AI Tools I Actually Use Every Day

I run two companies with AI as the operational backbone. Not as a buzzword. As infrastructure. Here is what I actually use every day, what it does, and why it matters.

No affiliate links. No sponsored recommendations. Just the tools that are running right now.

Claude

Claude is my primary AI. I use it for strategic thinking, content generation, code development, analysis, and decision support. It is the brain behind most of the automated systems I have built.

I chose Claude over GPT for a specific reason: it handles nuance better. When I am working through complex business strategy or writing content that needs a specific voice, Claude consistently produces output that requires less editing. For code generation, it understands context across large codebases in a way that saves significant time.

One of my clients, a financial services professional, put it this way after switching from ChatGPT at my recommendation: "We plug the same stuff into Claude, and it gave us much better results." Another coaching client described the output quality: "The documents I put together are beautiful. I sent one to the CEO of my other company, and he was amazed that it came from us." That reaction is common. The difference between Claude and other tools is not marginal. It changes the quality of what a small business can produce.

I run the Max plan, which means unlimited usage at a flat monthly fee. This matters because it changes the economics of AI usage. Instead of watching token costs, I can use Claude aggressively for analysis, brainstorming, and iteration without budget anxiety. I also use this same plan for all my code editing and book editing work. The Far Transfer manuscript, for example, went through five editing passes entirely in Claude Code. Zero API cost. That is the kind of leverage a flat-rate plan unlocks.

n8n

n8n is my automation platform. It is self-hosted on a Hostinger VPS, running in Docker. I currently have 56 workflows handling everything from email processing to data pipeline management.

The advantage of n8n over Zapier or Make is control. Self-hosting means my data stays on my server. The workflow editor is powerful enough for complex logic without writing code for every step. And when I do need code, I can drop JavaScript or Python nodes directly into the workflow.

The VPS that hosts n8n also runs 67 Docker containers. That is not a typo. Sixty-seven live applications, from client dashboards to game servers to learning platforms, all orchestrated through Docker Compose and Traefik for routing. The entire infrastructure costs less per month than most companies spend on a single SaaS subscription.

The PRISM Agent System

This is the custom-built layer that ties everything together. It is a multi-agent Python system with 34 autonomous agents organized across seven architectural layers:

Layer 1: Ingestion. Pulls transcripts from Google Drive and local Zoom recordings. The system has indexed 862 transcripts and counting.

Layer 2: Triage. Gemini Flash classifies each transcript by type and routes it to the right agent. Discovery calls, coaching sessions, proposals, internal meetings. Each gets different treatment.

Layer 3: Sovereign Agents. This is the command layer. Two sovereign agents sit above everything else. The PRISM Sovereign runs a math-enforced revenue pressure engine with pipeline coverage ratios, pacing percentages, and daily orders. The VersAssist Sovereign monitors delivery stability with throughput calculations, rework rates, and capacity load. These agents do not make suggestions. They issue binding directives based on real numbers.

Layer 4: Advisory Personas. Four agent personas (CEO, CMO, CFO, COO) analyze transcripts through specialized lenses and report to the sovereign layer. The CEO agent tracks strategic decisions. The CMO mines content opportunities. The CFO monitors revenue attribution. The COO flags operational bottlenecks.

Layer 5: Email and Communication. Email triage runs three times daily on the Hostinger server. Gemini Flash classifies incoming emails, drafts responses, labels and organizes the inbox. A separate triage bot handles VersAssist email. A content publisher agent generates weekly copy packages for LinkedIn, YouTube, email campaigns, and Skool recaps.

Layer 6: Prospect and Revenue Pipeline. A BizDev agent scores deal probability and syncs to GoHighLevel CRM nightly. A BNI intelligence agent tracks referral matching across 35 chapter members. A revenue attribution engine links every dollar to its source with confidence scores.

Layer 7: Specialized Agents. A session recap bot runs every 30 minutes between 8 AM and 10 PM, capturing every meeting with extracted action items. A watchdog monitors all agents and alerts me when something breaks. A content miner processes 20 transcripts per day looking for quotable moments. A YouTube QA agent scores thumbnails, titles, and content quality daily. A security audit bot runs monthly scans.

The system maintains shared state across all these agents, processes 288 verified client testimonials from 41 unique sessions, and produces a sovereign dashboard that shows both businesses' health at a glance.

