- Buffer: Schedules social media posts
- Zapier: Routes content to the right platforms
- Notion: Serves as the central hub and CMS
The Full Stack at a Glance
| Tool | Layer | Monthly Cost | Time Saved/Week |
|---|---|---|---|
| Fireflies.ai | Capture | $18 | 5 hours |
| Typeless | Capture | $12 | 4 hours |
| ElevenLabs | Creation | $22 | 3 hours |
| HeyGen | Creation | $48 | 6 hours |
| Notion | Processing/Distribution | $10 | 2 hours |
| Zapier | Processing/Distribution | $20 | 3 hours |
Total monthly cost: ~$130
Total time saved per week: ~23 hours
Effective hourly rate of the stack: About $1.41/hour
Compare that to hiring humans for the same work (easily $3,000-5,000/month), and the ROI is absurd.
Lessons Learned
After a year of building and refining this stack, here's what I wish I'd known from the start:
1. Start with Capture, Not Creation
Most people start by buying AI creation tools (video generators, writing assistants) and then struggle because they don't have good inputs. Start with capture tools (Fireflies, Typeless) and you'll have a constant stream of ideas, insights, and raw material to feed into creation tools.
2. Integration > Individual Quality
A slightly less powerful tool that integrates well with your stack is more valuable than a best-in-class tool that operates in isolation. Every tool in my stack has API access or webhook support.
3. The Human Layer Still Matters
AI handles about 60% of my work. The remaining 40% — strategy, relationship building, creative direction, quality review — is where I add value. Don't try to automate everything. Automate the repetitive stuff so you can focus on the work that actually requires a human brain.
4. Review Everything
I never publish AI-generated content without reviewing it. The tools are good, but they're not perfect. A 5-minute review catches errors that would take hours to fix after publication.
5. Be Transparent
I'm open about using AI tools. My audience appreciates the honesty, and it actually builds trust. "I used AI to help create this" is not a weakness — it's a sign that you're leveraging modern tools effectively.
Getting Started
If you're building your own AI assistant stack, here's my recommended order:
- Week 1: Set up Fireflies.ai for meeting capture
- Week 2: Add Typeless for voice-first documentation
- Week 3: Start creating audio content with ElevenLabs
- Week 4: Add HeyGen for video production
- Week 5-6: Build automations connecting everything
Don't try to set up everything at once. Each tool needs a week or two to become habitual before adding the next one.
Want weekly recommendations on the best AI tools and workflows? I share my latest discoveries, stack updates, and productivity tips in my newsletter: AI Product Weekly. Join 5,000+ professionals who are building smarter workflows with AI.
Disclosure: Some links in this article are affiliate links. If you purchase through these links, I may earn a commission at no additional cost to you. I only recommend tools I personally use and believe in. All opinions are my own.
Top comments (0)