Accenture Invests in Martian to Bring Dynamic Routing of Large Language Queries and More Effective AI Systems to Clients

Accenture has invested, through Accenture Ventures, in Martian, a technology company that has built a patent-pending Large Language Model (LLM) router. For any prompt, Martian’s router will find and use the LLM that will give users the best result at the lowest cost. As part of this investment, Accenture plans to integrate Martian’s advanced model router offering into its proprietary “switchboard” services that can customize large-language models (LLMs) for specific data sources and use cases for greater impact of AI systems.

Also Read: AI and Big Data Governance: Challenges and Top Benefits

“The capabilities of Accenture’s switchboard services and Martian’s dynamic LLM routing simplify the user experience and will allow enterprises to experiment with generative AI and LLMs in order to find the perfect fit for their business needs.

The rapid expansion of AI systems and large language models has created a complex landscape where each model presents unique tradeoffs in performance, cost, and speed. Martian’s model router addresses these challenges by searching across a vast number of LLMs in order to dynamically route prompts to the model that will yield the highest quality results based on specific performance and technical requirements for each enterprise. The router also improves reliability; prompts are automatically rerouted to other providers when a model or provider becomes unavailable, significantly reducing downtime.

“Many of our clients are looking to reap the benefits of generative AI in a responsible way that takes into account customized requirements, performance and cost,” said Lan Guan, chief AI officer at Accenture. “The capabilities of Accenture’s switchboard services and Martian’s dynamic LLM routing simplify the user experience and will allow enterprises to experiment with generative AI and LLMs in order to find the perfect fit for their business needs.”

Launched in November 2023, Accenture’s switchboard service allows clients to test, evaluate, deploy, and control the most potent combination of LLMs for their needs. Integrating Martian’s dynamic router capabilities, clients will be able to open up access to AI systems with customized models tailored to unique prompts and requirements, ensuring optimal performance and cost-efficiency. New LLMs will be automatically added in and analyzed as they become available, helping ensure that outputs will be accurate and current. Switchboard services can be applied to emerging technologies such as quantum as well.

Also Read: What is Return on AI – and How Do Companies Measure It

“We’re always exploring ways to accelerate our leadership in applying generative AI technology to create and protect growth and relevance for clients, with creativity at the core,” said Dan Garrison, chief technology officer for Accenture Song. “We believe Martian’s novel approach and capabilities can help our clients and people build dynamic AI systems fit for today and tomorrow.”

“The vast majority of enterprises building Gen AI applications are in early stages and finding that output accuracy is often too low to justify going into production. By integrating our technology with Accenture’s switchboard services, we can substantially boost accuracy so enterprises see quality results that exceed production threshold. Accenture’s technology and expertise will be critical as Martian continues to grow and simplify large language model management for enterprises,” said Shriyash Upadhyay, co-founder of Martian.

Martian is the latest company to join Accenture Ventures’ Project Spotlight, an engagement and investment program focused on working with companies that create or apply disruptive enterprise technologies. Project Spotlight brings promising startups into Accenture’s innovation network, giving them exposure to real business problems we’re solving for our clients and helping them tailor, refine and scale their solutions for maximum impact.

[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]