‘Narrower’ models sought for generative AI platforms
By Miguel Hanz L. Antivola, Reporter
Technology developers must build well-defined large language models (LLMs) for artificial intelligence (AI) to optimize business processes responsibly, according to global professional services company Accenture.
“It’s how you set it up for primary function,” Arvin Yason, innovation lead at Accenture Technology in the Philippines, said during a briefing on Wednesday.
“The narrower you set up the LLM, the more effective it is, also at preventing hallucinations,” he added on knowledge conflicts and gaps as avoidable circumstances in such rapidly evolving tech advancements.
However, seeking AI solutions must first identify business dimensions and functions it can benefit.
“Look at it in terms of your business use case and target outcome,” said Mike Lao, data and AI lead at Accenture Technology in the Philippines.
“What is the long term value for your employees and customers?” Mr. Yason added, noting the importance of human-centric design when introducing tech solutions in business.
In Accenture’s Technology Vision 2024 report, AI-curated responses through advice and a summation of results are expected to grow, where “searching now becomes synthesizing.”
“Business leaders who reimagine how information works in the organization and equip their people with AI-enabled enterprise knowledge tooling will realize exponential performance gains and competitive advantages,” the study said.
According to the Accenture Pulse of Change 2024 index, Asia Pacific companies are already moving toward customizing foundation models with proprietary data from proof of data.
It noted 77% of C-suite executives in the region planning an increase in AI-related spending this year.
The company is set to expand its network of generative AI studios in the Philippines as part of its $3 billion global investment announced last year, targeting industry-specific solutions and doubling of AI talent to 80,000.
“The studios will cater to a wide range of industry and functional needs, but each will also specialize in one or more industries including banking, insurance, telecommunications, public sector, manufacturing, renewable energy, chemicals, and mining,” Accenture said in a press statement.
The platforms, which were introduced during the briefing, include Companion, which enables utility and field service with optimized routes and chatbot inquiries, reducing manual inspection time.
It also allows field personnel to quickly assess damages via image input and generate a corresponding work order with bill of materials.
CareCoach, the training tool for tech support representatives, is able to build scenarios and simulate customer persona, even according to specifications like churn or level of difficulty.
It can be customized according to industry relevance, also providing personalized feedback and scores depending on the criteria set.
Moreover, Gen-E, the new innovation team member at the Accenture Philippines office, is an internal AI assistant prompt-engineered by the company’s own knowledge sources to simulate the viewpoint of an employee.
It can operate via text, image, and speech input, able to conduct tours in the office.
With such AI developments available to the market this year, Mr. Lao emphasized a close and continuous loop with companies leading with value, building its digital core, transforming work and talent, and fostering responsibility.
“It’s not something that is just technology-led,” he said. “It is making sure everyone is empowered for productivity and creativity in the workplace.”