It’s been a year since AI agents burst onto the tech scene, driven largely by the vision and product roadmap of Salesforce founder Marc Benioff, who is betting the company’s future on the “agentic enterprise”. Saleforce’s Agentforce platform, which allows customers to build and deploy agents in days or weeks, automating aspects of sales, customer service, marketing, and even software development, has racked up 12,000 customers in its first year. But Benioff said last week in San Francisco that there’s a gap between the innovation AI companies are progressing, and the business sector’s ability to adopt AI at scale. The halls of the Moscone Centre in San Francisco were full last week of early adopters, companies that have deployed AI agents with promising results, from PepsiCo, to Lululemon. Among them was One NZ, the telco that has spent years modernising its tech platforms and which uses Salesforce for its customer relationship management.  "Already across our AI investment, five times return," One NZ CEO Jason Paris told me on this week’s episode of The Business of Tech, describing how One NZ deployed an agentic tool to help prepaid customers migrate to better mobile plans.  “We've been deploying AI and all the variations of it, so robotic process automation or machine learning, for over 10 years. We've only been deploying agentic for last year, because no one even really knew about it 12 months ago,” he said.  “But yeah, five times ROI on our AI investment. So a lot of people are saying, is the value there? We can see the value. Absolutely. And we think this is just the beginning. We want more.” Orchestrating multiple agents The agent authenticates customers, presents personalised plan options, explains trade-offs, and completes the entire transaction autonomously via chatbot – all in a matter of minutes.​ Previously, a customer may have needed to make a phone call to the One NZ contact centre or visit a store to achieve the same outcome. Paris added that the newly deployed Agentforce agent has delivered a 400% improvement in customer engagement compared to traditional digital journeys.​ "It took us five weeks to build and deploy," Paris explained, noting that the first agent One NZ built with Salesforce took just eight hours to create, though two additional weeks were needed for data integration and cleanup.​ Sitting behind that simple chatbot designed to help prepaid mobile customers find a new plan, are actually seven AI agents working together.  "We're in this age now where this one agent can orchestrate multiple actions,” said Hamish Miles, Salesforce New Zealand’s managing director.   “In [One NZ’s case] it's like… a security check in for the trust layer. It's, have we got the right plan? Can I make some recommendations? We can do so much more as well." Miles acknowledged that addressing Benioff's innovation-adoption gap requires overcoming hesitation. "I think there's probably a little bit of reluctance to make a start," he said, noting that companies often worry about data quality. His advice is to at least get going on creating AI agents. "Make a start, because it's a low-risk entry. Is the agent going to be perfect straight away? No, it won't be. But will it learn? Yes, it will. Can we make corrections very quickly? Yes, it will. So start experimenting," he said. Salesforce itself now runs more than 200 internal agents managing 550 different tasks.​ Mark Benioff says use of AI agents is saving the company $100 million annually, much of it in customer support costs, including reduced headcount. Voice - the next frontier in AI agents The next frontier for AI agents is voice. Paris described the upcoming product Agentforce Voice, which enables natural conversations with AI agents over the phone, as "a game changer" and "probably, of all the things that were discussed, personally, [what] I'm most excited about".  With up to 80% of customer service interactions in some industries still coming via phone, voice-enabled agents could fundamentally transform customer experience.​ "If I can have a conversation where there's no latency, it's a humanised conversation that you can build rapport through transparency that you're talking to an agentic tool, it's probably an easier step for me to just talk to an AI agent in the same way I would talk to an agent in a call centre," Paris explained.​ As New Zealand's economy continues to face headwinds, the ability to simultaneously reduce costs and increase productivity through agentic AI presents a compelling opportunity, but many businesses still need to get their data and IT infrastructure in place to deploy AI agents effectively. AI expertise is also thin in our organisations.  "Leadership needs to come from the top,” said Miles.  “If you look at any successful transformation project over the last 10 to 20 years in technology, the one core ingredient that it got right was governance and leadership, and it came from the top." Listen to the full episode of The Business of Tech, powered by 2degrees Business, streaming on iHeartRadio or wherever you get your podcasts. Your weekly tech reading list NZ might have gone cloud-first, but public sector IT is still stuck on the ground: report - BusinessDesk The tech is real. The hype could ruin it - BusinessDesk Canva cash and human capital sprinkling Kiwi startups - BusinessDesk London Became a Global Hub for Phone Theft. Now We Know Why - Slashdot Please stop putting Peter's phone number on your accounts - Information Age ‘I left Microsoft to get my hands dirty’: Telstra’s $1.6b bid for AI riches - Australian Financial Review The Web Is Becoming AI’s Interface Layer - Richard MacManus  OpenAI’s AI-powered browser, ChatGPT Atlas, is here - The Verge Amazon Plans to Replace More Than Half a Million Jobs With Robots - New York Times Netflix goes ‘all in’ on generative AI as entertainment industry remains divided - Tech Crunch The podcasts in the SME Stream contain general information only, not financial or professional advice. Any opinions expressed in the podcasts are not necessarily shared by BNZ, or its related entities. BNZ is not liable for any losses resulting from the content of the podcasts. See omnystudio.com/listener for privacy information.

Ver todos los episodios