After around 80 years of incubation and development, Generative AI has emerged in 2023 as not only the most disruptive technology of the year but one of the year’s top mainstream news stories of the 2020s. A recent event held by the UK's digital authority, the Digital Catapult, provided some take-outs for SMEs, particularly those who have not yet embarked on a planned approach to AI.
With much hype surrounding the impacts that the technology might have on both organisations and wider society, the UK’s Digital Catapult – a part public-funded authority on UK digital innovation - held an event in London (21 June), on the “AI Communications Roadmap”, co-hosted with the telecommunications vendor Vonage, Contributions came from experts in technology, ethics, venture capital, R&D and a technology vendor and covered topics ranging from social impacts to technical configurations.
While there is much yet to unfold in AI’s story, there were several takeaways from the event that are relevant for SMEs today, particularly those who have yet to start their journey into AI:
Take out 1: New AI is different: Every business needs to consider it if they’ve not
“Thinking machines” can be traced back to Charles Babbage’s mechanical calculator of the 1830. However, modern AI is usually associated with origins in the 1950s, when teams of scientists in the UK (Alan Turing being the most famous) and US started to consider the concept of an advanced thinking and reasoning machine, based on replicating functions of a human brain. An early variant was presented at a famous two-month gathering of experts called the Dartmouth Workshop in 1956, where scientists and philosophers spent their days debating the prospects and the ways that they could evolve research.
Thereafter, the technology evolved with faster computers, more computer storage and better algorithms for the next 60 years.
However, Digital Catapult’s Robert Smith explained why the version of AI that has hit the headlines this year – so-called Generative AI – is distinctly different from the AI of decades before.
Commercial and industrial AI technology really came to the fore in the 1980s in the form of so-called “rules-based algorithms” or “expert systems”: Computer programmers would write code that comprised complex sets of rules, provided by asking human experts, When the rules failed or didn’t account for a scenario, a rule would be added or adapted to a programme. For instance, the Digital Equipment Corporation created a system called R1 in 1982 to configure computer orders and check their accuracy.
In the 1990s, these expert systems had become relatively sophisticated and started to gain notable successes, such as when IBM’s Deep Blue beat chess grandmaster Gary Kasparov in 1997.
However, the approach was limited for commercial applications, because adding more and more rules and knowledge to these programmes becomes increasingly expensive, and less valuable, as time goes on. This diminishing cost-benefit equation ultimately limits the viability of building expert machines beyond certain narrow applications and situations.
The present variant of AI – called Generative AI – works fundamentally differently and offers a profoundly different level of capability. Rather than computers being taught to recite human rules, programmers write code that teaches computers how to learn themselves (and effectively find and write their own rules).
Computers sift vastly greater volumes of data from the internet and learn to recognise patterns – in images, data and language – which they use to then recognise images, answer questions or make predictions. They also learn from their own mistakes, and the feedback of human users, extending their capability far beyond that of the expert machine.
ChatGBT by Open AI, is an AI chatbot that exploded into prominence in 2023, encapsulating a Generative AI technological breakthrough. This is not just because of the fluid language with which it responds to questions from a human user. It is because of the way it gets to the answers: answers are not derived from coded rules, but generative AI. The computer has sifted through billions of combinations of internet data to construct the most relevant response given the user’s prompt or question. It learns the structure of language to respond with a sequence of appropriate words to the user's entry. Moreover, it frequently manages this task in a second or two.
The acceleration of AI this year, following this breakthrough, has been astounding. In June 2023 alone, one web article pointed to there being 4,500 new AI applications. To date, AI has managed to progress with relatively little corporate or Government funding. However, this is now changing quickly. The UK Government had recognised AI as pivotal to the UK industry in their 2017 industrial strategy but published an updated national AI strategy in December 2022.
A £100m fund was announced in April by the UK’s innovation agency, Innovate UK, to drive the UK economy through AI. In its Artificial Intelligence sector study in March this year, the Government found 50,000 people work at 3,170 AI companies in the UK, generating £10bn/year in revenue.
As well as the panel talking about impacts on an industry scale, angel investor Elisabeth Ling pointed out that AI was also transformative for small SMEs: she talked of a company of 14 staff where the owner had recently set the challenge to transform the entire business using AI. Vonage’s Jon Lingard cited his company used AI in their communication management products for small businesses, which represent 90% of their business. The message was clear that we have now started a major transformation and all-sized businesses are affected.
