Here we are, kicking off the last quarter of the year, and it's been half a year since the Chat GPT Era began. And honestly? Most AI products suck. I've not yet come across an AI product that truly wowed me. It's like watching Thor's hammer, being used to build crooked houses. Modern LLMs are such a potent technology, but product-wise? Nothing that's knocked my socks off.

Even the mighty Chat-GPT; it’s great, but what is it for? Which problem is it solving? What about the UX?

Therefore, I've compiled a list of common errors to watch for when creating AI products, for engineers, entrepreneurs, product managers, and anyone who’s playing with AI no-code tools trying to create something useful & valuable.

Common Problems

Product-Market fit vs Market-Product Fit

Do things that don’t scale (1).  Instead of seeking customers for your tool, venture out and address real people's challenges. Conduct user interviews. If you're unfamiliar with the process, I suggest checking out the "Mom Test" book (2). But to sum it up: Pose questions and genuinely seek to understand. A good rule of thumb: If you're speaking more than you're listening, you're likely not doing it right

Lack of Originality on the Value Propositions

Many folks seem to disregard these rules from "The 22 Immutable Laws of Marketing" (3)

Law #1: Be the First in Your FieldLaw #2: If You Can’t Be First, Create a New Category

Merely having a presence in the market isn't sufficient. It's crucial to deeply understand the market and serve it with passion. Surprisingly, user research, an essential component, is frequently sidelined.

Poor User Experience

Even with technical products, functionality isn't the sole focus. While adopting a "Lean mindset" is commendable, the goal shouldn't be just an MVP (minimum viable product). Instead, in a highly competitive landscape, strive for an MLP (minimum lovable product)(4)

The illustration below provides a visual representation of this concept:

This perspective encompasses not only the user interface but also the performance of your products, including the performance of the AI models you work with - including when they hallucinate.

Unbalanced Founding Teams

The buzz around AI appears to be eclipsing basic team-building principles. I’m seeing a surge in teams made up solely of developers or engineers, missing a comprehensive business and design perspective. The most effective teams embody the HHH principle - Hacker, Hipster, Hustler. In simpler terms, this translates to a balanced mix of Business/Product, Design, and Engineering.

Overpromising capabilities

The mantra should be: underpromise and overdeliver. Be transparent about the current state of your product and resist the temptation to monetise prematurely. Consider your new-born product like a baby. You wouldn't send a baby to work; instead, you nurture it, educate it, and ensure a safe environment for its healthy growth.

Clearly articulate what your tool can and will do, as well as its limitations. Stay away from monetising too soon. Prioritise getting it into the hands of as many users or testers as possible and glean insights from their feedback.

Fine-tuning before achieving Product-Market Fit (PMF)

Start with a broad approach. Initially, leverage powerful, general models to understand what your users want/need. Only after gaining this understanding should you transition to use-case-specific AI models or dive into fine-tuning. Jumping into the latter prematurely can be costly, especially when you haven't fully comprehended your users' needs.

Overlooking legality, safety, privacy or regulations

We are all familiar with the saying, "better safe than sorry." While delving into legalities and regulations might not be the most thrilling task, it's essential. Dedicate time to understanding the regulations; In Europe, for instance, there's the AI Act. It may be broad and somewhat ambiguous, but expect more specific regulations to come soon. And let's not forget our old friend: GDPR. Always communicate transparently on your website about the data you share with third-party tools. Be clear and upfront.

Underestimating operational costs

Powerful models can be costly, mainly when aiming for heavy user testing before monetizing. It’s crucial to crunch the numbers and understand your financial runway. Determine how long you can sustain operations before needing to monetize, especially if you’re bootstrapping.


Go back to the basics:

  • Focus on Problem Space vs. Solution Space.
  • Prioritise Market-Product Fit over Product-Market Fit.
  • Be original**.** Find a unique niche and be creative with your value proposition
  • Grow evenly. UX & functionality
  • Team up right**.** Mix Product, Design, and Engineering.
  • Promise less, deliver more.
  • Start general, then fine-tune
  • Stay updated, know, and follow regulations
  • Assess costs; plan long-term.
  • Enjoy the ride!

Links & Sources






Assess your AI Maturity. Increase your organizations AI Value

Understand your organization's current maturity level & find out how to create more AI Value.