Creating compelling intranet content with AI

Syngenta


EMEA | SharePoint Intranet Festival 2025

Get inspired in this session as Melinda Schaller will share how Syngenta is leveraging AI and SharePoint capabilities to create a better content experience for colleagues. Hear how SharePoint knowledge bases make it easier to create better content, how they’ve embedded an AI Manifesto and Code of Practice, the steps they’ve taken to support colleagues in adopting AI functionality and what it all means for the future of the intranet.

  • But that's enough from me because I can see with a very stylish, snazzy, tech looking cool background, I can see Melinda Schaller from Syngenta is on the line. And Melinda, can I hand over to you and you share screen and share your presentation? Thank you. Absolutely.

     

    Thank you so much, Pete, and thanks for having me back again this year. I'm just going to share my screen quickly. Hopefully everyone can see it.

     

    Quick thumbs up. Yep. Yep, cool.

     

    All right. Well, I know we need to keep to time so I'm just going to dive straight in. Today I wanted to talk a little bit about how you can create compelling internet content with AI.

     

    I know AI has been a bit of a topic today. Some people are experimenting more with it than others. But hopefully this inspires you to think more about what it could be doing for your internet in order to get people more interested in that content.

     

    Because as we've heard today, it's hard to get people's time and attention. But maybe we can make it more compelling with some cool AI tips. Quick note about myself.

     

    My name is Melinda, as Pete said. I'm actually the IT capability lead at Syngenta Group, but I'm the IT business partner for our internal comms group also. And I actually came from that department.

     

    So I've had a lot to do with the internet over the years. My previous role was the internet and content manager. And so I still have a lot of links into that world.

     

    And it kind of is a world that won't leave me, which is good because it's a really fun area. Quick word about Syngenta. We're basically an ag tech company.

     

    So we help millions of farmers around the world to grow food, basically, with our technology. And we always aim to be the most collaborative and trusted team in agriculture. So technology is really a big part of what we do.

     

    And yeah, we try to bring that into our day to day a lot as well and really try out new technologies as soon as we can. And AI has certainly been one that we've been experimenting with for a really, really long time. So what I'll talk about today, however, is what AI can do for internets, how you can leverage it, and also why you should.

     

    So before I get too much into showing some examples, here's just some really quick ideas just to get you thinking in that direction of what AI could actually do for you. So content creation and editing, this is really an obvious one. And I would assume that most of us are doing that.

     

    Communication planning, bilingual communication, the translator services in AI tools such as ChatGBT are so good and so accurate and they're getting better all the time. We're actually really leveraging this for a lot of our more regional comms in offices where sometimes professional translation hasn't been readily available. So this is a huge benefit for internets in order to be able to personalize the experience for people as well.

     

    Media monitoring and analysis, internal comm strategies, brand messaging, summarizing information. I mean, this is one of the first use cases that really came about with ChatGBT. They were really kind of selling it on the fact that you could put in documents and just get a summary without having to go read loads and loads of different text and quickly summarize it for your stakeholders.

     

    Content repurposing as well, so that's turning articles into Yammer posts or LinkedIn posts or even Teams messages. Presentation and speech writing, training and development materials, this is something we also used it for quite a lot over the last year, particularly in the capability area, which is where I sit. Market research and competitor analysis, this will enable you to get up to speed with topics really, really quickly before going into stakeholder meetings, media meetings, whatever it might be.

     

    Data visualization assistance, so you can take some of those SWOOP for SharePoint reports, put it in and get some really good key messages out of it. SEO and content optimization, so helping your users to more easily find the content that's on the internet. Reputation management.

     

    It can also help you to keep SharePoint pages up to date. And also the research analysis, which I mentioned it's in there twice because it's a really important one because it's really going to help you to disseminate all the different pieces of information that you get, particularly on SWOOP. So, what I wanted to do also with this slide is just talk a little bit about some of the skills that you'll need to build to make sure that AI doesn't actually replace your job.

     

    Because whenever we talk about AI in Syngenta to some audiences, particularly audiences that are a little bit more hesitant to experiment with it. One of the reasons for hesitating is that they don't want to use it because they think if they do it will replace their job, it'll do maybe a better job than them, or maybe it'll just start taking over some of their key tasks and then their boss will notice and they'll be in danger. But actually by embracing AI, it actually allows a whole new set of new skills that we can bring to the table as humans.

