I’ve been integrating AI/ML solutions since before the current gold rush began. However, after years of real-world experience*, the reality of AI looks nothing like the early promises.
* Based on real daily team usage
Pre 2022
- BERT – for SEO and ASO. Sentiment analysis, classification, QA.
- StyleGAN2, ESRGAN – for static images and photography. Generation, manipulation, upscale.
- RIFE, Topaz – for videos. Upscaling, frame interpolation.
2022
- ChatGPT – QA, coding, SEO, ASO, summary, grammar check, emails
- Midjourney, StableDiffusion, DALL-E – for static images and photography
- Whisper – voice recognition
- N8N – automation
Wow! ChatGPT and MidJourney just blew my mind! It was incredible.
Built custom tools: advanced prompt generator, visual styles library for MJ
2023
- Claude, ChatGPT, Bard, Perplexity, Mistral, Llama, Grok, GLM, Qwen – coding, research, analysis, brainstorming, prompt building, assistants, knowledgebase, criticism, SEO at scale, automations, fine-tuning, marketing strategies, support, web chatbots, staff training, learning, translation, QA, ASO, text styling, grammar check, emails
- Midjourney, Ideogram, StableDiffusion – static images, design concepts
- Ada + Pinecone + Claude – RAG
- Puppeteer + Claude – crawling and parsing
- ElevenLabs – text to speech
- RunwayML – short videos for socials
- HeyGen – virtual video avatars
- Whisper – voice recognition
- Suno – music for socials
- Meshy – 3d models
- VS Code – IDE
- N8N – automation
I started building massive systems around the best available LLM models – from “second brains” for SMBs and sales web agents to virtual video avatars. I engineered the first tailored solutions for my partners.
2024
- Claude, ChatGPT, Gemini, Grok, Perplexity, ChatRTX, Mistral, Llama, GLM, Kimi, Qwen – coding, research, analysis, brainstorming, agents, knowledgebase, prompt building, SEO, criticism, custom tools, automations, marketing strategies, support, web chatbots, staff training, realtime translation, learning, QA, ASO, text styling, grammar, emails
- Midjourney, Ideogram, StableDiffusion, Flux, Imagen – static images, design concepts
- te3s + Pinecone + Claude – RAG
- Firecrawl + Claude – crawling and parsing
- ElevenLabs – text to speech
- Luma, Kling, Sora, RunwayML, Seedance – short videos for socials
- Whisper – voice recognition
- HeyGen – virtual video avatars
- OpenAI, ElevenLabs – virtual voice avatars
- Suno – music for socials
- Cursor – IDE
- N8N + LLMs – automation
I continued incorporating tailored AI solutions into the business processes of my partners and clients. My own team relied heavily on vibe coding, RAG, and AI brainstorming. Our workflows were mostly automated. We used AI for everything, every day, but with human supervision.
Built custom tools: voice recognition for desktop/mobiles, nutrition tracker, Instagram engagement bot.
2025
- Gemini, Claude, ChatGPT, Deepseek, Grok – code snippets, custom tools, agents, knowledgebase, web chatbots, realtime translation, QA, text styling, grammar
- Imagen, Nano Banana Pro, Ideogram – static images
- te3s + Pinecone + Claude – RAG
- Firecrawl + ChatGPT – crawling and parsing
- Veo, Kling, Wan, Luma – short videos for socials
- GPT-4o – voice recognition
- Antigravity, Windsurf, KiloCode – IDE
- MCP server, N8N + LLMs – automation
Eventually, I realized that the actual output requires a lot of human attention, much more than I was aiming for. I hoped it would improve, but to no avail. In many cases, it was more efficient to do things manually, the old way. Precise and predictable.
With limited AI involvement.
2026
- Claude, Gemini, Xiaomi – small code snippets, simple custom tools, web chatbots, translation, quick questions, text styling, grammar
- Nano Banana Pro – static images
- GPT-4o – voice recognition
- Veo – short videos
- te3s + Pinecone + Claude – RAG
- Firecrawl + Claude – crawling and parsing
- Cursor – IDE
- N8N + LLMs – automation
So...
After almost four years of mainstream, we can clearly see what AI is actually capable of and what was just marketing hype. In practice, AI has proven useful for some of our tasks (no doubt, especially image/audio generation), but we need to set realistic expectations and adjust our workflows accordingly.
When we can use AI, what to expect from it, and when it’s better to do the work manually.
Digital Growth Alchemist
Den is a Fractional CDO & Creative Director who builds comprehensive digital growth systems for ambitious SMBs and startups. With 5+ years in C-level roles overseeing $7.8B in transactions and leading teams of 860+ people, backed by 28+ years of technical experience, he combines executive leadership with hands on expertise in design, development, AI, and marketing. His work has been recognized with industry design awards and federal medals.
