2024 State of AI & What to Expect in 2025
AI from a builders perspective - This is a collective view from gAI Ventures CTO/Founders and its AI portco(s) founding teams
We are living through the dawn of the cognitive industrial revolution as described by Reid Hoffman. In 2025 and beyond, we’ll begin to feel the everyday impact of A.I.
The Current State of AI in 2024 - Key highlights first:
Maturity and Mainstream Adoption: AI has moved beyond niche and academic applications to become a mainstream technology. Generative AI attracted over $25 billion in private investments in 2023, nearly nine times the funding of 2022 (as per S&P GMI and Dealroom). S&P GMI reported that in the first three quarters of 2024, GenAI startups secured over $20 billion, indicating that 2024 is likely to exceed the 2023 total. This surge reflects growing consumer adoption as well as enterprise adoption, particularly in customer service, marketing, productivity and creative workflows
NVIDIA continues to dominate the AI industry, with a $3.4 trillion+ valuation. Established generative AI companies are now generating billions in revenue, and startups are rapidly gaining traction, especially in video and audio generation.
Buoyed by a bullish public market, the AI sector’s total value has climbed to $9 trillion (as per the State of AI report published by partners at Air Street Capital), with private investments maintaining strong growth
Companies like OpenAI, Anthropic, and Google advanced their multimodal models, making AI more accessible for diverse use cases, from content generation to agentic problem-solving
In 2024, AI agents are no longer a niche interest. Companies across industries are getting more serious about incorporating agents into their workflows - from automating mundane tasks, to assisting with data analysis or writing code. we are experiencing a new wave in AI, one defined by intelligent agents that can handle complex tasks.
And if you were not convinced by the words here is some data:
Generative AI is transitioning from a future technology to a fundamental business tool. AI spend chart by Menlo Ventures based on a survey of 600 U.S. IT decision-makers at enterprises conducted between Sep 24 - Oct 24:
ok now with this preamble I think its time to go deeper?
BigTech is moving towards vertical integration of chips + cloud + models it seems. Now with Amazon announcements at AWS re:Invent 2024 its becoming clear they all want the land grab.
Enterprise AI Scaling Up
Enterprises are actively embracing this momentum, pouring a staggering $4.6 billion into generative AI applications—nearly an eightfold surge from the $600 million spent in 2023. Based on a survey of 600 U.S. IT decision-makers at enterprises conducted by Menlo Ventures between Sep 24 - Oct 24
AI is becoming central to business operations, with significant revenue gains reported in supply chain management, marketing, and HR. However, issues like inaccuracies and cybersecurity risks remain prominent
Based on our own experience with B2B customer discussions at gAI.ventures in the US/Americas/EU/ME we concur with most of the findings of Menlo Ventures report. What’s striking about generative AI adoption today isn’t just the scale—it’s the scope. This year, generative AI budgets flowed to every department.
Focus on Specific Tasks: There are a lot more people whose emphasis right now is on building AI systems that excel at specific tasks rather than pursuing general artificial intelligence (AGI).
Source: Deloitte survey (May/June 2024 ) N (Total) = 2,770)
In 2024, AI agents are no longer a niche interest. Companies across industries are getting more serious about incorporating agents into their workflows - from automating mundane tasks, to assisting with data analysis or writing code.
Langchain surveyed over 1,300 professionals — from engineers and product managers to business leaders and executives — to uncover the state of AI agents. About 51% of respondents are using agents in production today. When we looked at the data by company size, mid-sized companies (100 - 2000 employees) were the most aggressive with putting agents in production (at 63%). Encouragingly, 78% have active plans to implement agents into production soon. While it’s clear that the appetite for AI agents is strong, the actual production deployment still remains a hurdle to many.
Data as a Competitive Advantage: Data is crucial for effective AI implementation. Companies with a data advantage are well-positioned to leverage AI.
Example - if I am a AI customer service platform company which develops customer chatbots for level 1, level 2, and perhaps even level 3 support, I could be in demand today. But if I am like Klarna which not only has tens of thousands of hours of customer interactions and 100s of millions of data points of back and forth bw customers and call centers + also strong willingness to adopt GenAI, I can be dangerous.
The Application Layer: Where Startups Have the Edge in AI
The application layer is rapidly emerging as the sweet spot for startups in the AI ecosystem. Unlike the infrastructure and foundational model layers, which are resource-intensive and dominated by giants like OpenAI and Google, the application layer offers startups an opportunity to innovate and build niche solutions.
In 2024, the spotlight was firmly on the application layer of generative AI. With foundational architectural patterns already established, application-layer companies have been harnessing LLM capabilities across various domains to drive new efficiencies and innovation.