     

    So things like the continuous learning, critical thinking, that's how you're prompting it, our ability to solve the deep complex problems by using that kind of human oversight about what we get out of AI, even the information that we feed AI as well. Things like prompt engineering so being able to come up with something that that looks at a topic slightly differently to get a different answer. Ethical understanding so not just taking those answers that that come out and saying yes that'll be fun and putting it on a report, but actually giving that human oversight to really say well you know based on everything I know in the human experience.

     

    Is this an ethical answer. Our adaptability, as well as also really really important so being able to work together with these tools and adapt to how they evolve. So it's really important that we sort of don't shut off from AI, but really see well how could we leverage that in order to keep our jobs rather than avoid it, so it doesn't take over our job.

     

    So using AI to enhance internet content there's a few tools that we're using here at Syngenta, we actually have our own version of chat GPT that we've had for a little over two years now, and it's called SynGPT. So I'll refer to that throughout today's presentation because it's just easier because it's so much into my vocabulary now I can't kind of not say it when I'm talking about AI. But the models that we have included in that platform is GPT-4 so it's originally based on the LLM from chat GPT so open AI.

     

    We also leverage Sonnet, Cloud, DALI, Gemini and now DeepSeek, and the kind of tools and capabilities that we have within that platform, the usual chat that you will see in chat GPT so you can get your document summaries, key messages, text refinement, improvement, etc. We have a separate feature in there for document translation so PowerPoint presentations, PDFs, Word documents, etc. We have an image generation tool.

     

    We also are using computer vision and computer vision is something we've actually rolled out to our customers as well so when farmers are on a field, they can literally take a photo of a plant that has a disease in there and ask what disease is this, what product can I use to treat that. And now we have also knowledge bases which sit within the platform, but also over the top of some of our SharePoint internet pages. So, some of the benefits of using AI, I actually wanted to show a quick video here.

     

    So I'm just going to switch my share screen and hope that this works. And you will get a bit of an insight to my Instagram here because I couldn't figure out how to get the video off there so just bear with me for one second and hopefully this will include the sound. Sexism in the workplace.

     

    Of course, sir. Defining the resolution. Like a human.

     

    Sexism in the workplace. More words, more words. Rhetorically speaking.

     

    Point evidence explained, goddammit. Succumbing to the fact that you sound like AI. I am AI.

     

    This will never get past an AI tracker. This is a bad policy. Hypothetically.

     

    Conclusion now. Conclusion. So hopefully that was visible you could all see it.

     

    And I hope you got a laugh out of that as well because I find this hilariously accurate and accurate representation of how we tend to use these AI or sin GPT at the moment, you start getting an answer immediately, but then you want to add more text to it constantly because you realize there's more and more things that you want from that. And this is one of the massive benefits, also of AI, is that it enables you to kind of get your, your thoughts together so it offers inspiration if you don't have those thoughts together also. And I feel like it can also take the stress away from trying to come up with that starting point, because as you're prompting it that starting point will will start to become clearer.

     

    But the efficiency and speed as well I feel like this can really play the role of an editor, as well. So, once you've created your content, you can quickly put it into something like sin GPT and ask for it to be cut down. You could ask for it to be checked for spelling areas, spelling errors, sorry, to make it more succinct, or to make it more SEO friendly so this is a process that would probably take a couple of reviews for yourself but also maybe an initial editor so the efficiency and speed in which you can create content has definitely changed with AI, but it also allows you to make more consistent content so it can maintain a uniform tone and style across all content.

     

    And it could be all content on the internet or all content that you're writing for your particular BU business unit. It also allows for content enhancement so it can suggest improvements for clarity or check tone, you can ask it to make something more exciting which is always good for the HR folks, or more sympathetic which we often use in it when we have to let people know we're about to delete all their data, or they were going to ask them to learn a new tool. Also the tone is a really really good one for communicators because you're often going from project to project and a lot of the times they don't have a lot to do with each other.

