Building trust in AI
Hi everybody and thanks for joining us today. We're going to be discussing AI adoption in the workplace. I'm Paul Joslin from the access group, I'm a head of engineering for Access Evo and access Evo creates a common user experience across multiple products and incorporates AI in an exciting but secure way to enable our users and customers to achieve more in their day-to-day work. Today I'm pleased to be joined by two guests,
Laurent Business Intelligence Manager at Martin's Rubber and Access customer, and Anil, Senior AI Cloud Solutions Architect at Microsoft. Laurent, would you like to just take a moment to introduce yourself and Martin's Rubber? Yes, sure, so Martin's Rubber is an SME, a manufacturing SME, who specialises in bespoke rubber products for industries such as Aerospace, Defense and Niche Automotive. I joined Martin's rubber four years ago as a BI Manager, so my role here is to maintain the infrastructure, maintain the IT hardware, software - the applications - and develop tools to help decision making, to run the business and drive improvement in manufacturing. Awesome and and likewise, Anil, if you could just do a quick introduction, maybe talk a bit about Microsoft with Access. So we have been collaborating with Access Group for more than one and a half years.
It is particularly an interesting client for us, it's the top software engineering company, they are creating really high quality software and they are almost in the same industry as us because they are focused on modern workplace solutions in a sense, and as Microsoft we also build modern workplace solutions and we can really empathise with them. So, I've been helping Access Group to build Access Evo copilot solution to augment their existing solution and used in a more interactive manner by their customers. Awesome, thanks Anil. So, Laurent, perhaps we could just start with you talking a bit about whether you or your team have actually already explored the use of AI in the workplace. So we're really at the start of it; it's a subject that is very exciting us at the moment but we haven't fully embarked into it. We're starting to look at candidates and areas of opportunity where AI
could play its role and starting to understand what do we need to do to be prepared for that journey and for that, let's say, let's use that word, that transformation into the AI world. And, obviously, we'll get to hopefully in the future about all the benefits AI can bring but I guess as part of that exploration at the start you're looking at some of the challenges and the technical hurdles of actually implementing AI in the workplace. Perhaps you could share a bit of, kind of what you've explored so far? Yes, so to me there's two main ones. My background is process engineering and in my previous days I remember that, as a process engineer, the quality of the data, of historical data, manufacturing data was a big obstacle to run predictive models or to write statistical tools and what I'm observing is with AI, we're moving into the same obstacle of the data that we would use for AI is not as clean or as available as it needs need to be. One example I would use is languages;
I could see that there would be huge benefits for us to use our own language, our own data, but the way it's written, the words that we choose are not consistent, are not clear, so we're not prepared to use AI on that subject. The other big one as a small manufacturing business, is making devices available to the workforce. If we want to deploy AI to the shop floor then these people need to have the devices, and it's not common that a small manufacturing business would provide, let's say, a mobile phone to every employee or a laptop to every employee. So that's really interesting, so you're talking about the data Integrity there at the start and you need to make sure the data is in the right kind of format actually for the AI to use to kind of move forward. But that's also a really interesting thing there I know you kind of work in the manufacturing space and I've heard previously about Access products having dashboards, for example, on the shop floor so that everybody can see the key insights and information but they might not have a device for themself to be able to use the AI day to day so that's very interesting. And Anil, obviously you'll be working with many customers; are there any
other technical hurdles that you're seeing these customers facing to actually adopt AI? I work with many manufacturing companies as well and I agree to Laurent that there is typically a single machine on the shop floor and the users, the employees working on the shop floor, don't have have immediate access to information, to the intelligence. I definitely agree on that, I think there is a path for digitisation and increased use of handouts inside the shop floor. There is definitely opportunity, there are definitely opportunities to use central consoles more effectively and to get feeds from the task management systems like what Access already provides to Martin's Rubber, to at least maximise the usage of individual users, the existing devices. So this is what we see, so at least an efficient usage of the existing devices could be an improvement in my opinion. Do either of you see across the kind of user base a kind of fear about security at all? Obviously there's a lot of sort of worries about will AI kind of use all my data in a bad way or anything like that; is that something either of you have actually kind of come across yourself? Yes, especially for us as a smaller business with smaller team there is that fear of AI will replace me because there is that, well, it's my job, I'm doing this part so if AI is doing it and I'm the only one that knows my job and my only job then what am I going to do? So, the smaller the team actually the more defensive I'm finding people are getting.
