杨莉明部长在微软 AI QuickStart 计划启动仪式上的开幕致辞

Good afternoon, colleagues and friends, Rachel, Tolgar, Kavita, Janet and Wee Luen.

Thank you for inviting me to be part of the launch of the Microsoft AI Quick Start programme in partnership with the IMDA as well as with UOB.

Some of you know that I started my professional life in the Economic Development Board, focusing on supporting Small and Medium Enterprises (SMEs). It was a very long time ago, and in those days, we were trying to help SMEs overcome financing problems . We have a scheme called the Local Enterprise Finance Scheme (LEFS), and we had another scheme called Local Enterprise Technical Assistance Scheme (LETAS).

Around that time, we also put together what was then referred to as an SME master plan and a retail sector development plan. So those were the kinds of things that occupied my attention back in the early 1990s.

What I can see is that with each successive wave of technology development, we've always paid close attention to SMEs’ adoption. This comes from the belief that in every vibrant economic ecosystem, SMEs feature very strongly. That's what we believe at its core. You need to recognise that they add dynamism, they bring a lot of value, and very significantly, they create good employment if we support them in the right way.

There is another reason why we believe that it's important to support SMEs when it comes to technology adoption. The studies and the research show quite clearly that technology diffusion is uneven. Very often companies at the frontier are able to pour in the resources, gather their teams and apply their minds to experiment and then create more competitive advantages for themselves.

But this ability to benefit from technology development very often, does not go beyond the frontier companies. It stays with them and there is a very long tail of a great many other companies that don't seem to be able to catch on. This is something of concern, but not an impossible task.

If you look at digitalisation today, well, it came after computerisation, and computerisation came after mechanisation.

If you look at SMEs in Singapore, the track record is a fairly strong one. Today, amongst SMEs, 95% of them would be using some form of digital technology. Maybe not all equally sophisticated, but it's not such an alien idea to them anymore.

We see this as something that is therefore worth our effort and worth devoting further attention to, and we have done so in different ways. For example, IMDA is the designer and implementer of the SMEs Go Digital programme.

Today, the challenge before us is to think about how AI adoption can also be made more widely available amongst SMEs. After all, we talk about AI as being a democratising technology, and what that means is that many people can access it. It's not confined to a few, but just because it is a democratising technology doesn't mean that it will be used to the same extent.

We have to ask ourselves, what challenges stand in the way of SMEs in particular?

We have some clue as to the impediments, the roadblocks that get in the way of SME adoption. If you look at their workforce today, well, according to our digital economy report, three in four of the members of our workforce, including those in SMEs, already use an AI tool, or maybe more than one on a daily basis.

However, when we look at the enterprise-level usage of AI, it is a completely different picture. For bigger companies, the number is closer to 60%. So even though three quarters of the workforce use it on an individual basis, even amongst larger enterprises, the current figure is about 60%. What about for SMEs? The number is closer to 15%. Large companies use AI tools at an enterprise level, four times as frequently as those among SMEs.

But if you think that this is reason for despair, I would urge you not to think this way. The SME track record in adopting technology is actually quite encouraging. To me, the bigger question is, how long will it take? How fast can we accelerate the process, and how deep can the AI usage go? So it's more a question of how long it takes and how deep it can go, rather than whether it will happen at all. I think it's a matter of time that the gap will close between what you see as the prevalence of AI usage in larger companies today – 60%, SMEs – 15%.

I'm confident that this gap will be closed. It's only a question of how long and how deep each one of them will be able to use AI.

The question then becomes, what we can do to speed up the process and help people to use AI in not just a superficial way, but in a more meaningful way that can help them truly transform their businesses and operate at a completely different level of effectiveness?

I see the AI Quickstart programme as one of the ways in which we can help advance this agenda. To turn this idea, the prospect and potential of AI being democratizing, including for SMEs, into reality, because it is not yet reality.

In our interactions with different companies, including those that are less well-resourced, it's clear that there are several gaps. There are several challenges to overcome.

The first is the issue of cost of experimentation. Building prototypes, these cost a little bit of money. Though in the AI age, this may not be so expensive, because to build a prototype about how your work processes can be automated, this can be supported in the use of agents. That process has now been shortened considerably, and it's not that costly. The real cost comes in how the prototype can run alongside and take over a currently live system for you, the integration with your installed base and other systems. That takes a bit of effort and that itself carries some cost. There is another cost of course. Two years ago, there was a big concern about the cost of compute, because AI is something that works only if you are able to access the compute capacity. So we've tried to plug this gap, and Microsoft is one of our partners in doing so. We have the Enterprise Compute Initiative that supports companies if they have an intensive use of compute capacity. This programme enables them to get credits in order to try out and implement their AI-enabled solutions. So cost is one impediment that we have to overcome.

