7 takeaways from the AI Summit London
This week I spent an insightful two days at the AI Summit at Tobacco Docks in London.
I watched talks from leaders at companies including Barclays, HSBC, NVIDIA, OpenAI, BBC, and Ocado as well as panel discussions with governments and regulators.
Here’s seven takeaways as you plan your own AI strategy.
If you prefer to watch - then I provide a 25 minute walkthrough of these topics in this YouTube video.
If you prefer to read - keep scrolling!
Digital reinvention not digital transformation
There is no understating the view that AI will completely reinvent the way companies run. This is not like the shift to cloud - “we have a box here and now we’re going to use a cloud there.”
This is about first principle thinking and going back to the very essence of why each process in our company exists.
Take a lead to cash process for example. You might think AI could improve the efficiency of parts of that process - automate some emails, answer a customer’s questions.
Instead first principle thinking asks what are we trying to do with this process, why does it even exist? We want to take someone that doesn’t know about our product to someone that pays for our product - how would we rebuild that process for the new world? Would we have the same team structure? Would we have the same process steps? Would we have the same systems?
Umang Paw, the incoming CTO of PwC UK explained, “we are looking at every product, every service and asking, what does this look like with AI embedded”
Jonathan Moore, Chief Operating Officer, Global Trade and Receivables Finance at HSBC added “AI is pervasive - it is going to be like the internet, like your mobile phone”
Stephan Gillich, Head of EMEA AI Centre of Excellence at Intel shared this graphic - explaining that we are moving into the world of AI functions - automating complex, enterprise-wide outcomes.
The next few years gives the opportunity for smaller competitors in a category to leapfrog the industry leaders - driving those leaders to invest to stay ahead. Its a race to win your category and everyone is on the same start line.
AI is a board room priority
Francesco Frederico, the Global CMO of S&P (the financial services company) explained that ChatGPT had made it easier to get into the board room.
“Before ChatGPT you couldn’t fight your way into the CFO’s office for investment in machine learning and AI - but now they are pulling us in”
The Group Chief Data and Analytics Officer at Lloyds agreed, “CEOs are going home and seeing their six year old child programming, creating videos and music - they understand the implications to their business and their industry”
Your data is the source of competitive advantage
The volume of data is exploding, but it is a different type of data - Jeri Culp, Senior Director of Data Science and AI Solutions at HP explained 80% of data is unstructured - video, audio, images, documents.
To prepare for the new world leading companies are focusing on getting a handle on what data they have and making it accessible to the business.
NVIDIA’s VP for EMEA Matt Horgan encouraged the audience to “free your data, make it accessible to your teams. Realise the value from the data you have and create breakthrough new products and services.”
Matt Harris, SVP for UK, Ireland Middle East and Africa at HP explained, “there is a race for advantage, but most customers don’t know where to start. Most projects are proofs of concept focused on internal use cases - very few are external value creation”
To give you a sense of scale JP Morgan’s internal enterprise data runs to 150 petabytes, whilst ChatGPT4 was trained on a dataset of just 1 petabyte (this was a quote from an insightful 20VC podcast I listened to on the way home from the AI Summit).
Ed Challis, Head of AI Strategy at UiPath shared this quote from McKinsey reminding us all to get our data centralised, organised and ready for our AI usecases.
Focus on your processes and your people
Whilst the tech is exciting, multiple speakers advised the audience to focus on their processes and their people first.
Ranil Boteju, Group Chief Data and Analytics Officer at Lloyds agreed, “For every $1 you spend on tech you’re going to spend $10 on business process reengineering. You need to understand your processes.”
Sara-Amanda O’Keane, Managing Director at Barclays added “AI is a team sport - its not an IT responsibility”.
Richard Hasslacher, Global head of Alliances and Channels at OpenAI agreed, “The challenge is not technology, its people. This is a change management problem”
Umang Paw, CTO at PwC (who have recently announced their global partnership with OpenAI, including purchasing ChatGPT Enterprise for their 100,000 employees), explained their approach,
“For the UK firm we have a champion network of 600 supporting a team of 25,000. We need to invest in our people as much as our tech”
He explained he had recently hosted an elective training day for 5,500 of their tech team with a selection of 26 AI modules.