     

    So this will allow you to say to really switch easily between those two things without having to think so hard about a different kind of vocabulary set. So for example, in my job for example I have one project that is really fun tone. And so I would do something comms wise where I'm making it really fun and then I'll quickly have to switch to something that is about sending out a comms where we're trying to ask people to delete all their data, free up disk space.

     

    And this can't be so fun. This has to be a little bit more more serious. And so this can really help to to get you on track to make those those distinguishes can also help optimize the readability as well.

     

    So to really check is this paragraph too long is the other words too complicated, particularly if there's non native English speakers. This will also help get your SWOOP your SWOOP health scores up so really good one to use as well for the analytics in the back can really help with productivity I mean this is an obvious one, it can really reduce time and spend agonizing over the wording that you're trying to find, especially if you're just trying to come up with super smart words all the time to impress your stakeholders, you might agonize agonize over some of those words. I know I do it in emails, particularly if you're emailing leadership.

     

    So this just helps you to find things quickly that are probably already there but in the back of your mind, but you can't quickly get to them because you have so many other things on your plate. Quality controls also an obvious one and this AI capability already exists across tools like Microsoft Office with the spelling check and the grammar check, but it just allows you to do it in one bulk action when you're actually creating the content cost effectiveness. This is also an obvious one because it links back to that productivity.

     

    It's just less time on the basic content. It's also going to help you to go through old content and quickly repurpose it without having to spend hours and days doing that as well. So if you think about the amount of time you might have spent trawling through your internet sites trying to bring them up to date trying to get in touch with all the content owners of those pages to ask them to do so to be able to offer a tool like this.

     

    And then go to your page owners and say you need to update your page but don't worry it's only going to take you 30 minutes because you can get some help from sin GPT, then this is a task that people are now more likely to do. So the cost effectiveness really refers to kind of money but also resources and time time effectiveness. Anyway, let me quickly get onto the exciting stuff which is showing you some examples about how exactly is working for us.

     

    The first thing is on that chat side that I mentioned is really getting the prompts, right. So the prompts are what you put into the chat to get the output that that you want and there's a number of different models that we use so it depends on which one, what kind of tasks you're looking for but if you're just looking for some text or for it to summarize something, then there's a kind of a pattern that we go through for the prompting that we took our users through so that they would get really good outputs and this is a little framework that you can see here that we used. I won't go into the details of it too much because you will get the slides.

     

    But it's just a quick overview of how we've taught people how to use it. And then we also using it a lot for image generation. The image generation is actually really great for internet content, because it's harder and harder to find images that are different and intention grabbing and we're lucky at Syngenta we actually have a library that we use where we have a lot of professional imagery so we have photographers come in.

     

    They go out to fields they take photos and come into our offices and they take photos. Sometimes these photos are overused so you see the same photo, a lot, if you're talking about a field or if you're talking about an office. So what this allows you to do is create completely bespoke images for your article based on a prompt.

     

    And there's a little prompt sample here because this isn't the easiest thing to use there's still a lot of work that the models need to do on this particularly around words that you can't have words in an image really. But this really really helps to just grab attention on your internet content and make something bespoke to you. It's not so good for using for images for external communications, but you can do that as long as you just classify that to your audiences that AI was actually used to create that image.

     

    But here's a quick example of an image I used I created for today. I asked it to create a image of the SWOOP conference so that's why everything is in red, and you can see these little virtual images, it's pretty good those images actually look like people. And when we first launched SWOOP in Syngenta probably about eight or nine years ago, we ran a campaign, because we were saying that SWOOP is going to look at the analytics.

     

    And so not to scare people we said it's SWOOP, not snoop, as in snooping on someone. I added that into this image but the depiction of SWOOP wasn't really appropriate for a business conference but I think you can kind of get the idea a little bit there, where the image is cut off. With the images though you also have to use your own judgment so this comes back to what I was saying before about having that emotional intelligence that human oversight so really really key skill when you're working with AI.

     

    And if you just trust images on their face value, without thinking too much about it then you could get yourself into a lot of trouble. This is an image that I generated for an article we were doing about women in tech. So it's a bit of a community that we have here in Syngenta, and I wanted to have an image of men supporting women in the workplace.

