Has anyone reviewed both the MIT success rate and the ChatGPT success rate?

Generative AI
2.5k viewscircle icon3 Upvotescircle icon5 Comments
Sort by:
Chief Technology Officer in Healthcare and Biotech14 days ago

Yes but we have done way better. I can explain if you want to.

Chief Information Officer2 months ago

This is a big issue. I’m involved in an industry survey to assess real adoption rates, prompted by the MIT study. Many organizations deploy Copilot or Enterprise ChatGPT without tracking actual usage. Education, training, and enforcement are critical for successful adoption. Firms with top-down mandates and enforcement see dramatically higher adoption rates, while others experience around 20% uptake. Monitoring usage and quality remains challenging. Preliminary findings suggest Copilot adoption is low outside specific functions like meeting transcription. Contextual intelligence is essential for effective AI use, and organizations must layer this capability on top of existing tools.

1 Reply
no title2 months ago

Microsoft has been slow to implement contextual intelligence in Copilot, requiring users to repeatedly provide context. Recent updates have improved this, but the ability to select different LLM models, such as those from Anthropic, may further enhance Copilot’s capabilities.

Director of IT2 months ago

I’ve reviewed both. The ChatGPT report provides metrics on usage, segmented by categories such as asking, doing, and expressing, with insights into effectiveness drivers. The MIT report analyzes corporate success rates in scaling AI from pilot to enterprise, identifying patterns for success. I combined both reports in Google’s notebook LLM to create a podcast.

Lightbulb on1 circle icon1 Reply
no titlea month ago

Oh please, mods please make external links work and Steve, please do share your Notebook podcast! That's a real benefit to the community.

Content you might like

Question for CIOs and heads of IT: What is the biggest challenge you face when you roll out GenAI to employees and why?

Read More Comments

Have you deployed AI agents? What’s been an unexpected use case you’ve discovered?

Which skills do you believe are becoming LESS important as AI interfaces improve? Select all that apply.

Technical prompt engineering14%

Understanding of model architecture32%

Manual data preprocessing35%

Custom model training25%

Specialized AI coding skills20%

Knowledge of prompt optimization techniques17%

Understanding of token usage and optimization13%

Something else4%

View Results

How does your company ensure control and consistency of its AI agents?

We don’t yet have a governance framework19%

We use manual checks or periodic reviews43%

We have defined policies and guidelines for the creation and deployment of agents, which each team ensures independently29%

We have introduced a platform to centrally manage policies, versions, and monitoring of agents5%

None of the above5%

View Results