“We need to provide our people with skills for now and into the future - it is an opportunity for us to reinvent our business.”
Safety is a concern, but not a blocker
With commoditisation of the foundation models, just using generative AI will not be a differentiator.
The key to competitive differentiation is how you use your enterprise data to add context to your models.
Returning to the JP Morgan example - their enterprise data runs to 150 petabytes - companies are balancing the need to use this data in their AI strategy, but without their unique intellectual property being shared with the wider community.
Stephan Gillich, Head of EMEA AI Centre of Excellence at Intel shared this graphic to explain the important relationship between a company’s data and the public AI models they want to use.
In the middle of his diagram is RAG - Retrieval Augmented Generation - and he explained how companies are using this approach to safely supplement their prompts with vectorised enterprise data.
Additional concerns raised across the speakers included:
inaccuracy
IP infringement
explainability (why has AI provided a specific response)
workforce displacement
fairness and ethics
reputation
To solve these challenges, Tom Betts, Group Data Director of Kingfisher (a large European retail group) described how they had set up an orchestration framework called Athena which centralises their AI strategy.
“At this stage it is unclear which model or which hyperscaler a particular usecase might use. Athena allows us to handle multiple LLMs and to combine outputs in a central place. It makes it quick to build applications. We want our teams to have less fear and more curiosity”
Having a flexible framework is essential he says, “Whatever works well today may not work well tomorrow”
Richard Davis, the ex-Chief Data Officer of Ofcom (UK media regulator) described the creation of AIOps teams to operationalise AI at scale.
David Norris, Chief Growth Officer at Holiday Extras (who had just won an award for the top AI implementer of the year) described their AI steering group - which defined roles and responsibilities, core AI principles at the company, included members from tech, legal, operations, people and all business teams - and is responsible for reporting to the Exec and the heads of business.
Lloyds calls theirs the Control Tower. It includes an AI ethics team. They encourage their teams to bring their ideas and use cases where they can be supported centrally.
Generative AI is more than text
Two surprising sessions focused on music and video.
Daniel Bedingfield (UK pop legend) landed from LA to walk the audience through creating a synthetic song, where he recorded a small snippet, and then an AI tool ‘filled in the blanks’ creating the rest of the track and instruments.
He was supported by Fernando Garibay who demonstrated an entirely AI-generated song where we captured thoughts and emotions from audience members, used ChatGPT to create the lyrics for a song, before Udio created two completely unique songs in just seconds.
Fernando explained that now you can focus on what you like doing - song writing, or singing or piano - and let AI do the rest. “It used to take you 20,000 hours to be an expert - now it takes ten”
Clive Santamaria, Chief Architect at ITV (a UK broadcaster) walked us through how AI is becoming a valuable tool in the media production process - we saw examples of translating into new languages, but with AI automatically reformatting the video to fit the new audio - no more bad dubbing!
We saw examples of de-aging characters like Harrison Ford in the latest Indiana Jones movie, or AI generated B-roll (supporting video that complements the main shot).
He described automatic recuts and reshoots without having to bring the actors back in.
Whilst these were media examples, my mind jumped to use cases in every industry - customer training, safety videos, employee onboarding - imagine your English speaking CEO being able to welcome Japanese employees or customers in their own language.
The HP VP quoted a US colleague, “everybody in the team should have a minor in AI.”
The people closest to the process are the ones that uncover the biggest AI opportunities - so make sure you give them the knowledge they need.
Everyone is learning
I wish I had a clicker to count up the actual number of times I heard the word “learn”. Every single presenter - “we’re learning”, “we’re helping our people to learn”, “our executive team is learning”.
ChatGPT is barely 18 months old, it doesn’t matter how big or well funded your competitors are - this is a level playing field for knowledge.
You may have noticed I included the titles of everyone I quoted. These are not integration or AI specialists, or even in many cases IT roles - these are senior executives focused on the business - it is telling that the companies pioneering in this space see AI as a business opportunity, not a technical one.
I for one am excited to be helping companies on this journey.
Start with the processes. Make the data available to your people. Safely introduce the technology.