     

    This isn't an image you can find on Google images either very easily so that's why I created it, but the image that came out didn't really look like that at all. Actually it was going against everything that women in tech actually stands for. So if you didn't have that human oversight, this could have probably got us into a lot of trouble so really important to always have that when you're using AI, particularly for content creation whether it's an image or words.

     

    The document translation tool that we have also enabled. And again this is just based on the LLMs from OpenAI so from ChatGPT, keeps the same format as well so if you upload a PDF, and you get it translated into any language you want, and there's over 100 different languages in there so it kind of meets meets our need for our 98 locations that we have, but it translates it and it keeps it to the same format so it's very very not only cost efficient but it's also very easy to use. It's not only cost effective, but time saving as well.

     

    This works in probably about 30 seconds flat. So you can imagine a huge cost saving against that to the business also. We can also use it through prompting and again you can do this with ChatGPT so if you don't have an internal instance like us, you can actually use a really clever prompt to get it to create PowerPoint presentations for you.

     

    And this is easily done by just generating a VBA code so visual basic code that is the underlying code of all Microsoft Office 365 apps. So, all you have to do is come up with what the presentation should be so you have an outline what's the topic going to be what's the audience going to be the purpose. We use something in Syngenta called a PO3 so where we have a document we do before every project that says what's the purpose what's the objective, what are going to be our outcomes and what are going to be the outputs.

     

    You can put that into a prompt like this, what are the key points and messages number of slides, etc. And then you can even ask it for some image suggestions, so that you can go and look for some images either on Google or get them created as well. So you ask for the code, and then basically you have to go back into your PowerPoint presentation and run this as a macro, and you can see here in the output, it will come up with a presentation for you.

     

    So if you want the step by step instructions for that you can ping me during the session I'll send them to you but this is a huge time saver, and allows you to quickly make presentations out of maybe an in depth article that you're putting on the internet or we see a lot as well as people create SharePoint pages, full of information that could have been a PowerPoint and easily distributed not taken up too much space as a site, but anyway. The other thing I wanted to show you guys with the knowledge bases, and I know there was a poll earlier in the session that Sharon launched about whether or not people were using chatbots for the internet. We this is something that we've been experimenting with so we have it in kind of two ways.

     

    The first way is we have a knowledge base which is part of our sin GPT platform, and these knowledge bases are basically linked to a SharePoint site, and you can ask that knowledge base any question about the content that is on that site so it pulls the information from that site as its as its knowledge. The downside to that is a lot of people don't keep their internet sites up to date, as you guys probably know, and the way that it's working is not indexing it by by date is looking for those patterns so you have to make sure that it's complete the information on the site is so up to date in order to give an accurate answer. Also the knowledge basis it's a different way of prompting it's more asking a question than it is asking a task, but we've extended this to our internet environment which is all based on SharePoint out of the box, where we actually have the chatbot, which is the knowledge base sitting over the top of the site.

     

    So you go into the site and this automatically pops up, and you can see this in the screenshot, and it just says ask me a question about any of the content that is on this site. And then you can ask a question it'll give you an output and you can basically go back and forth on that. So it's working relatively well as I said it only works on sites that are really well up to date or that have actual content managers for that site so we're experimenting with that at the moment, and it is something that we'll probably see a lot more of in the future.

     

    Talking about using sort of AI capabilities and SharePoint there are some that are already built in, and not sure if anyone has been experimenting with agents, this is something that Microsoft rolled out to kind of trial around with for free until June, and it's basically what I just showed that kind of chatbot. So you can build these yourself and put it over a site and they'll do exactly the same thing as the knowledge base it would just access the information that is on that site and you can then ask questions about it. So, why is this a game changer I've got a little use case here.

     

    We have a lot of sites that have a lot of scientific information on there they reference a lot of external material. Those sites are really good for the wealth of knowledge that they have but if you need to go to that site and pull information really really quickly to go answer a question from the media or a question from a stakeholder or you quickly need just a little bit of data to put into a presentation, rather than trawling through the site and try and navigate through it, you can actually just ask the agent to pull this information for you, and it will do that. So it provides immediate access to some of those points of data that might be buried in the site it's real time retrieval as well so you don't have to wait too long to get this, it's just quickly referencing that material and it's instant translation.

