Salesforce, Inc. (CRM) Earnings Call Transcript & Summary

Welcome, everybody, to the today's webinar on the Trust Imperative 4.0 GenAI: The Trust Multiplier for Government. I am delighted you're able to take some time out of your very busy day to join us today. I'm also delighted to say that I am joined by 2 very dear friends to Salesforce in BCG, Boston Consulting Group, who have been along us -- with us in this journey on the nexus between great government service delivery and increased levels of trust in government. I'm very pleased to present my co-presenters today. First off, we have Miguel Carrasco. He is a Managing Director and Senior Partner of BCG. He's also the global leader for BCG Center for Digital Government and BCG X for Public Sector. He has extensive experience working in government organizations globally and has deep expertise in a wide range of topics on citizen-centric digital service delivery, artificial intelligence, digital identity and legacy systems modernization. I'm very pleased he's also joined today by another dear friend of ours, Francisca Browne. Francisca is a specialist consultant at BCG Center for Digital Government. She has co-authored a number of BCG publications on GenAI in the public sector and has now been heavily involved in the Trust Imperative series as well as the BCG Digital Government Citizen Survey over a number of years. So thank you, Miguel and Fran, for joining us today. So as is also Salesforce way to note, it's very important for us all to recognize that at every point, our forward-looking statement is to be adhered to. Please make any buying decisions on the types of products that we have available in the market today rather than anything in the future. So let's get on with today's topic of Trust GenAI. Before we do that, a couple of housekeeping rules. Will this session be recorded? Yes, it will and will be in your inboxes straight after for your later perusal. We'd love you to join the conversation. In the sidebar, there is an opportunity for you to put in place any questions you might have along the way. And please let us know at the end of this webinar what you think. Any additional ideas and opportunities for further questions, discussions, we are all ears. So please, we want to make this as interactive as possible. Please put those questions in. We will get to them at the end. We'll do a presentation at the top and hopefully, have a really dynamic discussion at the end with Miguel and Fran. So here we are today with our Trust Imperative 4.0, which suggests that there has been 3 iterations of this study beforehand. First, we started with Trust Imperative back in 2020. And for those of you who might recall, 2020 was a very different world to the world that we're currently in today. And we were able to take a snapshot of citizen service delivery and trust in levels of government just before COVID hit our shores, which is an amazing moment for us to capture public sentiment and really understand the nexus between good government service delivery and increased levels of trust in government. We were able to find a direct correlation between good government service delivery and those increased levels of trust. In fact, we found that around 84% of Australia and New Zealand participants responded by saying that a positive digital experience actually increased their levels of trust in government. And we also found the opposite effect to be true as well. Drilling down on this finding and especially at the time when we were engaging with government for digital services at an unprecedented level, particularly because of the proximity issues posed by COVID, we then delved into this issue of what level of engagement do citizens actually expect from government? And what we found was actually a permission to personalize. We actually found that 84% of Australia and New Zealand does actually expect some level of proactivity from government and 73% expect some level of personalization. And that extends beyond just that initial e-mail that actually notes your name is that you are the actually intended recipient of a piece of correspondence and really understanding how data can be delivered to your circumstances and meeting you where you are as the citizen. Now of course, in order to be able to operationalize all of this, we looked at, well, how willing are people actually to share data with government. And not only that, how willing are they to allow government agencies to share that data between agencies as well in order to be able to deliver those personalized and proactive services. And we were overwhelmed with the response. In Trust 3.0, we were able to determine that around 91% of folks were actually willing for government to share the minimum amount of data in order to get that net tangible benefit of receiving personalized proactive government digital services. Enter the brand new brave world of generative AI, and of course, we went back out to market to determine exactly what people actually are comfortable with, with government experimenting with an efficiency tool like GenAI. And to take us through some of those findings today, I will hand over to Miguel very shortly. But one of the issues that we are talking about here is why bother at all. We know that an efficiency tool like GenAI is here to be able to deliver faster responses, more targeted digital government services and a clearer, more accurate form of communication with the end user. And in doing so turns the important flywheel of trust, because the citizens find that as they hand over data, they expect that data to be treated appropriately, with consent, in a transparent way and to deliver a net benefit to them as an individual or to their community. And that, in turn, gives them greater confidence to share data with government. And that turns that important flywheel of trust. Now to take us through our findings on how GenAI changes this equation, I hand over to my colleagues at BCG. Miguel, over to you.

