Recently, artificial intelligence (AI) has rapidly penetrated nearly every industry. This trend has product managers (PMs) asking: “How can I incorporate AI into my product?” Sometimes, this question leads to smart decisions. But often, it turns into a rushed and ineffective move just to keep up with trends.
According to McKinsey’s 2024 report, 55% of companies have already integrated at least one AI feature into their products or processes.
PwC predicts that AI will add $15.7 trillion in value to the global economy by 2030.
Gartner reports that products using AI have 25% higher customer satisfaction compared to those that don’t.
However, I recently tested an AI chatbot launched by a company I work with. It seemed interesting at first, but quickly became clear it was more about saying “we have AI too” rather than solving a real user problem or need. The key takeaway: Implementing AI isn’t just about creating a chatbot. It requires an ecosystem approach — identifying the right problem, working with quality data, and using technology to create real value.
Three key steps PMs should follow when applying AI:
Identify the right problem:
Starting with “Let’s implement AI” is risky. Instead, ask:
What user or business problem are we solving?
Is AI really the best technology for this?
For example, if users struggle to find what they want on a platform, an AI-based recommendation system can help. But doing AI just because “everyone else is” wastes resources and money.
Quality data is essential:
AI’s foundation is data. Without clean, complete, and sufficient data, results will fail. Ask:
Is our data clean?
Was it collected ethically and with permission?
Do we have enough data for at least a minimum viable product (MVP)?
Leading companies like Google and Meta invest millions in data quality — it’s a key success factor.
Build a new model or use existing ones?
OpenAI, Google, Amazon, and others offer many ready-to-use AI models (GPT, Gemini, Bedrock, etc.). The main question is:
- Do we need to build a new AI model?
- Or can we integrate an existing one?
Often, integrating existing APIs is faster and more cost-effective than building from scratch.
Common mistakes:
- Implementing AI just to impress CEOs or investors, without measurable user impact — this results in AI-flavored interfaces with no real value.
- Thinking AI equals chatbots. AI goes far beyond chatbots — it powers recommendation engines, automated responses, and decision-support tools behind the scenes. PMs should understand what APIs or models are driving these features.
- Trying to launch AI features without a dedicated AI team. AI implementation requires engineers, data analysts, MLOps specialists, and ethical oversight. If these skills are missing, either collaborate with the right partners or postpone AI ambitions.
AI can make our products smarter, but only when implemented for the right reasons and in the right way.