And I think that's one of the key things that we've noticed. Access actually did a survey recently, which I think will be published alongside this if anybody wants to download it, and the key use of AI that I think a lot of people are really embracing is the AI there, is an assistant to you, so it's not to replace you in your job, it's to empower you in your job. We found some really interesting information in this, that 72% of adults had already got an understanding of AI and half of the people surveyed actually used it already in their workplace, with 93% of respondents saying that AI was having a positive impact in their workplace. So I think that's one of the key things; AI isn't there to replace you, it's there to empower you and your business to be able to do more. Anil, like, security on the data, that must be a key thing that we've really got to focus on, making sure that the data is kept in a secure way? Yes, this means that you should be embracing still the best practices for enterprise governance; you would need to ensure that the data is accessed by the right people, you need to ensure the confidentiality problems, the individual confidentiality problems, you should be careful about how you use copyrights. So it was a typically one of
the first conversation points that we have almost with all clients - they want to ensure that their data is used in the right manner and we are doing a lot of myth busting actually on this topic, but at the same time the platform creators for these AI based tools have a serious responsibility of ensuring that they take their necessary guard rails to prepare the potential consumers for effective usage and a safe and secure usage in a sense. Yeah and I think that's it, we've worked quite closely with Microsoft .With Access Evo we've got three tiers of security so we talk about it being company secure permissions aligned and personally private, so whenever you interact with the AI services in Access Evo they are completely private so you can put your company data in there and not worry about that perhaps ending up in a training a model somewhere, and likewise the way that we integrate the products into this, we make sure all the permissions and access controls are exactly the same as using a product, so the data won't be available if you can't access it in your underlying products. So it's really great that we talked about the technical challenges because I think that's a lot of the fear people might have, but now let's get on to the exciting bit - how can AI actually improve productivity and streamline processes in the workplace? Laurent, I'd love to hear your thoughts in where you work. Yeah, so the areas that we looked at go in automation of routine tasks, things that people do day in day out, let's say the receipt of a purchase order or matching invoice with goods-in receipts and having an assistant to carry out these tasks.
The other one is yeah, the early preparation of work packages, so for me it's preparing a code, a piece of code that I'm going to use and I've used Copilot once to help me with transferring the price list of a customer into the ASC ERP system from the Access Group, so that was about 200 lines that normally somebody will go one by one in the ERP system and enter the data. The code that the Copilot gave me took 5-10 minutes to work and all the transactions were done within half an hour. That's a huge time saving, so it's that part of please help me with the part - I'll fine tune it, I'll make it work, but give me the body of the code I need.
And that, I think, comes back to what we talked about earlier about AI. AI is there not to be a scary thing, it's to be an assistant, it's to be able to give you more help doing things. I know again referring back to that survey they kind of asked many users what they use AI for today and it is the gathering and analyzing data, generating ideas, content creation, automating tasks, summarizing info, it's all those things that actually can benefit you on what you are already doing day to day. And Anil I don't know - oh, sorry Laurent. So can I can I just add to what you just said, I think there's also that part of if the assistant is going to do the trivial task then there is most likely a need for that person to learn new skill, so in that example that I've mentioned about purchasing, that free time for that person to do something else, and that something else could be negotiating with a supplier, spending more time negotiating suppliers, but that's new skills that that person needs to develop I think that's a great point actually, with the AI piece you can perhaps free yourself up to focus on the things that really do make the biggest impact in your workplace, and it's not to sort of say that, as you say, AI is not going to replace you but it can assist you on the things that perhaps you would have spent lots of time on and then actually let let you have the bigger impact in your workday on the real important tasks. Anil, I was going to say are there any other things that you've seen around improving productivity, perhaps even how it can provide valuable insights, for example? So, even before a generative AI boom I can say that we were always working with our manufacturing customers on improving the productivity to ensure that we optimize, we streamline their processes so this was a topic of general purpose AI even before generative AI. But
making human beings as a part part of it, being more involved and contributing to the process is also important to improve the productivity journey. So we observe that human beings at least delegating these kind of very repetitive tasks to AI tools compare unstructured information very quickly, that helps a lot. Human beings accessing to knowledge documents to information to data with their natural language and searching through knowledge bases effectively is really accelerating their day-to-day work so these were the some of the top use cases I can say that. There is obviously more advanced implementations of it where these tasks could be even distributed to the multiple human beings to ensure that the overall team is working very efficiently, and I think we will increasingly see these kind of use cases in our manufacturing customers.