Suno

I use Suno for AI music generation. It is how I produced the 18-track "Mansa Musa: More Than Gold" concept album, which is on Spotify under "Dr. Jeff." The quality has reached a point where the output is genuinely listenable. Not a novelty, but actual music you would choose to play. I maintain a catalog of 43 tracks across multiple projects, and the production workflow involves style prompting, metatag formatting, and careful lyric structure to get consistently good results.

Gemini Flash

Every automated bot in my system runs on Gemini Flash. Not Claude. The cost discipline matters: Claude is for strategic work (Max plan, flat fee). Gemini Flash is for volume operations where you need thousands of API calls per day without budget anxiety. Email classification, transcript analysis, content mining, QA scoring. All Gemini Flash. The cost per operation is fractions of a penny.

This separation is deliberate. Most people pick one AI and use it for everything. That is like using a sledgehammer for finish carpentry. Claude handles the work that requires nuance, long context, and sophisticated reasoning. Gemini Flash handles the work that requires speed, volume, and low cost. The two together cover every use case I encounter.

Vanilla JavaScript and Docker

I build web applications in vanilla JavaScript. No React. No Next.js for the games and interactive tools. Pharmageddon, a browser-based game I built about pharmaceutical industry dynamics, runs entirely on vanilla JS and is deployed as a Docker container on the VPS. The game explores real dynamics from my 18 years in the pharmaceutical industry, and it runs in any browser without a build step, a bundler, or a framework dependency.

Same approach for District Zero, Referral Rush, Max the Flying Chicken (my daughter Layla's game), and other projects in the portfolio. The pattern is simple: build it clean, containerize it, deploy it behind Traefik. The entire game portfolio runs on infrastructure that costs less than a single cloud function setup on AWS.

The choice of vanilla JavaScript is intentional. Frameworks add complexity and maintenance burden. When you are running 67 containers on a single VPS, simplicity at the application layer is how you keep the whole system manageable. Every container starts fast, serves static files efficiently, and does not require periodic dependency updates that break things.

Google Drive and Otter.ai

Every coaching session I run gets recorded and transcribed through Otter.ai, then stored in Google Drive. The transcript pipeline pulls these automatically, indexes them, and feeds them into the agent system. This is the raw data layer that powers everything else. The 862 transcripts in the index represent the actual conversations that generated the 288 testimonials in the database. Every claim I make about client results traces back to a specific transcript with timestamps.

This matters because it creates an accountability loop. When one of my clients says "we plug the same stuff into Claude, and it gave us much better results," that quote exists in a transcript with a date, a session name, and a timestamp. It is not a marketing fabrication.

YouTube

I run a YouTube channel with 103 published videos. The production pipeline is increasingly automated. A content miner agent identifies 30-to-60-second clip candidates from transcripts daily. A Shorts producer generates output using Kokoro text-to-speech and AI image generation through FFmpeg. A QA agent scores every piece before it goes out. The channel serves as both a content engine and a proof-of-concept for the AI workflows I teach to clients.

One client saw a music video I produced using these tools and said: "It was like, amazing. It was Hollywood." The production quality that is achievable with AI tools today is not a compromise. It is genuinely competitive with professional output, at a fraction of the cost and turnaround time.

GoHighLevel (GHL)

GoHighLevel is the CRM that holds the prospect pipeline. The BizDev agent syncs 248 prospects into GHL nightly, maintaining six pipeline stages. This is the system of record for where every deal stands, what the last contact was, and what the next action should be. The integration is automated. Prospects flow in from coaching sessions, speaking events, BNI referrals, and inbound inquiries. The CRM keeps them all organized without manual data entry.

The Integration Layer

The tools matter less than how they connect. Claude generates the intelligence. n8n orchestrates the automation. The 34-agent system provides persistent context and autonomous operation. Gemini Flash handles the volume processing. Suno handles creative production. Docker and Traefik handle deployment. GoHighLevel tracks the pipeline. Google Drive stores the raw data. Each tool does what it does best, and the connections between them are where the real leverage lives.

One of my clients said it clearly after a single coaching session: "All that you did in that one hour, it was just like, oh my God. And I don't get excited about a whole lot of stuff." That reaction comes from seeing how these tools work together, not from any one tool in isolation. Another client, after watching the file organization capabilities, said simply: "Now this is just opened up the door for, okay, now I have to do X, Y and Z, oh, I'm going to use Claude."

That is the actual stack. No magic. Just systems, built over time, running every day.