     

    If you need it in another language which is also a really really good capability. So some of the other SharePoint AI capabilities that you can leverage that everybody should have if you're using SharePoint, and you probably don't even know that this is AI powering this because it's been there even longer than we started to talk about chat GPT. And that's things like the smart content discovery so using the AI in the background to show you content that is most relevant to you, based on things like your, your job title your business unit your location.

     

    So, that's if you have your SharePoint connected to your ad systems obviously. But this is how it also gives you the customized news and content feeds. Also, depending on how you built your internet you will have content that is already targeted to you.

     

    This will show you content that is similar to content that you might have read before that's targeted to you topics that you might be interested in or writing about or, or whatever. There's also these inbuilt automated workflows and these are particularly good for things like the document libraries, if you want to be notified when something changes, or someone deletes a document, this is something that we always encourage the use of. And this is basically an AI automation that is built into the tool and that's there available for everyone without these other fancy tools like send GPT.

     

    But as I alluded to in some of the other slides before using any of the AI tools for content creation, or even sort of posting the outputs of it. It is important that you are aware of any kind of code of practice, or any guidelines that your company might have. And this is why we actually stopped the use of something like chat GPT because it was external and we couldn't really control it.

     

    Chat GPT also learns from you, which means it can keep your data to train itself and then basically reveal that data to others, which is why we have an internal system that is locked down safely within our Syngenta environment. But we still have these guidelines and what we call a manifesto for people to help people when they're using AI and it is basically things like I mentioned, all throughout the presentation about having that human oversight so not just trusting the first answers that it brings you but actually reading it through. You also probably need to spend a bit of time editing content that it comes up with as well just to check that it is in the kind of right tone as well if you've not prompted it to give you a certain tone.

     

    You always need to have this human oversight just to make sure that it's also not having any biases or hallucinating. And you need to make sure that everything that you're doing is fully compliant with any kind of legal framework that's out there and we have one in the EU specifically which is the EU AI Act, and that's to ensure the safe use and the ethical use as well. So we have this which really guides our users.

     

    It's not intended to put them off from using AI, it's actually meant to make them feel safer when they're doing it, because these safeguards are automatically built into our systems. But we also make people aware of it and we do that with some mandatory trainings and part of those trainings actually are part of a bigger program and you'll see that my background safe you can see the little logo. It actually says amplify better we have a whole program that's part of our work better campaign, where we are constantly upskilling and driving adoption of AI functionality across all of our tools, but particularly in how you can create content with AI safely as well whether that's content for your, for your internet or for other areas.

     

    But this is really a key pillar of our training and adoption wider initiative that we have in it where we give weekly trainings and targeted trainings to specific use cases and specific business areas. We do mandatory trainings as well once a year where we really teach people about AI how they can leverage it, what are the watchouts, etc. And these are learning paths that we have existing where people can go learn more about it, and basically get more comfortable with learning it as well.

     

    There are a lot of predictions out there about how much productivity gains companies can actually get out of leveraging AI tools like this. And I think it's it's at about 30%. You may see reports also that say a lot of companies haven't actually realized that, but that's also because the adoption is actually still going on particularly internal inside companies where people have a little bit of fear still, and it really is down to the kind of training and adoption that needs to be done.

     

    So we have a lot of awareness campaigns that we do as well and actually this may we have dedicated a whole month to just doing AI training so we have those every single Wednesdays. That's what the gift is on the screen as well. And we have lots of resources and information so we have a whole site dedicated to this area where we actually have all the case studies all the benefits that I went through with you today as well and a lot of different examples.

     

    We have those frameworks set up for how to write good prompts as well so that people can very quickly get into the tool without too much training it's really a learn by doing kind of tool anyway. And here we always try to keep on top of the trends as well to really to really drive that adoption. So, this is just a little bit about the work better campaign that I mentioned for which amplify is one pillar.

     

    And that's where we do all that AI trainings, but we won't go too much into that. I wanted to finish with the future of the internet. And so I had some pictures created about what I would hope the future looks like so that we humans would just be sitting in a nice park or sitting by the pool, and some robot somewhere is going to be doing all the work and when they're finished they just come to us, and we check it while we just have a little drink break.