Yeah and, like, I think one of the things that excites me at the moment when I'm seeing Evo, obviously we use it internally in Access so, as you said, getting the knowledge from policy documents, things like that, but then likewise we've got a feed which brings all the insight and actionable notifications directly to you wherever you are but they've actually now got a new feature which can launch the Co-pilot prompt off the back of that to go and get me the additional information that allows me to decide how I'm going to make that decision, so it's very very exciting. And, I guess, thinking about the benefits of AI, Anil maybe you will have an example around this; can AI actually enhance the customer satisfaction and loyalty as well? Some of the top use cases where we have driven the value was about customer satisfaction. There is a very well-known public customer story about it, it's one of the earliest clients who embraced our services around generative AI and assistance, it's the story of Vodafone; there are many a customers there who is trying to use AI at a certain level but what really differentiates Vodafone is they are able to really generate substantial business value out of it, and this business value is thanks to the customer satisfaction because telecommunications organisations are highly focused on customer service and they were achieved, able to achieve, with their new assistant much better net promoter scores and they were able to improve their call resolution times, they were able to improve their completions, successful completions, with the help of their assistant. This happened thanks to their way of organising organisation with our account teams in a sense because they are firstly asking the business value to start a project, and they are really planning everything around return on investment perspective, so it's more business value driven than technology driven in a sense in their decision making, and they are defining their dos and don'ts very well. So,
I can say that this has been a very successful case study we do a lot of this kind of customer satisfaction focused use cases across our clients, across the different industries. Yeah. fantastic. So, across the board I think there's a lot of pluses there for using AI in the workplace, whether it was helping with tasks, improving satisfaction, finding insightful information that help people make decisions. I guess, Laurent,
when you're actually looking at adopting AI in the workplace, are there particular financial resource implications from actually adopting this? Yes. the answer is yes. and I think that's that's a big hurdle, is that the upfront cost is for a business can be quite high and the time involved in setting it all up can be quite high, or potentially an initial project or solution with limited benefit. And what people need to bear in mind is that upfront cost is an investment, it will enable to deliver so many more projects after, because the right infrastructure will be now be in place, the right data and the right quality will be in place that you can much more easily deploy the next project, but the first one is the harder one and you really need to give people confidence of the journey ahead, of the 10 years to come. And do you have, kind of talking about the people in the workplace, there might be some people are not familiar with AI do you actually need to plan to upskill them on how to use this new technology? Yes, so I'm I'm not too sure actually. Yes you are correct there are people that are not aware of AI and I was taken a bit aback when we were having a conversation with one of the team, office based team, and I mentioned Chat GPT and they had no idea what it was, they never heard about it. I know, I was like, um, so yes there is that education, that learning to be done but
I come from a world of statistics and in the back of my mind is in a way people don't really need to know AI, they don't need to know how a statistical model works, all they want is the function, the what it delivers, so if you take think about our own life at home, we use AI all the time in some form; we've never been trained on it, it just happened, and we used it and there's minimal training required so I think in the workplace actually it's probably be more of a mindset change in what could AI do for me rather than oh how do I get trained on AI? I think that's like a fantastic point, right, if we think about the modern AI that we're using and looking to use across our products behind the scenes they're essentially using what we call a large language model. So, a large language model is a model that has been trained on human language and when we actually interface with the AI we're talking to somebody in human language, which, coming back from an engineering background myself, and I think obviously an engineering workplace that you work in as well Laurent, I'm used to dealing with zeros and ones and code. Actually the modern interface for interacting with these AIs isn't code, it's it's human language, you speak to it as if it's a person and it will do what you ask it to, so it's a great point that the actual amount of training that we need to give to people, once they've got used to the tools, might actually become quite natural to them once they actually adopt it. You also mentioned talking about 10 years' time thinking about long-term plans, and AI is accelerating at a crazy rate, right? It's insane how fast it's gone over the last couple of years. So what does the future for AI look like in our working lives, our personal lives, in the future? I don't know, Laurent, if you want to comment first and then we'll go to Anil, I'm sure he'll have a lot of ideas from Microsoft.