Thanks, Gisele. And I might also call on my colleague, Fran to join me. So the latest iteration of the Trust Imperative is based on BCG's global digital government survey, which we run every 2 years. And it's the longest running, most comprehensive voice of the citizen survey in the world. And this year, it captured the views of over 40,000 people regular Internet users in 48 different jurisdictions. And that, just to give you a sense, represents nearly 3/4 of the total population. And 8 out of 10 people in OECD countries -- 9 out of 10 actually people in OECD countries. So it's a very comprehensive wide-ranging survey, looks at people's usage and satisfaction of the digital government services. And this year, in particular, looks into the question of AI and generative AI and citizens attitudes towards the use of AI by government. Fran, do you want to keep going?

Sure. Absolutely. Thanks, Miguel. So when we look at what the survey tells us and what our research says, the good news is that net satisfaction has improved over the past 2 years. So it will improve 7 percentage points in Australia and 9 percentage points in New Zealand. If you look at the data underneath this, net satisfaction actually improved across all of the Australian states and territories. And this really means a lot because we also know that citizens' expectations are really high. So 74% of ANZ respondents told us that they expect digital government services to match the world's best private sector organizations in the world and global digital leaders. They did -- we did also identify some service opportunities as well. So there's improvement opportunities as well. So 62% of ANZ respondents told us that they encountered problems using online government services in the past 2 years, the top 3 problems being long and difficult processes, remembering user names and passwords and technical issues. And to Gisele's point earlier, we know this is really important because from our previous research, citizens told us that their everyday interactions with government online has a direct impact in trust and confidence in government of the day. And in this respect, GenAI presents a significant opportunity for governments moving forward to address these issues and achieve a leap in service satisfaction as a tool that can make processes easier to navigate, services more accessible and personalized. And can even make use to make updating digital services cheaper, faster and easier. So I'm going to turn over to Miguel to walk us through what the citizens told us about how they feel about government using GenAI.

