Article

AI is poised to reshape industries and redefine possibilities in enterprise software

George Mathew, Michael Spiro | January 24, 2025| 2 min. read

This article originally appeared in TheMarker

Artificial intelligence is evolving rapidly. Before 2017, AI was primarily focused on prediction tasks like classification and recommendation. In 2017, Google’s Attention Is All You Need paper introduced the transformer model. This breakthrough model architecture has driven a transformative era of generative AI, with AI now capable of creating new content, solving complex tasks/problems, and reasoning.

Insight Partners has deployed nearly $4 billion into more than 75 AI/ML companies since our founding almost 30 years ago, investing in AI long before the generative AI era and the release of ChatGPT. This history of investing in and partnering with AI businesses gives us insights into the challenges, and the opportunities, in AI.

The state of AI

Technological and research breakthroughs, paired with decreasing training costs thanks to specialized AI chips, have enabled the scope and possibilities of AI to expand drastically over time.

The evolution of AI can be represented across four phases:

  1. Early Analytical AI (Pre-2017): Early AI systems (primarily used for classification and recommendation) had limited data and high computational costs, making them accessible to only a handful of enterprises and academic institutions.
  2. Improved Accessibility (2017-2022): The transformer architecture democratized AI by making it faster, cheaper, and more accessible to train and deploy AI at scale.
  3. The Generative AI Boom (2022-2023): The release of ChatGPT on November 30, 2022 (just over two years ago!) showed the public what was possible with generative AI. Quickly, a flurry of founders arose building generative AI applications and platforms, and consumers and enterprises alike grew in excitement to try new generative AI tools.
  4. The Current Era of AI (2023 – Today): AI applications are tackling a host of consumer and enterprise challenges, from content creation to decision support to complex enterprise workflow automation, driving a new wave of creative and productivity-enhancing applications. Enterprises are seeing massive changes in productivity and how work is done, powered by both horizontal and task-specific AI applications.

The opportunity in GenAI

The adoption of generative AI is leading many enterprises to rethink their application architecture. Traditional databases are evolving into vector and graph databases, enabling AI systems to retrieve and process information more effectively. Workflows like retrieval-augmented generation (RAG) combine generative models with structured knowledge, making AI systems more context-aware and capable of reasoning.

At Insight, we are proud to partner with several companies building generative AI applications and the infrastructure and tooling enabling generative AI’s adoption and deployment at scale. Companies like Weights & Biases, Atlan, Fiddler, and CrewAI are helping developers, ML researchers, and organizations at-large build, deploy, monitor, and scale AI applications across their organization and customer base. Companies like Writer, HourOne, Jasper, Swimm, Relevance, and Sourcegraph are building applications with AI at their core, helping drive enhanced outcomes and productivity for end users and customers.

Insight has also been an early partner to successful Israeli AI companies like Deci AI and Run:AI, both of whom were acquired by NVIDIA this year. Israel is an incredible source of innovation and continues to be a successful tech hub through a very challenging time, with founders and entrepreneurs exhibiting resourcefulness and desire to find product-market-fit and solve customer needs.

Challenges on the horizon

Despite its promise, generative AI does not come without a number of challenges and potential limitations. Transformer models, for example, struggle with hallucinations, limited context windows, high computational costs, and potential limits to scaling. Emerging innovations, such as small language models (SLMs) and mixture of experts (MoE) architectures, aim to address these limitations, making AI systems more efficient and accurate.

The deployment of advanced AI systems requires a strong foundation in responsible AI practices. Frameworks like the NIST AI Risk Management Framework and the EU AI Act emphasize transparency, fairness, and safety. These principles are vital to scaling AI systems responsibly while mitigating risks such as bias and misinformation.

What is next?

We are in the early innings of the AI revolution and are excited for the next wave of AI application companies to arise, who embed AI at their core, seamlessly integrating reasoning capabilities into enterprise workflows and leveraging data flywheels to continuously improve the performance and reliability of AI systems.

Artificial intelligence and generative AI are catalysts for innovation, efficiency, creativity, and fundamentally new possibilities. While challenges remain, the opportunities are large, and AI is poised to reshape industries and redefine what is possible in the years to come. Enterprises that adopt responsible AI practices and embrace next-gen architectures will be best positioned to harness the full potential of AI.

 

Insight Partners has invested in Weights & Biases, Atlan, Fiddler, CrewAI, Writer, HourOne, Jasper, Swimm, Relevance, Sourcegraph, Deci AI, and Run:AI.