Pragmatic AI Business Path | Generated by AI
Question: How does this next plan sound? This is from a WeChat conversation with a friend about my future direction.
Answer:
Overall, I think it’s a reasonable and pragmatic direction, especially compared with trying to immediately build a venture-scale AI product from scratch.
From your messages, your plan seems to be:
- Start an AI consulting or AI software outsourcing business.
-
Offer services such as:
- LLM deployment
- Model training/fine-tuning
- Agent development
- AI integration for companies
- Register a company and work with partners.
- Look for projects in the Guangdong–Hong Kong–Macao Greater Bay Area.
- Spend time in Hong Kong and network with higher-end clients and founders.
- Build relationships first, then see what opportunities appear.
- Eventually transition from a service company into a product company if you discover a real market need.
I think this progression makes sense because it reduces risk.
Why I like this plan
You already have several strengths:
- Real software engineering experience.
- Hands-on experience training and deploying LLMs.
- Experience running GPUs and open-source models.
- A technical blog and public portfolio.
- Genuine curiosity—you actually experiment instead of only reading about AI.
Those are much easier to monetize through services than through a new AI SaaS product immediately.
Many successful AI startups actually followed a similar path:
consulting → custom AI solutions → reusable internal tools → product → SaaS
Services generate cash flow while exposing you to real customer problems. Those problems often inspire the best products.
Hong Kong could be useful
Going to Hong Kong to meet founders, investors, and businesses could be worthwhile if your goal is higher-value B2B AI work.
Potential advantages include:
- Companies with larger AI budgets.
- International business environment.
- Easier access to overseas clients.
- Networking opportunities with startups.
The key is to go with a purpose rather than hoping to “meet someone.” For example:
- Attend AI meetups.
- Meet startup founders.
- Talk with software agencies.
- Meet companies that need internal AI systems.
Having concrete offerings makes networking much more effective.
One thing I’d change
Instead of saying:
卖人力 (sell manpower)
I’d think of it as:
Sell expertise and business outcomes.
Clients generally don’t buy “developers.” They buy things like:
- “Automate customer support.”
- “Deploy a secure private LLM.”
- “Reduce manual work by 70%.”
- “Build an internal AI knowledge assistant.”
Framing your work around results is often more compelling than framing it around engineering hours.
I would also narrow the services at first
Rather than offering “anything AI,” pick a few repeatable services, for example:
- Private LLM deployment
- AI coding assistants for companies
- RAG systems over internal documents
- AI agents for business workflows
- Fine-tuning open-source models
- AI infrastructure and GPU deployment
These can become standardized packages instead of one-off projects.
A possible roadmap
- Months 1–3: Take consulting and outsourcing projects, build case studies, and generate revenue.
- Months 3–12: Develop reusable internal frameworks and templates that reduce delivery time.
- Year 2: Identify a recurring customer pain point and turn your internal tools into a product.
- Later: Shift from primarily selling engineering time to selling software and recurring subscriptions.
This approach aligns well with your background because you’ve already been experimenting with LLM training, model deployment, GPU infrastructure, and AI engineering. Those experiences can become differentiators if you package them into solutions that solve concrete business problems rather than presenting them only as technical achievements.
One final thought: don’t view the service business as merely a stepping stone. Some AI consulting firms become highly profitable on their own, while others use the customer insights and revenue they generate to build successful products. Remaining open to both paths lets the market guide your next move rather than committing too early to a single outcome.