Confronting the Future: Challenges and Opportunities in AI Adoption | Richard Crampton

Hi, everyone. For those who don't know me, - my name is Richard Crampton, I work for Atlassian - as a senior IT service management solution engineer. And I have the great privilege of half an hour of your time today - to discuss some of the challenges and opportunities - in adopting artificial intelligence. So, let's get started. In terms of agenda, I'd like to start, if that's all right, - with a bit of background and history around AI. From here, we'll move on to some of the challenges - facing everyone adopting AI, before moving forward - with the opportunities that AI can present. Lastly, I'd like to summarize some thoughts about AI, - and then we'll leave time, around five minutes or so, - for any Q&A at the end. So, hopefully, that all works for everyone as an agenda. So, let's start with the background. AI has been around in various forms for a number of years now, - but the capability and adoption has really grown exponentially - in the last two to three years. But let's take a step back. So, as you noted, AI is not new. This is a headline from the New York Times in 2018, - noting that AI tiptoes into the workplace. Now, we're now much further on with AI adoption than 2018, - and it's making fundamental changes - to the way that we work and collaborate together. Indeed, at Atlassian, we truly believe - that AI is changing the fabric of teamwork itself. Whether you're working in a Dev team, building the next amazing product, - working in a marketing team, creating a very unique campaign, - or as a service agent supporting customers, - you will be partnering with AI in order to succeed. So, we strongly believe that the winning teams of the future - will consist of human teammates supplemented by AI teammates. And through this major shift in teamwork, - Atlassian will be here to help welcome AI into your business. So, I've got an interesting statistic. We did a survey called the State of AI in Service Management, - facilitated by CITE Research, and it took in responses - from around 500 professionals in August of this year. Now, this survey found a remarkable 88% of organisations - are already leveraging AI in some capacity or other. Furthermore, the top motivators for this investment and adoption - were, one, using AI to fuel data-driven decisions at 80%. This isn't actually all that surprising if we look at other studies, - such as an IDC study, that found that 75% of execs, VPs - and directors are investing in decision intelligence initiatives. Second place was boosting workplace efficiency. 78% of responders said they were really interested in AI for this. In third, it was AI being the champion for customer satisfaction at 64%. Now, these findings align with broader trends observed in recent studies. For instance, a global benchmark study on generative AI found that, - while enthusiasm remains really, really high, - many organisations are now taking a more measured approach - to implementation in 2024. Only 63% of companies are planning to increase AI spending - this year and next year, compared to 93% in 2023. That's as they grapple with the rising implementation costs, - data security concerns, and challenges in moving projects - beyond the initial pilot stages. And organisations are also leveraging a diverse set of KPIs, - key performance indicators, to measure the success of AI implementations. And customer satisfaction, or CSAT, leads the pack, - 42% of organisations using it as their primary metric. For me, that underscores the critical role that AI is playing - in enhancing the customer experience. There's also a strong focus on operational improvements, - such as cost, again with 42% using operational cost savings - as a measurement of success of their initiatives. One percent less to 41%, organisations have a KPI - of how accurate their AI data model is and its overall performance. And lastly, time savings with 36% focused on the time - being saved by AI-driven efficiencies. Now, these figures highlight a balanced approach, - valuing both financial outcomes and the quality of AI-driven processes. Now, some organisations also have AI-specific KPIs, - with 20% of them monitoring AI adoption rates - within service management processes, showcasing a commitment to measuring - the penetration of the AI technologies they're introducing. It's also worth noting that only 3% of organisations report - not tracking the success of their AI implementations. This near universal commitment to measurement - demonstrates the strategic importance of AI initiatives - and data-driven approaches to evaluating the impact. And the variety of KPIs being used - reflects the multifaceted nature of AI's impact on service delivery, - spanning customer satisfaction, operational efficiency - and technological advancement. This comprehensive approach to measurement suggests - organisations are taking a holistic view of AI's value proposition. Next, let's have a look at some of the deeper challenges - of adopting AI in your organisation. So, again, going back to the survey, - data security remains a key consideration. 72% of respondents in our survey have worries - about the security of AI tools, highlighting the need - for robust security measures as AI adoption increases. Skills is also a concern. 32% of respondents noted concerns regarding skills - and talent shortages in their business, - affecting both the implementation of AI, - but also, the value delivered as an outcome. Aligned to this concern, almost half the respondents - are planning investments into training and upskilling initiatives - around AI for their core business. It's also worth looking at the different stages of AI utilisation - and the maturity reserve from our survey. So, we said 88% of organisations are using AI in some degree, - which means 12% aren't using AI at all. But of the rest, 23% are at the understanding stage, - learning how AI works, exploring the use cases - they think will be valuable for their business. 