2024 State of
AIAgents

AI hits new highs in 2024—here’s how it evolved. After serving 184 Billion tokens, 786 million API requests by 36K developers, we collected insights from over 3,400 developers building AI agents.

Who did we survey?

The 2024 State of AI Agents Survey received over 3,400 responses from 100+ countries. 46% of participants were in C-level roles, reflecting AI's importance at the top.

The survey captures insights from developers, industry leaders, and professionals across diverse sectors and company sizes worldwide.

After handling an incredible 184 billion tokens and 786 million API reqs from 36K developers, we’ve gathered invaluable insights from 4K builders, the next-gen of AI agents.

Which LLM providers are developers using to build AI agents?

Developers rely heavily on OpenAI for AI and LLM services, but Google is quickly emerging as a strong competitor. Anthropic follows as a popular choice, while Meta’s Llama, Mistral, and Cohere maintain smaller but growing presences, highlighting a dynamic market landscape.

CohereCohere4%MistralMistral11%Meta (Llama)Meta (Llama)38%AnthropicAnthropic47%GoogleGoogle59%OpenAIOpenAI76%
Usage
(% of respondents)
Providers
Serverless AI Agents
Ship personalized AI agents with ease. Build, collaborate, and deploy with Langbase’s powerful serverless platform.
Sign Up for Free

What specific purposes do you use various AI models for?

OpenAI leads in tech and marketing applications, with strong adoption in translation tasks. Anthropic is highly favored for technical tasks but sees lower use in marketing and translation. Google's models dominate in health and translation, showcasing strengths in language and medical domains. Meta is widely used for tech and science applications, while Cohere is valued for balanced use across multiple areas, including science and marketing.

OpenAI
Strong in tech and marketing but lags in health applications.
76%
Tech
48%
Science
83%
Marketing
37%
Health
56%
Translation
Google
Dominates translation and health while maintaining strong marketing capabilities.
58%
Tech
31%
Science
82%
Marketing
76%
Health
86%
Translation
Anthropic
Excels in tech and science but has minimal focus on translation.
87%
Tech
54%
Science
41%
Marketing
45%
Health
18%
Translation
Meta
Balanced in tech and science but the models are weak in translation.
77%
Tech
76%
Science
48%
Marketing
37%
Health
15%
Translation
Cohere
Strong in tech and science with moderate capabilities across other domains.
74%
Tech
63%
Science
65%
Marketing
52%
Health
50%
Translation
Mistral
Focused on tech and science but lags significantly in other areas.
61%
Tech
54%
Science
23%
Marketing
15%
Health
9%
Translation
Multi-agent AI pipelines
Simplify the design, collaboration, and deployment of multi-agent AI pipelines with Langbase.
Start Building

Top concerns preventing wider adoption of AI agents in your worklows

Complexity in scaling and deployment leads as the primary concern, followed closely by data privacy and security compliance. The lack of robust monitoring tools and high infrastructure costs also hinder adoption. Resistance or skepticism about AI-driven solutions reflects ongoing apprehension, signaling the need for more transparent and user-friendly AI implementation platforms.

Concerns like...
High infrastructure costs (e.g., storage, APIs, operational limits)
35.0%
Complexity in scaling and deployment
59.0%
Lack of robust observability and monitoring tools
50.0%
Concerns around data privacy and security compliance
55.0%
Resistance or skepticism about adopting AI-driven solutions
40.0%
Near-Zero Hallucinations
Reduce hallucinations and improve AI performance with Memory Agent.
Deploy Your First Memory Agent

What factors influence choice of LLM?

When choosing a large language model (LLM), most respondents prioritized accuracy, followed by security and customizability. Cost was the least influential factor.

Factors like...
Performance
45.0%
Security
24.0%
Customizability
21.0%
Cost
10.0%
BaseAI: Ship AI Faster
The first web AI framework for building serverless AI agents.
Start Building

What are the biggest challenges in deploying AI agents to production?