Take out 2: AI can already offer significant benefits for managing communications
Vonage’s products help businesses to better communicate with their customers across a range of communication channels by integrating those channels with each other and back-office business systems. Imagine you call a company about an order you’ve placed, and start the conversation on the phone, but ask the company to notify you in WhatsApp with an update, before later sending a query email to request where the order is. Vonage’s product links the customer’s communication to both staff and the back-office systems, like inventory systems, where the information is held. Customers can get real-time, personalised information about their transactions using their device and channel of choice.
Vonage is one of many software-based companies that build business systems that are “plugging in ChatGBT”, lending them the intelligence to be able to answer users’ queries or resolve their issues, without recourse to calling support or trawling through data themselves.
Businesses are already able to buy AI-enabled software products, usually stored on the cloud, that offer the potential to automate processes, save money and improve customer experience. For instance, Vonage noted that when AI resolves customers’ queries, rather than a real call agent or operative, the cost is about 10%.
While it was noted that some consumers might still presently prefer human contact to an AI bot, this was believed likely to change: The panel noted that when ATM cash machines became more convenient than visiting a bank, the former became the consumer’s method of choice for getting cash. It was expected that once AI offered consumers the opportunity to get queries accurately answered more quickly than waiting to talk to a person, they will generally prefer the real-time AI solution.
Other benefits on offer were noted for businesses that had started to use AI: Used appropriately to “augment” rather than “replace” human work, ChatGBT can offer productivity benefits. It can already improve the way we search the internet, providing direct information to a question rather than a set of search results that the user has to interrogate themselves. This characteristic means that AI chatbots, like ChatGBT, could come to replace Internet search as we know it.
Elisabeth Ling also noted that where she had seen companies adopt AI, there had been a surprising lift in staff morale who enjoyed using it as, effectively a personal assistant, as it allowed people work quickly and "delegate" routine tasks.
Take out 3: Start small, build incrementally.
The ultimate capabilities of AI for companies can be profound. Robert Smith discussed how AI has helped aeroplane manufacturers simulate many different prototype designs to find the optimal before the manufacturers started work: of course, building hundreds of real prototypes to find the best design would have been impossible.
The range of applications for AI transcends consumer communications and includes all aspects of a business. It was recognised that this can make AI feel daunting for many businesses, and it can be unclear where to start to apply AI.
The panel spoke unanimously about the best approach as being one of starting small with experimentation and iteration. Vonage’s Jon Lingard said that his favourite business customers were those that would constantly carry out small, controlled projects, try, fail and improve. This approach saw them make headway more effectively.
As for choosing the first AI project, the panel’s advice was consistently to choose a current, manageable, and constrained business problem; find an area that is not hugely impactful to customers in case there are bugs and apply AI to improve it. The starting point did not seem to be critically important.
Other advice included forming a small cross-functional group to think about AI. If a business has multiple departments, ask each to brainstorm use cases where it could improve productivity.
Robert Smith advocated that businesses start to use ChatGBT, but should also ask it questions in a subject of expertise, which helps demonstrate its high value for non-complex research where the answers have low human impact, but partly also to also demonstrate its limitations. ChatGBT does not understand the knowledge it conveys: it works by understanding the structure of language and finding strings of words that it calculates are best related to the user’s string of words. It can misunderstand and provide nonsensical answers by misunderstanding context. That said, ChatGBT has an IQ of 155 and is able to get 96% A in a Maths A-level!
In different guises, the point was made by the panel that the businesses that will ultimately derive the most benefit from AI will be those that use it to intentionally drive sources of competitive advantage, which should therefore be the guide for selecting projects.
Take out 4: Ethics are critical. Involve the board.
Ethics plays a huge part in AI discussions, from a number of perspectives.
At the global level, the discussion between delegates and the panel frequently addressed the polarised views in the media on whether AI will ultimately create human good or harm. At one end of a very wide spectrum, some herald the dawn of a new era of humanity in which no one will have to work. On the other end, AI will trigger doomsday scenarios or mass job losses and poverty. Robert Smith of the Digital Catapult suspected the truth might “end up lying between the two”.
Smith noted that the pace of unchecked progress this year has motivated some 1,400 industrial leaders to sign an open letter in March calling for a six-month pause to AI development beyond GBT, including Elon Musk and co-founder of Apple Steve Wozniak. In May, one of the “godfathers” of AI, Geoffrey Hinton, quit Google to speak out about the “dangers” of the technology he had helped develop. Discussion continues about the nature and extent AI will be regulated.
Particularly given the current limitations of AI, this makes ethics extremely important for anyone starting to use AI-based systems in communicating with customers. The consensus was that risk – and particularly risk to consumers - should be taken seriously by everyone concerned, including any and all businesses that adopt AI.