20%, or a fifth, are at the piloting stage, - running a limited number of projects with specific use cases - to demonstrate feasibility and the potential ROI. At 17 are at scaling, - expanding their AI pilots to broader areas in the organisation - or exploring additional use cases. And 29% are at optimising, - continuously optimising their AI models, algorithms, - and/or processes based on performance metrics and feedback. So, while 29%, the largest group, is optimising the AI usage, - having built out their maturity, - 71% are still in the earlier stages of adoption. It's also quite interesting that of that 29%, - most of those come from organisations with fewer than 5,000 people, - which demonstrates that agility is real key - to time to value where AI is concerned. One final note on this is that - budgetary constraints was considered a challenge for 31% of respondents, - suggesting that while the potential for AI is recognised, - securing the necessary financial resources - remains a significant barrier for many. Doubling back to data security, - let's discuss how you keep your data safe in the AI world. So, I'm going to do it from the perspective of us as Atlassian. Firstly, we build all our technology, - including the teamwork graph, Atlassian Intelligence, and Rovo, - in line with our responsible tech principles, - which we publish online for everyone to see. Secondly, we look to make sure that everything respects - the enterprise-grade permissions of the source content. We tailor every answer to every user in real time, always honouring their level of access. And thirdly, we continue to work towards - meeting compliance standards that you need. Rovo will be client with ISO and SOC 3 before the end of the year, - and we've also joined the EU AI Pact with fellow industry leaders - to stay ahead of the upcoming regulations. Now, the fact remains that AI is driven by data, - and great AI needs great data. But in my mind, there's actually three different areas of focus for this. One is the quality of the data. Now, I'm sure many of you are aware of the very old catchphrase, - "rubbish in, rubbish out", or some of its more colourful derivatives. If you feed AI bad or incorrect data, - it can, albeit very, very efficiently, produce incorrect outputs. But the second I want to think about is the scope - and the reach of AI into the data. So, the ability of AI to take detailed, correlated data - as an input to allow deeper, more thoughtful analysis. We'll come back to that one. And the third is the actual outcome. Even with the correct scope and the quality of data, - 70% of the participants in our study were worried - about the quality of data emerging as an output from AI tools. This does suggest that organisations are grappling with ensuring the accuracy - and reliability of AI-generated outputs. Now, at Atlassian, we have put a specific focus - on that scope element, providing what we call the teamwork graph - under our system of work initiative. So, the teamwork graph connects all of our tools - and all the underlying key elements within these tools - into a chain that can be fully accessible by Atlassian Intelligence. But what does that mean? So, to illustrate the power of the teamwork graph, - let's just walk through an example. Using a typical AI tool, we ask for updates - that our team has made while we've been away on vacation - and how this might impact our upcoming campaign. So, the AI tool can request the context - and get data sources for the user, their work, and the team. The output is a response to inform the user the team finished the sprint, - completed the epic, and closed three bugs. Useful, but not providing any real insight - into the impact we might have in our upcoming campaign, - which was the key thing we were looking for in our question. So, let's look at exactly the same thing now, - but with Atlassian Intelligence powered by a teamwork graph. The prompt is the same, but the context is much deeper and richer, - providing visibility into milestones, specifications, decisions - and research, in addition to the user team and work attributes. This time, the output is much more detailed, - providing insight into a decision to exclude a persona, - noting the ship date was moved and that research validated - a specific feature as the strongest value proposition. Now, AI comes in various forms, from distinct standalone tools - to agents, to functions embedded within a product. So, it is worth considering if you consider AI - the product itself or a feature of the product. Atlassian Intelligence is an experience within our cloud platform. So, it's native to the product. And we do believe that this can be significant in terms of accessibility - of the underlying interconnected data. We also believe that AI will succeed when it's baked into the product - to enable advanced experiences and outcomes - rather than AI as a separate product. So, having viewed some of the challenges in adopting AI, - let's have a more detailed look at some of the opportunities. Now, the Atlassian survey results reveal a strong appetite - for leveraging AI in service management - to drive innovation and gain competitive advantage. Process automation and optimisation emerges as the forerunner, - with an impressive 87% of respondents expressing interest. This high level of interest underscores the widespread recognition - of AI's potential to streamline operations and boost efficiency. Hot on its heels, repetitive task automation - garners interest in 86% of the response to the survey. This strong showing highlights a growing desire to free up - human resources for more complex value-added activities. And AI-driven analytics for actionable insights - also command significant attention, with 85% of responders showing interest. This enthusiasm reflects increasing recognition of AI's power - to turn data into strategic business intelligence. So, I'd like to introduce, some of you may be aware of this, - these are the ITIL scales for value, which is a helpful tool - for looking at the wider business case for adoption of AI. Now, let's work through this using an example of someone looking - holistically at the business case for AI in service management context. So, on the right-hand side, we have the supported outcomes. That could be better search capabilities, - better customer-centric language, and the cost removed, - much less cost to find data if you're not relying on a human - to go and find it and you're providing it all in context. And risk removed. Maybe we can confirm the correct procedures - have been followed by an agent using AI. But we've got to balance that with the left-hand side. So, what effective outcomes? Well, some customers may prefer to have conversations - with humans rather than AI. They might distrust AI. Cost introduced, there's always going to be cost to implement - and maintain AI and monitor it. And the risk is introduced. We've already seen there's major - concerns about data exposure, and I'd limit that. So, it's all about making sure you are balancing this correctly - to understand where you're going to get value, - what that investment is, and what the potential trade-offs for that are. It's also worth noting at this stage, there is a difference - in terms