Deploying AI agents to production faces key challenges: difficulty in customization, limited evaluation methods for quality assurance, and a lack of reusable infrastructure. Fragmented tools, integration issues, and scalability concerns further complicate the process, emphasizing the need for simple pipelines and robust support tools.

Challenges like...
Lack of reusable infrastructure for multi-agent pipelines
85.0%
Difficulty in creating or customizing agents
90.0%
Limited AI acumen in the team
95.0%
Integration challenges with current systems
70.0%
Limited evaluation methods to ensure quality outcomes
88.0%
Fragmented tools requiring integration
75.0%
Managing scalability and performance issues
65.0%
Infinite Context With Memory Agents
Break the universal limit of the context window with the Memory Agent for infinite context and precision.
Create Memory Agent

What are your primary goals when adopting AI technology?

Automation and simplification are the top priorities for AI adoption, enabling efficiency and streamlined processes in the company. Custom solutions and improved collaboration reflect a growing interest in flexibility and shared system access.

Goals like...
85.0%
Automation
35.0%
Productivity
65.0%
Custom
75.0%
Code
Serverless Chatbots
Build and deploy AI chatbots with personalized memory and real-time data integration using Langbase.
Build Your Chatbot

What use cases does your company use LLMs for?

LLMs are widely used in software development, with notable adoption in marketing, IT operations, and text summarization. Customer service, HR, and legal show emerging interest, signaling potential for broader adoption in 2025.

Text gen and summarization
59.0%
IT & Operations
48.0%
Marketing & communication
50.0%
Software development
87.0%
Customer service
43.0%
Legal & Compliance
15.0%
HR & Finance
26.0%
Accelerate AI Development
Build smarter, faster, and with less code using Langbase's SDK.
Get Started

Which AI infra feature is critical for your immediate needs?

Most respondents need multi-agent retrieval-augmented generation (RAG) capabilities to improve contextual information processing. Evaluation tools are also important to ensure AI systems are working as expected. Multi-agent pipelines are key for implementing complex tasks in production.

Features like...
82.0%
Retrieval-augmented generation
Memory Agents
68.0%
Multi-agent automation pipelines
Pipe agents
75.0%
Evals & Observability
AI Studio
Langbase Studio
Build, collaborate, and deploy AI agents effortlessly with Langbase Studio. Real-time testing, seamless scaling, and team collaboration in one platform.
Try it out

When orchestrating AI pipeline, what type of tools do you prefer?

The majority of respondents prefer dev tools that offer flexible, foundational primitives to design tailored AI pipelines. While prebuilt, point solutions address specific problems, they are less customizable, indicating a strong need for customization in AI workflow design.

We prefer...
Prebuilt drag & drop point solutions
24.0%
Flexible, composable primitives for agents, tools, memory
76.0%
Docs & Support Agent
Transform your documentation into an intelligent AI assistant. Enable fast, accurate answers powered by your content.
Try Docs Agent Now

What is the primary factor that would influence your choice of an AI development platform?

Developers value AI agent version control as the top feature for an AI development platform. A strong SDK or library ecosystem and local development environments also valued. Team collaboration and experimentation are important but secondary considerations and resource monitoring dashboards are less critical.

AI agent version control
86.0%
Experiment tracking features
65.0%
Team collaboration capabilities
72.0%
Local development environment
78.0%
Resource monitoring dashboards
55.0%
SDK/library ecosystem
82.0%

How extensively is AI used in your company?

The majority of developers use AI for both experimentation and production. Developers are steadily adopting AI for production.

We use AI in...
Experimentation
95.0%
Production
35.0%

Who did we survey?

The 2024 State of AI Agents Survey received over 3,400 responses from 100+ countries. 46% of participants were in C-level roles, reflecting AI's importance at the top.

The survey captures insights from developers, industry leaders, and professionals across diverse sectors and company sizes worldwide.

Thanks to everyone who contributed to the 2024 State of AI Agents Survey. We hope these insights provide a valuable understanding of the evolving AI landscape and inspire innovation across industries.

The only and most powerful

Serverless Developer Platformfor building AI products.