This's project AI Assistant - 2023-2024
What's a AI Assistant?
AI Assistant
How We Built an AI Assistant for Traders at Bitsgap?
The Problem
What were the hypotheses?
First Hypothesis before the start
What if we created an AI assistant that curates news and trading opportunities tailored to each user's strategy and risk appetite?
Hypothesis
Has a customer survey been conducted?
1 - Respondents
1,000+ users contacted, 258 responses received.
2 - What do they use?
18% said they read news daily, even while using automation.
3 - How many agreed?
40 users followed popular traders for insights.
4 - Conclusions
8% tracked Twitter, news channels, and forums - but only traded on spot, prioritizing safety.
Next, we launched a survey to validate our assumptions.
This survey was designed to understand how traders make decisions and how AI can help reduce research overload.
Customer Survey
Have you continued your research?
Top answers - 1
How they discover opportunities?
Top answers - 2
How they discover opportunities?
Top answers - 3
Where they feel uncertain?
What happened next?
From MVP to Real Users
Has the hypothesis been confirmed?
1 - Testing the hypothesis
Created a landing page with a pre-order waitlist and real paywall (card pre-auth)
2 - Result
Ran a targeted campaign → 58 paid subscribers ready to wait for the product
Wow! Cool
Subscription Tiers
What is the result?
APIs (Reddit, Twitter/X, news feeds) cost a lot at scale. After technical review, we pivoted
1 - Testing the hypothesis
Used our own 5 years of trading data
2 - Integrated
Integrated backtesting engine
3 - Focused
Focused on internal AI, not external NLP APIs
And what did you come up with in the end?
Subscription Options
Is it a completely different product?
Yes, we are focused on the new product.
For example, there were selling phrases like "try our artificial intelligence trading assistant, created based on our verified data, and not based on news analysis."
What were the next steps?
We started designing and prototyping
Our north star is a zero-friction installation. We divided it into 4 steps:
1 - Onboarding
One-click onboarding
2 - 3 risk levels
Conservative, Balanced, Aggressive
3 - Pair recommendations
Conservative, Balanced, Aggressive
4 - Clear UI
Clear UI with probability ranges, not just "Signals"
What's next?
Component Library
Multiple widgets - 1
Strategy selector
Multiple widgets - 2
Profit simulation
Multiple widgets - 3
AI suggestions panel
Multiple widgets - 4
Smart alerts & notifications
Was there a product launch?
Launch - 1
Updated marketing site
Launch - 2
Created onboarding email sequence
Launch - 3
Added announcement banner on web app
Live Prototype 2
Live Prototype 1
The Outcome - 1
80% early users joined the test
The Outcome - 2
Increased subscription revenue
The Outcome - 3
Growth in MAU and CRR
The Outcome - 4
Positive user feedback + low churn
What was the result?
Result - 1
User needs: clarity, safety, speed
Result - 2
Business goals: monetization, retention, differentiation
Result - 3
And we did it without burning money on expensive APIs.
What are the metrics?
Product Adoption 1
Product Adoption 2
Excellent results! Is the product still working?
Yes, the product works, brings money to the company and helps users earn money in trading.