The second is capacity. Capacity is potentially an issue at two levels. One is at the leadership level, because it is leaders that set the tone, and it is also leaders that articulate the ambitions, the extent to which they hope to see AI transform their business. If the leaders are not ambitious, then it's quite unlikely and difficult for the company to dream bigger. But if the leaders have the capacity to dream bigger, then even if they did not achieve all of their goals, they come closer to breakthroughs. But the capacity cannot remain only at the leader level. Leaders are only leaders if they have followers, so if the entire organisation does not have a complementary capacity to help fulfil a vision that has been articulated by the leaders, then the AI adoption will also not go very far. The staff must have is the ability to bring what was just a prototype, something more likely to be at the fringe, to the core of the company, the core of their processes, the core of how they deliver products and services to their customers, the core of how they interact with their key suppliers in order to fulfil their promise to their customers. So bringing things from the fringe to the core is not a capacity that gets built overnight. You need to invest in growing it. So that's the second thing, the challenge of capacity.

The third, and this is the one that we really stand a good chance of building up through this programme, is confidence. I say confidence because at the start of every experimentation, there is risk. So if you don't have some confidence, it will be hard to imagine an attempt that is significant enough.

Speaking of this, I was reminded of something that I was confronted with about two years ago. I was trying to put in place a display cabinet. This is nothing very high tech, it's just a display cabinet, but I was attempting to do it myself. I decided that getting something from Ikea might help me do the trick. The difficulty for me is that it is a glass cabinet, so I thought it's a bit dangerous, because we're dealing with glass, and I've not done it before. What gave me the confidence are a few factors.

One factor is that when you go to Ikea, you see lots of other people bringing home packages. So if you see other people being able to do assemble things themselves, maybe you have some confidence that you can do it well too.

Part of the confidence also comes from the fact that on previous occasions I have assembled tables and drawers. It turns out the experience is not so frightening. It is still a bit of a leap, because now I'm trying to deal with glass, but it's not as though I have no experience.

I share this with you because I half suspect that AI experimentation on an enterprise level requires a few of these elements. You need to build confidence seeing other people succeed in their endeavour. You need to build a little bit of confidence by having attempted smaller projects yourself and realising that you don't actually fail so miserably. In fact, you can have small successes. But I think also very importantly, what the IKEA experience tells us is that you need dummy proof instructions. I don't say this in a pejorative sense. You need instructions, you need products, you need tools that have been simplified so that even if you don't have very sophisticated skills as a carpenter or assembling things, you can still put together something workable, functional that actually creates utility for you.

That is the goal worth working towards, and something I hope programmes like QuickStart will help us to build up an understanding of what it takes to make progress with AI adoption.

I want to just sum up by saying that the approach we are taking is based on partnerships. It is based on the belief that we each bring something to the table that is helpful to the SMEs.

With Microsoft, we bring technical expertise. It's the equivalent of the IKEA customer service. If all else fails, you call them and you say, can you please come and fix it up for me? Hopefully you don't have to call them, but if you need to, they're there, and that helps to build a degree of confidence in proceeding with AI adoption, with AI experimentation.

UOB, as a partner, also bring something important to the table – your networks, your reach to a very wide base of customers, your touch points with them. When the experimentation gets big enough, your ability to offer financing solutions that will help the businesses go further.

IMDA also tries to do its part in capacity building. The Digital Leaders Programme, which has been running for some years now, has now got a new track focusing on Generative AI, specifically what Generative AI can help businesses do.

With this partnership, we hope to bring to life the idea that AI adoption for SMEs can move from the fringe to the core by overcoming the challenges of cost, capacity and confidence.

Our ambition is that for a start, we would like this programme to be able to help 1,000 SMEs, so that even if these businesses are not AI native, they can be AI pacesetters.

To all our prospective AI pacesetters, participants in the AI QuickStart Programme, I am very confident that this is a programme that will enable us all to learn together and turn the potential of AI being democratising and transformative into reality.

On that note, all the best. Thank you again for inviting me.