Senthil Ramani, global lead – data & AI in Accenture, said, “AI has been there for decades. GenAI has become a catalyst for people to understand this technology. I don’t think there is any other technology like genAI which is more human-like. GenAI has cracked the code on language and others have cracked the code on numbers. So, you put numbers and language together, you can run any enterprise,” he said. Last year, Ramani joined Accenture’s strategic decision-making body called global management committee (GMC).
Accenture has defined the R12 talent model that encompasses primary and proximity skills and aligned specialized learning paths for each of these 12 roles. Some of the roles include full-stack large language model (LLM) development that entails developing LLM-based applications, fine-tuning LLMs to improve performance, accuracy, and alignment of the general-purpose models. It also includes developing and integrating LLMs and/or LLM applications into products and services.
Another role is on visualization and business intelligence (BI) engineering focuses on developing information that drives adoption, decision-making, and behaviour change, using storytelling to ensure adoption in business users.
To grow its data and AI talent pool, Accenture is doing it in three ways. “One is the rotation, second is re-skilling and third is rotating effectively and there is a subtle difference from the first one. First, we identify people who have strong cloud skills and if you understand those better, you start to understand the newer tech even better,” Ramani said. He also said how architectural skills are going to become critical in the future. “In most of the use cases, genAI coexists with classical AI — this is coexistence of architectures. So, people need to understand how to architect this coexistence,” he said.
Ramani alluded to the fact that how responsible AI delivery for an organization begins with data and it requires a holistic approach factoring in bias, fairness, transparency, and accuracy to create an effective framework. Accenture’s research showed that 9% of the 1,500 companies have built the capability for continuous reinvention. These leaders, defined as reinventors, have been weathering disruption pulling away from less-prepared competitors. The report projects the gap in revenue growth between this set of leaders and all other companies will increase 2.4 times from 2022 to 2026. Industry giants are leading the charge — the number of reinventors in this group, defined as having revenues more than $50 billion, has more than quadrupled in the last year (growing from 4% to 18%).
Ramani said that just 5% of organizations are spending on reskilling. Take the banking industry where mainframe skills are still in vogue, for instance. “Taking the person along with the technology is very important because they understand what happened in the past and they are the only ones who can validate whether the target state was right or wrong. They have the institutional knowledge to do that.” Accenture has also partnered with UC Berkeley for the Accenture-Berkeley Foundation Model Architect Program that will focus on the technical aspects of foundation models and how they are affecting the development of AI.
When asked about a greater emphasis on LLMs than SLMs, Ramani said, “I think there has been an over-indexing on LLMs and an under-indexing on SLMs (small language models). If you start with SLMs, we have been tracking over 1,020 models within our model garden.” To help customers unlock the power of genAI, Accenture has launched a Switchboard that will help companies customise and manage foundation models. “It’s a functionality which helps you understand routing and circuitry of models within your client environment as you have the choice.” For example, one of the largest entertainment companies is testing the Switchboard to compare how the same prompt would be interpreted by different models and how they perform before deciding on which one to use.