Great. So one of the things we wanted to test was the extent to which citizens are comfortable with a range of different use cases. And we gave some examples and asked them to say to what degree they felt comfortable with governments using GenAI in those instances. And we found actually a reasonably high degree of comfort with some use cases. So for example, using it to translate or communicate in multiple languages for customer staff using it as a support tool or using it in terms of internal administration and streamlining internal processes, documentation and so on or even things like chatbots and being able to use virtual assistants to access simple government services. But there were clearly some use cases where the level of comfort was a lot lower. And the 2 that stood out were being used to make decisions about access to services or assessment decisions and also any kind of surveillance or monitoring of public sentiment. So this is important when -- for governments in thinking about where can they and should they use GenAI. Obviously, understanding the social license and where there is acceptance is important. And we expect these to change over time. We know that there's a link between awareness and education and the people's perceptions around the level of comfort. So it does -- it provides a useful starting point for where might it be okay for government to start doing experiments and pilots in terms of using AI and GenAI. And in those areas where there's a high degree of comfort are probably a good place to start. The other opportunity, of course, is not just improving service quality, but efficiency. And another study that we did and published earlier this year, we estimated the potential productivity and efficiency benefits of GenAI in government across all levels of government and globally, this is worth $1.75 trillion by 2033. And then in Australia and New Zealand, we estimate it to be about $40 billion and $8 billion, respectively. And a lot of that is linked obviously to being able to generate efficiency and productivity improvements, staff costs. That's not to say that it would necessarily be harvested as savings because often there's pent-up demand and that increased productivity can be used to address unmet needs or improved service quality as well. The other area that we wanted to look into was to what extent can -- what were people concerned about and what could government do to address those concerns. So when you -- when we ask citizens to outline their major concerns, the #1 thing that stood out, and this was true across country was the impact potentially on the job market and potential for job losses. But there was also concerns around accuracy, the ethical issues, the degree of potential bias or discrimination in models and whether there was enough internal capability. And so what this suggests from a government perspective is that there's an important communication challenge is how can you assure people that when AI and GenAI will be adopted, that it will be adopted in a way that effectively manages the transition. We know that from experience more often than not, it is about augmentation, not necessarily automation. And so being upfront with people and managing those concerns, it will be a key part to adoption and implementation. We also heard that not everyone is yet convinced that the benefits outweigh the risks. And half of the survey recipients said they didn't trust government to use AI or GenAI responsibly. And this actually -- this is important feedback because if we rush too quickly without bringing people along, without making sure that people understand the benefits and are aware and understand what the risks are, we obviously -- there's a risk of some resistance or reluctance and that could potentially erode trust in government. What we also explored though, were some ways that government could increase trust. So as I mentioned earlier, one of the things that we found was that the more the people were educated and aware of GenAI and AI, the more likely they were to say that the benefits outweigh the risks. And so education and awareness raising is a critical intervention in terms of helping people understand. Obviously, the beginners and the more inexperienced tended to be more fearful and more distrusting and that's understandable. But actually once people understand it more what it's capable of, what it's not capable of, what the real risks are, et cetera, they tend to be more supportive. And then there were some specific interventions that would help increase trust. So one of them is having clear laws and regulations about how AI can be used. The second is having some specific rules and safeguards on how personal information will be protected. Obviously, people are concerned about their personal information, maybe ending up in models are being released or revealed. And then the third 1 was around how to make sure that there is transparency disclosing, for example, when AI has been used in a process or in a decision-making process. And another one was also just transparency around breaches or lapses, if there's been issues, don't try to cover it up, be upfront, be transparent if there's been errors. So what does that all mean? We put this all together in terms of a road map in 3 parts. So the first thing is to get started so GenAI is one of those technologies where to really understand what it is capable of and possible what's possible and not possible, you actually have to explore the boundaries and the frontier of possibility and you have to trial and work out what it does well and what it doesn't do well. And it's not always easy to predict where that frontier will be. We talk about it sometimes as the jagged frontier because sometimes it does things really well that surprise you and other times it does things not so well that you would have expected it to do. And that's where pilots and proof of concept can be really useful at building that learning through trial -- through structured trials, randomized trial, trials that proper consult groups. Secondly, as I've mentioned before, thinking about the use cases, where are the ones where there is a social license and where there is a level of comfort and support. They tend to be a good place to start. And using the trials as an evidence base for them further investments. There's -- this is a new area where everyone is upskilling and reskilling very quickly. And I think it's important to invest in upskilling the workforce so that they can use the tools effectively and responsibly. And then finally, we know that good quality data and scalable technology platforms continue to be a prerequisite for being able to move beyond proof of concepts and pilots. In terms of the trust prerequisites, as I mentioned before, having clear laws and regulations is important. And it's not necessarily the case that those will be set and forget. I think we can expect the laws and regulations will need to be agile and adopt as the technology continues to evolve. Encouraging a very clear set of responsible AI frameworks and policies and more importantly, creating a very strong responsible AI culture and an AI safety culture set by leadership from the top. And that is a fine balance between being sort of an advocate and a promoter for adopting the technology, but also doing it in a way that's safe and responsible. Being transparent with citizens about when AI is being used and how it's being used. And then finally, recognizing that the best outcome -- I think our view is the best outcome will always be the combination of the best of humans and the best of machine's brain working together, not either or. And then finally, there's probably also an important role for government to play more broadly as an advocate and a builder of skills across the economy. So given the productivity benefits and potential and efficiency that's available, it could spur another wave of productivity across the economy, and this becomes -- but unlocking it will require investments in literacy of the population in AI and in how to use it effectively and responsibly. There's always the risk of exacerbating inequity. So in fact, using the technology to address some of the inequities. We've seen some terrific examples of how some of these technologies and particularly help, for example, vulnerable populations, people with disability, those who have traditionally been excluded from participating digitally, either because of lack of access or education or otherwise. And actually, some of the new technologies allow us to break down some of those barriers and reduce some of that inequity. And all of this becomes a real catalyst for innovation and economic growth. So I'll hand it back to Gisele, but that gives you a bit of a flavor of some of the things that we found in our deep dive. Over to you, Gisele.