     

    And then we never have to do anything again but relax and actually enjoy being human and and living life, but actually that's probably not what's going to happen. And it's not anytime soon anyway. So, what is the future of the, the internet with the introduction of AI or the adoption of AI as well.

     

    And I think if you look at it from thinking about all the problems that you currently have with the internet, or maybe they're around adoption or keeping content up to date. If you think about it from a problem point of view then it's probably easier to then think about how AI could actually help you to solve that, rather than the other way around so thinking of the business problem first and then the tool that could help you. But some possible ways that I could come up with is a shift away from content repository to more of a dynamic hub so internets are likely to move away from being kind of static repositories of documents and news stories and come to be a more of a dynamic place where people are going for more collaboration, communication and knowledge sharing and the knowledge sharing would definitely get easier with things like the knowledge bases and agents as well.

     

    It's going to make it a lot easier for people to discover the information that they need more quickly, but it's also going to personalize it for them as well. So the internet won't be this thing that is an annoyance to some people. I mean I still get emails asking me for particular content that I might have shared and I'll say well I just can't be bothered looking for it can you just send me the link or I know you shared this with me before but I can't find it.

     

    So, we know that internets are like this because they're just so full of information and not everyone is an SEO expert and not everyone has filled in their metadata, and this is where AI can really really help with internets is just getting people to the information that they need more quickly. The AI powered search and knowledge discovery is kind of already there and that SharePoint search and that's showing you things that it thinks are most interesting to you. It's now including things like information that is in your emails and information that is in your chats.

     

    And this is probably going to be the way that people use internets more in the future is not even going to the homepage but going to the search box because they're mainly going there as a starting point when they're looking for something, rather than, I'm just going to go there and have a quick browser than you. That's how that's using our company anyway. The automated content creation and curation that I kind of went through today.

     

    This is something where AI is really going to change internets. I think the watch out is that we will just create too much content because it becomes too easy. But the flip side is that you're going to be able to create really good bespoke interesting funny sympathetic whatever kind of content that you need to do and you're going to be able to do this in a way that that that is just easier for you to write but easier for people to consume as well so things like making things shorter not over explaining things, making it more succinct, maybe doing it with storytelling through images, for example, or maybe using AI generated videos which is another tool that we're experimenting with as well.

     

    Instead of having so many long form news articles. I won't go into it too much I think I'm almost out of time. Just quickly go to some of the watch outs and I mentioned through the session but the trust and transparency is really really important, particularly if you start using it for a lot of internet content.

     

    You kind of don't want people to know it's AI if you think back to that video when he was asking the guy that was playing the AI that make it sound more human. If it starts sounding like it's really AI generated because you haven't had that human oversight, then people will switch off to it. So you really got to make sure that you have the tone right that you have the human oversight, and that is not using words that are really really common like for us it's something like thrilling if I see thrilling written somewhere I think yes someone's used in GBT, not that that's a bad thing.

     

    It's not cheating or anything like that it's actually very smart because they save time, but it just means that maybe they've not gone and had that human oversight. So being aware of also which model is being used when you're asking it for information because some of the models that we have with the last update was kind of June last year. Some models are getting a lot better and they'll actually reference internet internet sources so this is getting better but good to be aware of that.

     

    Have a code of practice like I mentioned, always check for the accuracy and the reliability as well. Check the quality data security is important when you're using external tools because as I said these systems will train themselves on your data. You can actually have an opt in and opt out for that on chat GBT now but still some people are not fully aware of what they're going to do with that data, and then always remember the end user experience as well.

     

    This is key with anything to do with internet as long as whatever you're doing. If the whole purpose is improving that end user experience, then, then you can't really go wrong. So, with that I'm going to leave it.

     

    I think I'm over time sorry Pete. That that's okay it's great content. Thank you so much for sharing it.



Meet the speaker:

 

Melinda Schaller
Global IT Capability Lead, Strategy & Operations

 


Previous
Previous

Q&A Panel (All EMEA speakers)

Next
Next

Designing an intranet experience that’s unique to each user