‘Global firms are adopting large AI models to cut costs’

Enterprises globally are adopting Large Action Models (LAMs) that understand complex goals communicated with natural language, and they follow up with autonomous actions to achieve them, said Adnan Masood, Chief AI Architect at UST, a California-based digital technology services provider.

LAMs are advanced AI models. They go beyond generating and interpreting text like ChatGPT, and can perform tasks interacting with applications and devices in the real world. Mr. Masood said, ‘‘They can autonomously handle complex tasks like end-to-end vacation planning, job application automation, investment portfolio optimisation, and personalised social media content creation, all while continuously learning and adapting to user preferences.”

Cutting labour costs

This has helped cut significant labour costs at U.S. insurance firms by automating claims processing with LAMs, Mr. Masood added. He also pointed to a major European airline that uses LAMs to enhance customer interactions from bookings to loyalty programmes; and in Asia, retailers boost sales conversions by 25% with personalised recommendations powered by LAMs, he said.

Also, extremely powerful language models can now deeply understand intent from multimodal (speech, video, text) inputs and neuro-symbolic AI to navigate user interfaces like a human would. ‘‘Mix this with the proliferation of APIs (application programming interfaces) and IoT devices, demand for automation, increased computing power – all of this makes for a perfect environment for this next generation of AI agents,’‘ he explained.

Early examples like the Rabbit’s LAM-powered device showcased CES (the annual Consumer Electronic Show held at Las Vegas) are quite promising where a single LAM-powered assistant could handle a wide variety of tasks seamlessly. ‘‘I believe companies will likely develop their own proprietary LAMs or purchase access from LAM service providers. LAMs could become as integral and commonplace for businesses as SAAS applications or hyperscaler cloud enablement in the coming years,’‘ he reckoned.

Potential job loss

Responding to a query on whether LAMs become a threat to employment as they seem dynamic, independent and intelligent, he agreed that LAMs would likely automate many knowledge work tasks currently done by humans, so in that sense, they may displace some roles. However, like other things with AI, they are likely to create more jobs than they replace by enabling new capabilities and allowing humans to focus on higher-level, creative tasks.

Emerging roles will support LAMs, including training, oversight, and explaining model actions. The development of LAMs must emphasise transparency, fairness, and security, with stringent testing for biases and misuse and clear accountability. The shift in work types would necessitate proactive reskilling and job transition planning, and government policies to ensure upskilling, Mr. Masood added.