Historically, “wrappers” around existing platforms were seen as shallow plays. However, the dynamics have shifted. Today, innovative applications built on top of generative AI models are delivering substantial value by tackling specific use cases. According to the Langchain survey, The top use cases for agents include performing research and summarization (58%), followed by streamlining tasks for personal productivity or assistance (53.5%). Efficiency gains aren’t limited to the individual. Customer service (45.8%) is another prime area for agent use cases, helping companies handle inquiries, troubleshoot, and speed up customer response times across teams. For example: FastTrackr.ai leverages generative AI for automated email drafting, scheduling meetings, notes, saving hours of manual effort in corporate workflows. Harvey.AI focuses on legal tech, offering tools that streamline processes for law firms. These examples show that startups can create meaningful value by refining and customizing AI for targeted business problems.
According to a Deloitte survey, organizations are getting very particular about measuring the outcomes from AI investments and projects.
Different Approaches at Play - Startups are taking varied paths to excel in this layer:
1. Prompt Engineering-Based Systems:
Some focus on mastering the interaction between users and AI through advanced prompt engineering. This approach enhances the precision and relevance of outputs, making it valuable in dynamic sectors like customer service and content creation.
2. Domain-Specific Models:
Others are building models tailored for industries like healthcare, finance, or e-commerce. These models reduce the risks of inaccuracies associated with general-purpose LLMs. For example:
• Erica by Bank of America delivers personalized financial advice, showcasing the power of a domain-specific solution
• NVIDIA’s synthetic medical data models help researchers overcome real-world data limitations.
By focusing on specialized applications, startups can sidestep the competition in foundational AI layers and build high-value, defensible niches. The key is to understand the pain points of specific industries and create AI tools that directly address them.
Data Lakes as the Foundation for Enterprise AI - Often scattered across departmental silos and legacy systems in enteprises, data can be difficult to consolidate. Without integration into a unified data lake, scaling AI across the organization becomes nearly impossible. To deploy AI agents effectively, enterprises need to consolidate scattered data into a unified data lake thereby putting it into a cohesive, accessible format.
Even pioneers like Klarna took significant time to get this right. The company understood that AI applications are only as effective as the data feeding them. This lesson serves as a critical reminder: for anyone building AI solutions for corporates, start by understanding their data infrastructure.
Open Source - The dominance of proprietary models is being challenged as open-source models close the performance gap. Developers are leveraging frameworks like LoRA and quantization to fine-tune smaller, more efficient models that can operate locally, lowering infrastructure costs
The Rise of AI-Delivered Services: A Big Opportunity
Call it Results as a service (RaaS by Hubspot founder), outcome as a service, Infosys 2.0 or AI delivered services. Its becoming increasingly clear that AI-delivered services represents a massive opportunity. why? because it can scale and provide much better margins than IT services. Also because during this initial phase there is a lot of customisation and implementation work required. During my recent interactions in the US-India Corridor, many professionals shared the belief that this sector could evolve into a transformative force, potentially creating an “Infy 2.0” for the AI era. Even VCs are taking notice. For example — Insight Partners recently invested $10M into Sirius AI, signalling confidence in the scalability of AI-driven solutions. Hyperplane.ai, another key player, was acquired by Nubank, a move that highlights the growing demand for AI service innovations in the fintech space. It’s an exciting time for the space, with both startups and investors aligning toward this vision.
List of notable AI partnerships in 2024:
Government and Defense-Focused Collaborations:
1. Anthropic, Palantir, and AWS: Anthropic’s Claude AI models integrated into Palantir’s AI Platform (AIP) for U.S. intelligence and defense agencies, hosted via AWS
2. OpenAI and Anduril: Partnership to deploy AI technology in defense systems, focusing on responsible use to protect military personnel
3. Anduril and Palantir: Joint efforts to develop AI applications for military and defense, leveraging each other’s expertise
4. Meta with Anduril, Palantir, and Others: Deployment of open-source Llama 2 models for U.S. national security through partnerships with multiple contractors
Enterprise AI Integrations:
5. Microsoft and Adobe: Integration of Adobe Experience Cloud with Microsoft Copilot to assist marketers with automated workflows
6. Amazon and Anthropic: A $4 billion investment by Amazon, with Anthropic utilizing AWS and Amazon’s custom AI chips for its workloads
7. Oracle, Cohere, and Meta: Integration of Cohere and Meta’s Llama 2 models into Oracle Cloud Infrastructure (OCI) for generative AI services
8. Hewlett Packard Enterprise (HPE) and Nvidia: Collaboration on AI/ML software stacks and products leveraging Nvidia’s Blackwell platform
Specialized AI Applications:
9. Google and Nvidia: Joint efforts to advance hardware for AI workloads, combining Nvidia’s GPUs with Google’s AI frameworks
10. IBM Watsonx and CrushBank: Integration to enhance IT service management, providing AI-powered knowledge management systems
11. DataRobot and AWS: Tools for monitoring generative AI use cases, focusing on cost tracking, data drift, and model accuracy
Regulatory Evolution
2024 saw a fragmented yet accelerating regulatory landscape. The EU’s AI Act and U.S. executive orders have begun shaping AI governance, balancing innovation with caution. In China, enforcement of AI laws and domestic hardware efforts intensified. India seems to be taking a "pro-innovation" approach to AI regulation, balancing potential and risks
Pricing Evolution
I think everyone building AI agents is experimenting with the pricing. One transformative idea in the application layer is the shift from traditional seat-based software pricing to outcome-based models. As AI increasingly performs work directly, businesses will be paying for results rather than licenses. For instance: AI tools that optimize marketing campaigns could charge a percentage of revenue uplift. Compliance solutions, like Synthea AI, may price services based on audits completed rather than users onboarded. At FastTrackr.ai we are still charging per user per month as its a productivity tool.