of valuable outcomes between efficiency gains and productivity gains. Efficiency is the ability to achieve an outcome - with less resources than previously. Productivity is the ability to achieve more outcomes - or better outcomes with the same resources as previously. Now, a specific gain can deliver value across both, - but the outcome tends to favour one over the other. Let's have a look at some examples. So, let's take the example of a support agent experience. An efficiency gain could be the ability for AI to highlight - similar historical incidents and suggest ticket resolutions. On productivity, it could be the ability to automatically analyse - the sentiment of the customer and draft empathetic comments. For an end-user experience, efficiency gain could be using - a virtual assistant to find knowledge and self-help. For productivity, it could be full self-service, - delivering these services with no human interaction. And for level two, level three support teams, an efficiency gain - could be quickly generating drafts of post-incident reviews, - and a productivity gain might be reducing the noise - by grouping your alerts together using AI. Productivity gains means you're delivering more - with the same resources, while efficient gain means - you are further freeing up the time for your agents - to better focus on those value-creating activities. So, we discussed some high-level use cases in adoption of AI, - as well as discussed balancing the opportunities with risks and costs. We also briefly delved into differences between efficiency and productivity. Next, I would like to introduce four different roles that AI can play. So, we have AI as the enforcer. AI could enforce policies and procedures. And this works very well when situations - and appropriate responses are clearly defined. AI can enforce the response. So, automatically opening or updating a ticket, - highlighting the relevant knowledge base article, - or an example, confirming the agent has followed the proper procedure. That's enforcement. We also have AI as the enabler, enabling quicker outcomes, - and it works when situations require some investigation and analysis. So, AI can enable the response, - finding similar incidents, suggesting knowledge-based articles. It can also be a champion to empower collaboration in the business, - and this works when situations and responses - need to be fluid or fast-changing. AI can enhance the response, - maybe with drafting a knowledge-based article, - listing action items for meeting minutes, et cetera. And AI can also enhance user journeys. So, knowledge workers want polished interfaces and experience. AI can enhance that interaction, offering help as a virtual agent - or making information easily accessible to agents. Now, each of these four roles do not exist in isolation, - but they rather overlap each other. Nonetheless, each delivers an advancement in a specific area. So, enforcement improves your compliance. Enablement increases efficiency. Enhancement improves employee satisfaction, - while empowerment promotes collaboration and productivity. So, let's talk about this for a real customer, - and forgive me if I pronounce their name wrong. It's Elkjøp. They had a specific challenge with the number of tickets they were getting in. So, in the first eight months, they had 80,000 support tickets, - of which 62,000 were raised directly by end users. So, around 350 tickets a day, and 100 core IT staff. They needed to drive both productivity and efficiency gains - in order to deliver a great customer experience. And I'm pleased to say they're still advancing on that now, - using new features of AI to help move this forward, - drive both efficiency and productivity. And this was not an isolated case. Forrester recently interviewed a set of Atlassian customers - and found that they could resolve 30% of all their tickets - through Jira Service Management Virtual Service Agent alone. And it enables them to save 25 minutes per request, - thanks to their experience. So, I'd like to ask you, like all of you, just have a think. What would 30% of all your IT requests - being deflected mean for your organisation? What would be that cost saving? What other value-add activities could your teams be investing their time in? Finally, I want to put together some overall thoughts on AI adoption. So, I find frameworks can be incredibly helpful - to provide structure and governance. And personally, I find myself increasingly drawn towards ITIL, - which is IT Information Library, as a framework for IT service management. And the application of the guiding principles of this framework - are also increasingly relevant for AI adoption. So, the first is, focus on value. Focus on what the desired outcomes of the AI initiative are, - what it will enable, what will be the overall value. And please note, value is subjective. It will not be the same for a CIO versus CFO. Second, start where you are. It can be really difficult to start using AI, - and we've seen on the maturity, it takes time, - especially so if you may be concerned of the quality of your current data. But you can't analyse it without testing. So, this could be spilling up a new test instance, a sandbox, - doing some experimentation, starting with that data feedback - to understand where you are today. Thirdly, progress iteratively with feedback. Very much an agile approach, right? We've seen almost all organisations are measuring the success of AI initiatives. And our measurements are absolutely essential, - but also, an agile mindset and the ability to make small changes, - small iterations and test quickly is key. It is no coincidence that the most mature organisations - are those most able to embrace this agile mindset. It's already part of their culture. Collaborate and promote visibility. Democratize AI and promote and evangelize the successes - to excite and motivate others. Think and work holistically. Think about all aspects. It's not just about the technology, - but the people, the mindset, and the skills required for success. Keep it simple and practical. Don't overcomplicate it, especially at the start. Avoid bloating your AI initiatives with all the cool things it could do, - and focus on what means most in getting that out the door. And last, and this could have been written specifically for AI, - optimise and automate, right? Think about how AI can fully or partially automate processes, - but also optimise others. There are so many different use cases, - and AI is absolutely key in driving this one forwards. So, I'll finish with some thoughts. While AI is exciting, it hasn't been significantly intuitive - for