Thanks, Miguel. And what we're seeing from that exactly, as you said, is that within the guide rails that folks have given government to use a tool like GenAI, there is a huge opportunity that sits there. And using a tool like an AI as an accompaniment and augmentation tool to help deliver services more effectively and more efficiently while still preserving the absolutely crucial element of a human in the loop as the final decision-maker, the final point of the interface between government and the citizen is hugely important. And so that really takes us through -- to the discussion we have here about what is it exactly that you want to get out of GenAI? As government is rightly looking at this tool as a trust multiplier as well as a productivity multiplier, we look at the ways in which it's being applied across government. And we see the opportunity and the experimentation begin, which is a great thing, that familiarity breeds comfort element that we've seen in the data that came out of the research. But it also means that we need to take a step back and understand why we're even looking at these products in the first place. And so we need to define what that problem or opportunity actually is. So what is it that you want GenAI or any kind of the AI tool to be able to assist you with? A really clear articulation of what it means, both for the citizen as well as the public servant agent who is using the tool. What risks and responsibilities have to be considered with the ordinary execution of that kind of task? And how does that change when you introduce a tool like GenAI. So these are the -- it's the basic risk assessment that most good organizations already undertake. Then it's about understanding what -- how you assess data and availability within your organization and what data you need to actually create that augmented outcome, and understanding, of course, the compliance and support structures that you will need to underpin your organization's use of a tool like GenAI. And this really takes us into an understanding of what are the tools that you have available to you. Is this an automation element? Is it the use of a predictive AI? Or is it really a job for GenAI and really understanding the differences between the tools out there in the market. And of course, start small and iterate. But really, which Miguel has gone into in some detail. But it also makes you think about, when we think about GenAI these days, I think a lot of people think about it in the format or context of consumer GenAI, which is very much the form of a GPT tool where you type in a prompt and it scours the Internet for an answer. But really, is that what you're actually looking at for an organization where trust and the integrity of your data is fundamental to an accurate outcome for the citizens that you are trying to serve. And really, that's when we start looking at -- talking about organizational enterprise AI. And so when we think about that, are we talking about GenAI as an admin assistant, as a digital colleague or as a customer agent. And in each of those instances, we can map GenAI in terms of its proximity to the service agent as the decision maker. And I would suggest at this point in time, given the data and the research that we're seeing to date, that digital colleague narrative appears to be the way in which citizens feel comfortable with the public service using a tool like GenAI to help augment to create efficiencies, be more targeted in that approach, but not be a final decision maker in itself. So -- and we've seen a number of questions come in. But at the end of the day, this is a huge productivity opportunity for government, but putting in place both the right processes and tools for your organization to embed trust in each layer of that process is going to be fundamental to the why of doing what we do every day, about ensuring that both the service agent as well as the end user, the people we're -- the stakeholders we're trying to serve see a better, more effective outcome and thus, increased levels of trust in government service delivery and therefore, overall government.

So I've seen a number of questions come in and I'd love to move to the Q&A section now. Now what we've seen -- 1 of the first questions that came in was what are the ways in which government is actually using a GenAI? And I might throw to Fran in the first instance to answer that question.

Yes, absolutely. Thanks, Gisele. Great presentation as well. So I think that globally, what we're seeing is that there's lots of proof of concepts and pilots happening. Certainly, the private sector is adopting GenAI a lot faster than some of the government counterparts. But there's still lots of opportunity. And some of the biggest areas of opportunity I see, both in terms of increasing trust and also gaining productivity benefits, are really in the areas of customer service delivery. So whether it be GenAI bots to provide customer service to make it easier for people to self-serve, to navigate services or providing customer service agents support tools that are GenAI-enabled and also to streamline internal administrative task to free up capacity, to spend more time serving customers and meeting that unmet demand that Miguel was meaning before. I think in the rest to get started, one of the key things that or one of the key pitfalls because learning by doing is going to be really, really important, but one of the key pitfalls to avoid is just really remembering for governments to put rigor around the pilot proof of concept process so that you're testing and like measuring impact so that you go to scale because some GenAI investments are not without cost in going forward. What are you seeing Gisele?