Big Promises, Hard Reality: Why Some AI Startups Struggle

Introduction

AI companies often make big promises. They show plans in steps or “tiers,” starting small and aiming high. These step-by-step plans look clear and exciting, but not every company reaches the top.

Step-by-Step Plans: Good and Risky

A plan with clear steps outlines what the company is currently doing and what it intends to do next. For example:

• Tier 2: AI sign language that looks real, about 60% easy to understand, and checked by humans.

• Tier 3-4: About 80% clear, faster, and better grammar.

• Tier 5-6: Almost perfect, instant translation.

• Tier 7 or later: As good as a human signer.

  
  

But if progress is slower than promised, trust can drop.

AI and Sign Language Technology — Some Companies Have Closed Down

Photo taken at a conference last year.

Why Some AI Startups Close

Recent examples:

• Builder.ai — Once worth over $1 billion, it closed after people found it used more human work than real AI.

• Some AI startups like Wuri and CodeParrot closed because they couldn’t grow or get more money.

• Scale AI cut jobs, showing even big companies face problems.

Common problems include running out of money, making promises they can’t keep, strong competition, and losing trust.

Extra Risk in Accessibility AI

For sign language AI, the risk is bigger. It must be right in meaning and culture. If the AI signs badly or uses wrong grammar, it can upset the Deaf community. Working closely with Deaf people, testing often, and being open about results are key.

For the last 40 years, many major companies have focused on other areas of disability support, often leaving Deaf and sign language solutions behind. This may be because they believe human interpreters are enough, so they haven’t invested much in new tools. In mid-September 2025, I will share some tools that challenge this thinking. For now, waiting until after August, when most people are in holiday mode and not paying attention.

Tips for AI Founders

1. Promise less, deliver more.

2. Work with the people who will use your product.

3. Have more than one way to earn money.

4. Be honest about what your tech can and can’t do.

   
   

Conclusion

AI is exciting but also risky. Step-by-step plans can work if the company keeps money flowing, beats the competition, and earns trust. In accessibility tech, slow and steady is often the safest way forward.