There are various models being tried as covered in this tweet:
Future Trends and what to expect in 2025:
There are two very positive views about the future of AI:
1. The disruptive view believes that AI marks the "end of software/SaaS” as AI democratizes software development and creation costs go down. This numerically is a $500 billion market today.
2. The growth view believes that AI agents can expand software's addressable market ~10x. AI will compete for labour budgets which are much larger than SW budgets. This numerically is a 5 trillion dollar market today
Whether its 1 or 2, no one might be able to tell you today but its a great time to build.
Challenges to expect in 2025 - Reid Hoffman said to NYT that there are concerns that we might “run out of data” to achieve the next level of needed scale. “I would like to see a concerted effort by all stakeholders on licensing schemes and creative solutions. It’s essential that we engender a shared belief that “using data is not theft but creates a public good.” Making more data available to improve A.I. will yield better tools for society, the economy and the individual.”
Autonomous agents - we are going to experience a new wave in AI, one defined by intelligent agents that can autonomously handle complex tasks.
Open Source’s Growing Role - The dominance of proprietary models may start fading as open-source models close the performance and usability gap.
Federation of Models: AI systems will increasingly rely on a combination of specialised models to achieve complex tasks.
Compound AI Systems: These systems will consist of multiple components working together, allowing for better debugging and improvement.
Breakout Applications: Expect a viral AI-powered app or game, likely created by non-coders, to dominate headlines. This will signal the democratization of AI development
Localized AI Models: The rise of smaller, domain-specific models (like small language models) will grow as businesses prioritize data privacy and operational efficiency. This trend will particularly impact industries like healthcare and finance
Safety and Governance Maturity: As risks around misuse and inaccuracies become more apparent, companies will invest more in AI safety frameworks, explainability, and governance boards
Tracing and human oversight are needed to keep agents in check
With great power comes great responsibility — or at least the need for some brakes and controls for your agent. Tracing and observability tools top the list of must-have controls, helping developers get visibility into agent behaviors and performance. Most companies are also employing guardrails to keep agents from veering off course.
The Need for Ethical Considerations: As AI becomes more powerful, it's crucial to address ethical implications and ensure responsible development and use.
The Importance of Human-AI Collaboration: This collaboration between humans and A.I. agents, which combines human creativity and critical thinking with A.I.’s precision and scalability, will form the cornerstone of the future workplace. Effective AI implementation requires a strong partnership between humans and machines. Humans will continue to play a vital role in AI development and deployment, especially in areas like evaluation, debugging, and setting ethical guidelines.
Final words:
AI is transitioning from experimentation to execution, reshaping industries at scale. 2025 will likely bring more practical applications, broader adoption of open-source frameworks, and refined regulations to ensure responsible innovation.
Although companies begin to make the journey from model to product, long-term questions around pricing and sustainability remain unresolved. While there's immense potential in AI, it's important to be realistic about its limitations and the challenges involved in scaling and deploying AI systems.
Ending with a quote "Generative AI is the key to solving some of the world's biggest problems, such as climate change, poverty, and disease. It has the potential to make the world a better place for everyone." - Mark Zuckerberg
Let me know what you think ? Linkedin / Twitter
Sources:
https://menlovc.com/2024-the-state-of-generative-ai-in-the-enterprise/
https://www.langchain.com/stateofaiagents
https://www.mescomputing.com/news/4191299/epic-big-tech-genai-collaborations-2024
https://x.com/amitTwitr
https://gai.ventures/
https://fasttrackr.ai/
https://www.stateof.ai/
New York Times