It’s easy to get lost in the hype and speculation of AI. But it’s not all hype, and it's important to stay informed of what's really happening on the ground.
Amazon CEO, Andy Jassy, just posted on LinkedIn about how using Amazon Q helped them save 4,500 developer years of work (yes, years) by helping update applications from old Java versions to Java 17. AI is clearly capable of saving huge amounts of time but most companies aren't getting those kinds of results.
So how are businesses and developers actually using AI, and what impact is it having?
Retool surveyed 750 tech professionals across various roles and industries to get some solid info about how people actually use AI for business.
Here’s an overview of their State of AI 2024 report.
AI sentiment in 2024
Despite the AI hype cycle reaching fever pitch in in late 2023 and early 2024 – the sentiment around AI in 2024 remains surprisingly measured. Most respondents consider AI to be slightly overrated.
- 53% consider AI overrated
- 22% consider AI fairly rated
- 24% consider AI underrated
This tempered view isn't necessarily negative. Rather, it reflects a growing understanding of AI's current limitations alongside its vast potential.
As one respondent noted, "AI is being shoehorned into products without really adding value." There's a sense that the aggressive hype has temporarily obscured AI's true potential, leading to unrealistic expectations and misapplications.
However, this pragmatic outlook is coupled with optimism about AI's future utility and breadth of applications.
The consensus seems to be – this is just the beginning.
Slow and steady adoption
Contrary to sensationalist headlines, AI adoption isn't skyrocketing across the board. Only about 30% of respondents consider their companies to be "running" or "flying" when it comes to AI adoption.
This represents a slight decrease from 2023, with those feeling they're leading dropping from 13.4% to 9.8%.
At the same time, AI tools are becoming increasingly integrated into daily work. Almost three-quarters of respondents use AI tools like copilots or ChatGPT in their work at least weekly – with 56.4% approaching daily use.
Product and Engineering teams are leading this charge – with 68% and 62.6% respectively reporting daily(ish) adoption.
Productivity: More use = more return
Whether AI is overrated or not, 85.5% of people who use it report increased productivity. So it’s definitely not all hype.
The people who think it’s overrated or not really useful likely don’t use it much, if at all. It seems like you have to use it every day to really get the most out of AI.
It also turns out that 64.4% of daily users report significant productivity improvements, versus 17% of weekly users and 6.6% of occasional users.
This suggests that the more familiar and comfortable people become with AI tools, the more value they're able to extract from them. Which makes good sense.
Use cases beyond LLM chatbots
While AI-powered chatbots have become the poster child for AI applications (with 55.1% of respondents having built or used one), the survey reveals a diverse range of use cases across different departments.
The top internal use cases are:
- Writing code or queries (42.1%)
- Knowledge base Q&A (36.4%)
- Support chatbots (33.9%)
AI use is spreading across departments too – with Engineering (62.5%), Marketing (43.9%), and Data Science (43.2%) leading the pack.
Interestingly, respondents see more promise for internal AI applications (33.7%) than external ones (8.5%), suggesting that companies are focusing on enhancing internal processes before customer-facing applications.
The modern AI stack
When it comes to the tools powering these AI applications, some clear leaders emerge:
- LLMs: OpenAI continues to dominate – with its models accounting for 76.7% of usage. GPT-4 leads at 45%, followed by GPT-3.5 at 25%. But Claude 3.5 Sonnet might change that.
- Vector databases: Usage has skyrocketed from 20% in 2023 to 63.6% in 2024, with Pinecone, Weaviate, and Milvus leading the pack.
- Development tools: 59.1% of companies are using AI development tools, with HuggingFace and custom solutions being the most popular.
Interestingly, the majority of respondents (51.9%) aren't using an inference platform at all, which may reflect the associated hardware requirements and costs.
AI customization challenges
Most respondents are customizing their AI models to some extent – with fine-tuning existing models (29.3%) and using vector databases or RAG (23.2%) being the most common approaches.
However, this customization comes with its own set of challenges.
The top pain points in developing AI applications are:
- Model output accuracy/hallucinations (38.9%)
- Lack of available technical expertise/resources (38.2%)
- Data access/security concerns (33.5%)
These challenges highlight the need for continued investment in AI education and infrastructure, as well as robust data governance practices.
Is AI really worth the hype?
While the initial hype is being replaced by a more grounded understanding of AI's capabilities and limitations. Companies are finding real value in AI applications – particularly for internal processes – but are grappling with significant challenges around accuracy, expertise, and ethical considerations.
The focus seems to be shifting from flashy demonstrations to practical, ROI-driven applications. The rapid adoption of tools like vector databases, RAG-based LLMs, and the increasing customization of models suggest that companies are getting serious about integrating AI into their operations in meaningful ways.
You can read the full 2024 State of AI report from Retool here.
The industry still has work to do. Addressing bias and fairness, closing the expertise gap, and solving data access and security concerns will be crucial for realizing AI's full potential.
If you’re in a place where you need help with AI at enterprise scale, we can guide you and start building AI tools for your business.
How do I hire senior AI engineers to build internal AI tools?
You could spend the next 6-18 months planning to recruit and build an AI team (if you can afford it), but you won’t be building any AI capabilities. That’s why Codingscape exists.
We can assemble a senior AI development team to start building internal AI tools for you in 4-6 weeks. It’ll be faster to get started, more cost-efficient than internal hiring, and we’ll deliver high-quality results quickly.
Zappos, Twilio, and Veho are just a few companies that trust us to build software with a remote-first approach. We’d love to help your company harness the power of generative AI and LLMs.
You can schedule a time to talk with us here. No hassle, no expectations, just answers.
