Add AI to your product or website, in ways people actually use
I add AI where it does real work: chatbots that answer from your content, search across your documents, assistants, and agents. One AI feature costs $5,000–20,000; a chatbot or search over your content, $15,000–60,000.
Scope
What I build
AI chatbots trained on your website or documents
Search over your own content (RAG)
AI assistants inside your product
AI agents that take actions in your tools
Personalized content and outreach at scale
AI added to a product that's already live
Pricing
How much it costs to add AI to a product
Freelance build prices in US dollars, plus the monthly AI running cost (the fees you pay the model provider). The more the AI can see and do, the more it costs to build and run safely.
| Type | Price | Timeline | What you get |
|---|---|---|---|
| One AI feature on an existing screen | $5,000–20,000 | 2–4 weeks | Summaries, drafting, or classification in one place. Running cost usually $50–500 a month. |
| Chatbot or search over your content | $15,000–60,000 | 4–10 weeks | Answers from your website, documents, or data, with sources. Running cost usually $200–2,000 a month. |
| Assistant inside your product | $20,000–80,000 | 6–12 weeks | Knows the user's own data and the product, and helps them get tasks done. |
| AI agent that takes actions | $40,000–150,000 | 8–16 weeks | Changes records, sends messages, or runs workflows, with limits and approvals. Running cost usually $1,000–10,000 a month. |
Real example: The Chat Company, an AI chatbot SaaS, cost $20,000 over 9 weeks and now handles 100K+ messages a day.
Approach
How I add AI that people actually use
Most AI features fail because they're a chat box added for the sake of it. I start with a task your users already do, then decide whether AI makes it faster or better. Sometimes the honest answer is that it doesn't.
Start with a real task
What your users do today, where it's slow, and whether AI would actually help. If it wouldn't, I'll say so.
Ground it in your data
The AI answers from your content and data, not guesswork. Each customer's data stays separate.
Keep costs predictable
The right model for the job, limits on how much text goes in, caching, and usage tracking, so the monthly bill doesn't surprise you.
Make it reliable
Fallbacks when the AI is slow or wrong, limits on what agents can do, and monitoring in production.
Capabilities
What I can handle
What users get
- Chatbots and assistants
- Search that understands questions
- Drafting and summaries
- Personalization
- Actions and workflows
Your data
- Website and document ingestion
- Search over your content (RAG)
- Customer data kept separate
- Answers with sources
Cost and speed
- Choosing the right model
- Caching
- Token and usage limits
- Background processing
Production
- Fallbacks when AI fails
- Guardrails on agents
- Monitoring
- Embedding on any website
Typical stack
- OpenAI
- Vercel AI SDK
- Vector search
- Next.js
- TypeScript
Difference
What goes into the build
AI running at real scale
The Chat Company's chatbots handle 100K+ messages a day for 30+ businesses, replying about 3x faster than typical chatbot platforms.
AI inside the product, not beside it
On DelightLoop the AI works inside campaigns and personalization, not as a separate chat window. The platform has run around 10K campaigns.
Answers from your content
Chatbots answer from the business's own pages and documents, with each customer's data kept separate, so answers stay accurate and private.
Costs you can plan for
Usage is tracked per customer and capped where it matters, so AI costs grow with value, not by surprise.
Process
How a project works
- 01
A call about the task
What your users need done and where AI could help. You get a written plan with a price, timeline, and estimated monthly running cost.
- 02
A working prototype on your data
A quick version tested on your real content, so you see the quality before the full build.
- 03
Build it into the product
The feature built properly into your app or website, with data separation, limits, and fallbacks.
- 04
Launch and tune
Live with real users, with usage and cost tracking. Then I tune answers and cost based on how people use it.
Clients
What clients say
“We upload a contact list and an hour later everyone has their own page. We run it for every campaign now without needing a developer.”
Product lead Personalization engine: 1,000+ campaigns and 100K+ pages generated
FAQs
How much does it cost to add AI to my product?
One AI feature on an existing screen costs $5,000–20,000 to build. A chatbot or search over your content is $15,000–60,000, an assistant $20,000–80,000, and an AI agent $40,000–150,000. You also pay the AI provider a monthly running cost, from about $50 to several thousand dollars depending on usage.
Can you build a chatbot trained on my website?
Yes. The chatbot learns your pages and documents, answers questions from them, and can be added to any website. The Chat Company, which I rebuilt, does exactly this for 30+ businesses.
How do you stop the AI from making things up?
Give it your own content to answer from, tell it to say when it doesn't know, show sources where possible, and test it on real questions before launch. It won't be perfect, but it will be far more accurate than a general chatbot.
Is my data safe when I use AI?
Your data is sent to the AI provider only to generate each answer. Major providers don't train on business data sent through their APIs by default. Inside your product, each customer's data is kept separate.
Can you add AI to a product that's already live?
Yes. Most AI work is adding it to something that already exists. I build it into your current product without a rebuild.
Is AI actually worth it for my business?
It's worth it when it saves your users real time on a task they do often, or answers questions your team answers repeatedly. If it's only there to say you have AI, it usually isn't.