What you talked about, people asking as if they were talking to a human, I think I think yes, I think the the keyboard will disappear. You will just be talking to your hardware, whatever that looks like, and they will hopefully start to have convergence between the data, the information that we hold on our servers and the manufacturing equipment, and I could see that people - operators - will, will be able at some point to to talk to their machines, for the machine to do the work that they told them to do, not only being what we have at the moment which is very computer based, you know, I'm asking the computer to do something. I think there is a potential for manufacturing to move into, yeah, the manufacturing process becoming told what to do. Fantastic, very sci-fi-esque when you think about it, but it sometimes feels like we're living in those sci-fi movies that I grew up watching, we're finally there. Anil, I don't know if you've got any thoughts about this? I think I love really Laurent's answer, it was very in line with what I'm really envisioning and Microsoft envisions for actually where these assistance, where the extension - more extended usage of AI - will help us and where it will change things for us in our day-to-day interactions. What's really interesting about this innovation in the last years is it was really coming from bottom to top - your research shows the same thing I guess, it's very interesting that in many organisations actually people were interested in using GPT, they were really regard even without informing their employers they were, they they started on their own pace and, in our initial interactions with our clients, was they were trying to bring some kind of a order to this cause in a sense, to make it more enterprise friendly and in line with their organisational policies. So it came really from the people so, as it is adopted very well, people want
in their lives AI; we can see that it's a reality people wants AI in their lives, so the good, the important thing, next step is using responsibly AI in your life to be productive and to do, to make your life better actually and then your work life better as well. So all these things are very connected just as Laurent mentioned, he mentioned his machines, of course as a manufacturing expert he will say so, for us the computers, obviously, personal computers will evolve into something else like personal assistants in a sense, and our interaction with these assistants will change, they will be probably totally part of our life and following us through this journey and they will, they will, our interaction model will totally change that's the summary I can say that. And who knows, right? That's the thing, it's gone so fast the last couple of years it's almost hard to predict the future but it's definitely an exciting future and I think you mentioned that survey again and one of the things that I was amazed by was that employees are already taking advantage of AI so it's so important that the employers of those employees are embracing the tools that actually allow their end users to use those tools effectively in a safe and secure manner. Bringing it back from way into the future, back to now, Laurent I wondered if you could talk about kind of what you've actually seen today, where you have adopted AI in the workplace and what are the sort of benefits you've seen from that so far. So yeah, I mentioned the helping me with coding which especially for me because I'm I'm not an expert at any language, I just use what I found and what I can so it's great to have Co-pilot next to me, helping me with getting me started, and that enabled me to free so much time. The other
one that we started to use is around building our company policies and making sure they are aligned with current legislation, so having an assistant to help with the generation of these documents, it's a great time saver, great time saver, it still needs us to have a read through, correct, tweak, check, but that 80% of the work is done and, let's face it, it's not it's not the most interesting part of my job to start writing policies and procedures, so it's great to have something like that to help. And the other one I would like to see actually, going slightly off topic thinking about the Access Group, is the training; training materials and having people who use the product being able to ask these questions on how do I do this with this software. No, that's a great one and and everywhere in the business every product right now is adopting this Evo platform and they're looking at those powerful use cases so, watch this space, I'm pretty sure that will be coming and thanks for sharing that feedback. I know we're almost
out of time but, since both of you have got quite a lot of experience on this, I just wanted to sort of finish on the final question which is, some of the people watching this might be brand new to adopting AI in the workplace. What advice would you give to companies that are just beginning to explore AI solutions; are there any best practice or pitfalls to avoid? I don't know Laurent if you want to just go first? I won't go to best practice because I don't know them, but my two pieces of advice is, just try, just if you think of something you get to know it with doing something; just have a try at AI to see if it could help, don't don't be afraid of it. And the other one for me that we need to do here as well, is think now about your language. I've got this thing now where I'm starting to spot that even with it one site we, for example, name things differently; a team will call an area one way, the other team will call it another way, so when the time comes and you want to build your own language models then you will get stuck because your language would be all over the place and you won't have enough Big Data around you to make sense of it so, when you write documents, when you write procedures be consistent and be thoughtful of data the - yeah - the data and the language that you use. Excellent, thank you for sharing that and Anil, briefly, I don't know how much time we've got but maybe the best practices/pitfalls to avoid? What can you share? I can say that, as I mentioned the Vodafone story, there is you can just start to try, trying is really good you know, just trying reduces the fear in a sense make it, ensure that it's used and tried well in your organisation so you can maybe start maybe creating a restricted version of, you know, other open AI in your organization with security, you know, requirements set up, just putting some rights restrictions, making make it popularise in the organisation break the fear in fear about open AI in a sense. But, later on, really think about what are the business, what's the business value you can obtain from this what are the top use cases that you should really prioritise, that you can create positive outcomes to your users, to your internal users and your customers in a sense, and this will actually create further investment and further motivation to invest more in AI and in your organisation, yeah. So you need
create these success stories in the early stages, you need to discover firstly and then you need to create some success stories internally that will feed back to later further maturity levels. Fantastic. I was just going to say I think we're almost out of time so I just want to say thanks for your time today, it's been a really insightful conversation about adopting AI in the workplace thank you very much, thank you.
2025-01-03 17:33
