Share

cover art for How Salesforce are using LLMs with Greg Bennett

VUX World

How Salesforce are using LLMs with Greg Bennett

•

Greg Bennett’s the final Conversations2 interviewee focused on large language models!


When I heard that Salesforce have adopted LLMs at scale (and also heard that they had started to design and build this prior to ChatGPT’s release) I wanted to know more!


I’ve known Greg for years and have always found him to be warm, knowledgeable and an expert at making complex subjects understandable. He didn’t disappoint here - we spoke about thinking in systems and how that relates to conversational AI, how LLMs affect conversation design (and how conversation designers are key to the best usage of LLMs), how his team of designers work with the wider team, and Salesforce’s recent push into Generative Conversational AI.


So listen up to Greg’s thoughts on the best ways to use LLMs in conversational AI! Salesforce have gone all-in on generative conversational AI and Greg’s team have made that possible.


00:00 Start

00:29 Greg's introduction

03:55 What's your favourite bot?

05:11 What's the most useful thing you use at work?

06:21 Which conversation designer should we all know about?

08:15 Conversations are systematic, and design should be too

10:17 What does 'generative conversational AI' mean to you?

11:38 How have you been incoporating generative conversational AI into your work at Salesforce?

16:05 We need expertise when writing prompts for LLMs

17:45 How does your team collaborate with the wider Salesforce team?

20:50 What benefits did you envision and how close are you now?

24:23 How has this affected the day to day work of conversation designers?

27:46 Now we need to be able to educate people about how to do conversational AI well

29:52 We're thinking conceptually about language

32:43 Is this a new phase of conversation design?

36:32 How do you minimise the risks?

39:49 Do you continue to iterate together with engineering?

41:37 The model has a risk of learning from bad behaviour

42:46 How do you stay up to date with generative AI?

44:05 Researching attempts to break LLMs on social media gives great REAL data

45:49 When will you next speak live?

46:46 Outro


More episodes

View all episodes

  • CSAT and NPS are dead. Good Riddance. With Richard Smullen

     01:09:01
    Richard Smullen is the CEO of Pypestream, an AI automation platform making waves with its approach to design and operationalising conversational experiences. Tune in to find out why CSAT and NPS are dead, why happiness should be your true north KPI, the future of the contact centre and what we’ll do with all the jobless call centre workers **grits teeth**. Learn the 4 missing pieces of your AI strategy and why you’re struggling to fill them.Myself and Richard share so many similar perspectives on how to do AI right in the enterprise. This will be a bone shattering journey of cognitive dissonance if you’re not open to change. Be open. Dive in.1

  • How we built AI assistant, Sandy, with loveholidays

     01:08:39
    The loveholidays AI team join us to share their lessons learned from implementing the award-winning AI assistant, Sandy.Eugene Neale, Director of Business IT and CX Engineering, and Tom Smith, Telecommunications Operations Manager at loveholidays, the UK's fastest growing and largest online travel agent, share use cases, lessons learned and best practices from years spent designing and building Sandy with Google DialogFlow.Subscribe to https://vux.world/subscribe for weekly updates on how to implement AI effectively in your organisation.Unparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here - https://www.unparsedconf.com/

  • The role of AI in music with SoundCloud's Hazel Savage

     59:14
    Hazel Savage is the VP Music Intelligence at SoundCloud and the founder of Musiio; a tech company using AI to make music easier to catalogue and find.Hazel joins us to discuss the role of AI in music, the opportunities, challenges and future.Subscribe to https://vux.world/subscribe for weekly updates on how to implement AI effectively in your organisation.Unparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here - https://www.unparsedconf.com/

  • Unlocking AI use cases with speech recognition with Ricardo Herreros-Symons

     01:00:54
    Ricardo Herreros Symons is the VP Corporate Development and Strategy at AI speech to text company, Speechmatics.

  • New updates from OpenAI: a rundown of GPT-4o and it's impact on CX

     01:02:00
    OpenAI GPT4o: What's new and what's the impact on CX? Today I'm joined by Marc Bernstein, CEO, Balto, to discuss the new updates from OpenAI, GPT-4o, and what it means for the future of CX.

  • Chatbot and LLM Analytics with Eric Griffing

     57:52
    From the granular details of measuring intent and LLM-based chatbot effectiveness to the strategic implementation of AI across business units, Eric provides an in-depth analysis on the importance of data-driven decisions in the era of AI transformation.Join us as we discuss how businesses can leverage analytics to optimise customer interactions, the common pitfalls in current AI strategies, and the future of AI in customer service.In this episode00:00 Introduction and UNPARSED conference - Find out more at https://unparsedconf.com4:32 Welcom Eric Griffing7:25 Best practice for measuring LLM agents12:10 Differences tracking ML and LLM bots14:00 Goal completion19:00 What's missing22:00 How to set up tracking on chatbots24:50 Tracking LLM chatbots30:37 Air Canada's chatbot issues33:38 LLM guardrails37:00 Using data to inform your AI strategy41:00 Effort over satisfaction44:30 Tracking customer effort in chatbots48:40 Market demand for AI analytics55:40 Wrap upSubscribe to https://vux.world/subscribe for weekly updates on how to implement AI effectively in your organisation.Learn more from Dashbot:https://www.dashbot.io/blog/360-bot-frameworkhttps://www.dashbot.io/blog/inside-the-black-box-analyzing-your-llm-bothttps://www.dashbot.io/blog/navigating-the-shift-from-intent-based-to-language-model-chatbotsUnparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here

  • AI that filters voices and soften accents, with Ofer Ronen

     52:00
    Ofer Ronen is an ex-Googler with some immense experience in deploying AI systems for large enterprises including the infamous example of Agent Assist at Verizon. Ofer’s third startup is tomato.ai, a company that has AI technology that’ll strip away your accent! Or soften it, rather.Ofer joins us to share his lessons learned from years of deploying AI in the enterprise, as well as the use cases, value and technicalities of this new breed of AI technology: voice filtering and accent softening.Unparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here - https://www.unparsedconf.com/

  • Lessons from 21 years in conversational AI with Chris Ezekiel

     01:10:58
    Chris Ezekiel is founder and CEO of one of the longest standing conversational AI platforms, Creative Virtual. Chris joins us to discuss the changing landscape of conversational AI and the recipe for longevity and long term success.Unparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here - https://www.unparsedconf.com/

  • Using AI to make the job search easier with Total Jobs Group

     01:04:35
    How can AI make your job search easier? I'm joined by Head of Product - Conversations at Total Jobs Group, Somnath Biswas to discuss. Som has immense experience in project management and AI and knows what it takes to make AI work at scale. He joins me to share some lessons in implementing conversational AI in a highly stressful and emotional situation: the job search.Do customers want this? What conversations do they want and need? How can you make sure you can serve those needs at scale? What people, process and technology do you need to make it happen? What's the role of generative AI and large language models? And what does the future of job hunting look like?Let's dive in!00:00 Introduction 04:27 Welcome Somnath 09:25 Organised groups11:50 Conversational interfaces17:45 Measuring lifetime value20:15 Next steps after conversational formats28:00 Making improvements36:00 Model producing responses38:55 Multi agent46:40 Augmented retrieval 52:22 User / response input stage 58:50 Different prompt criteria1:04:40 OutroUnparsed, the world's most loved conversational AI conference returns with more punch, more passion and more impact! Buy your tickets now here - https://www.unparsedconf.com/