Many supranational entities like UNESCO, universities, government bodies and industry groups have been discussing AI’s ethical implications for some years. Paula Nordstrom, a consultant on the panel at the Catapult event, highlighted the key issues relating to customer interactions included the impact on consumer privacy, surveillance, and the exasperation of discrimination.
In a world where consumers swap personalised data for personalised experiences, it was acknowledged that consumers already yield more data than they might realise to both companies: internet usage data, social media posts and mobile phone location data. Ethical questions are created by scenarios such as if AI-powered businesses which can crunch this data to predict consumer behaviour know more about consumers’ likely future decisions than the consumers themselves.
The discussion also covered questions such as whether companies should disclose to consumers when they were interacting with AI rather than human staff, since AI is now getting so realistic that it can elicit human sympathy which influences consumer behaviour.
It was advocated that, while businesses should start to experiment with AI, it would be a danger to leave the activity unmanaged and purely in the hands of the R&D or technology teams. Boards and owners should have oversight and set the rules in light of industry parameters, changing ethical expectations and the company’s values. The Digital Catapult now publish an Ethical Framework for individuals and organisations developing AI-based products which can help.
Take out 5: Help and information is available.
The Digital Catapult’s focus tends to be larger businesses and specific industries where their finite team will have the greatest effects, such as aerospace, manufacturing and creative industries. Therefore, the panel discussion had occasionally seemed understandably skewed towards larger organisation scenarios (such as with a discussion about whether to ring-fence several staff to explore AI or create internal working groups, which is likely to be a luxury for smaller-end SMEs).
An audience member asked the panel how the UK could possibly compete in AI against the US and China, and nobody quite had a compelling answer beyond the hope that the UK should attempt to differentiate. I asked a question of the panel how the UK’s SME population – particularly the c5.5m businesses under 50 employees – might be supported to maximise the value they get from AI, thereby helping the wider economy, given highly varying levels of awareness and confidence across the population.
The Digital Catapult were keen to point out that they are open to businesses of all sizes via their events and online information. Paula Nordstrom acknowledged, though, there is a “gap” in support needed for small businesses to exploit AI and is discussing initiatives with the Institute of Directors to close it.
There are several sources of insight however: Many vendors offer AI support: Microsoft wrote a guide called “AI for small business: a beginners guide”, as well as offering courses. Many AI vendors have Account Managers or Success Managers to help their clients to use their solutions effectively.
There are growing numbers of independent AI consultancies.
Low-cost online AI courses are available from the likes of Udemy or Coursera.
Several conferences are also being hosted in the UK (and Europe) this year, notably the AI World Congress (28/6), the International Conference on Artificial Intelligence and Machine Learning (29/7-30/7) and the AI & Big Data Global Expo (30/11-1/12), all in London. There is also the International Conference on Advances in Artificial Intelligence at Birmingham City University (13/10-15/10). There are many other events and AI appears as a topic in various industry conferences too.
Of course, there are a plethora of online information and news sources.
At the point of a SME needing investment, more funding is becoming available for AI innovation projects and start ups, that includes but transcends the £100m AI fund. Innoavte UK also help businesses of all sizes, including at pre-revenue stage, to connect across their network.
Universities received a £54m funding boost in March, which opens up partnership potential. There are a number of funded programmes for the development of AI in specific sectors, like the NHS, such as a £20m fund to address “Cancer Grand Challenges”.
Global AI start ups also raised $50bn in 2022 (Global Data) and London has been a particularly strong performer in the global market, with 750 data science start-ups existing in May.
The dots are gradually joining. For SMEs, there are already plenty of entry routes and routes to continue with an AI journey.
In conclusion
The broad tone of the Digital Catapult’s AI Communications Roadmap event held in London (21/6), along with other developments thus far in 2023 suggests:
AI will now have a major impact on most, if not all, organisations and sectors. The emergence of generative AI is a disruptive breakthrough and, while there are risks, the message sounded was the applications and opportunities demand a considered response from businesses of all sizes.
However, risks and uncertainties remain, which means that AI projects should be controlled and not left solely employee’s pure discretion, as decisions have ethical implications and reputational risk. Business owners and/or boards should define the parameters of experimentation based on ethical considerations and risk.
Starting on the AI journey can be daunting. Therefore, businesses should start with a planned and considered approach to how they will use the technology.
There are a range of ways for SMEs to either start their journey or to become more considered about their approach. Software in which AI capabilities are included are already providing businesses of all sizes with opportunities to derive benefits.
Comments