WEBVTT

1
00:00:00.000 --> 00:00:01.410
Carolien Dekeersmaeker: the recording,

2
00:00:02.110 --> 00:00:17.919
Carolien Dekeersmaeker: There we go, very shortly. And we will be sharing that recording afterwards as well, so if you want to rewatch some of the webinar, or you want to share the presentation with some others, we will make that available, on our website, and also on the Thought Farmer Community side.

3
00:00:18.480 --> 00:00:28.680
Carolien Dekeersmaeker: If you have any questions at all during the presentation, we will have some time for Q&A after we finish the presentation and demo.

4
00:00:28.680 --> 00:00:52.519
Carolien Dekeersmaeker: But feel free to ask any questions at any point in time while we're still presenting. There is a Q&A, box in the, in the toolbox, the toolbar there, of Zoom, so go ahead and post that question there, and then we will… we have a few Thought Farmer people standing by who can already answer questions, but otherwise, we will, pick those up, near the end

5
00:00:52.520 --> 00:00:53.559
Carolien Dekeersmaeker: Of the webinar.

6
00:00:54.230 --> 00:00:58.589
Carolien Dekeersmaeker: And with that, Darren, I will pass it on to you to get started.

7
00:00:58.590 --> 00:01:22.079
Darren Gibbons: All right, thanks, Caroline. My name is Darren Gibbons, I'm the president of ThoughtFarmer, and before we get into the demo part of the presentation, I'm going to just cover a little bit more information about our approach to AI, talk a little bit about, you know, how we see… we want to try to build responsible AI features within ThoughtFarmer. Caroline's going to give that live demo, and we're going to wrap up there with the Q&A.

8
00:01:22.360 --> 00:01:38.070
Darren Gibbons: So yeah, so the next couple of slides here, we're just going to get into our overall approach to AI. Now, what we're focusing on right now is building AI tools inside ThoughtFarmer. We want to try to use AI throughout the product and really try to make it as useful as possible.

9
00:01:38.070 --> 00:02:02.029
Darren Gibbons: So right now, there are two features that are already enabled there in ThoughtFarmer, AI Create and AI Enhance. AI Create helps you to create new content faster, and AI Enhance allows you to improve the tone and formatting and structure within your content. So those features are there in place today. If you're using them, that's great. If not, you can go in, you can activate those, and take advantage of them. They show up in

10
00:02:02.030 --> 00:02:05.040
Darren Gibbons: inside the edit page experience inside ThoughtFarmer.

11
00:02:05.290 --> 00:02:29.560
Darren Gibbons: Now, for the rest of the presentation today, we're going to be talking more about the AI Assistant, which we are, recently launched, it's actually now available on your internet, but it is disabled. We're launching it, we're calling it a Beta, and what we mean by when we say beta for this AI Assistant feature, what we're referring to here is, it's like we're still under active development, we're improving it, we're adding new features, we're helping to make it answer questions better.

12
00:02:29.700 --> 00:02:47.540
Darren Gibbons: But we really want to try to get it into the hands of our customers so they can start using it. We've actually been running this beta for several months now, for over 6 months, and we've been improving it, we've been getting some great feedback from customers, and now we want to try to get it broader, into a broader release for our customers to use.

13
00:02:47.540 --> 00:02:56.069
Darren Gibbons: As part of the beta, one of our goals here as well, too, is to finalize our approach to pricing for AI. I'm going to talk about that later in the presentation there as well.

14
00:02:56.910 --> 00:03:18.869
Darren Gibbons: So let's get started here, and just talk a little bit more about our AI Assistant. And the way AI Assistant works is it's a chatbot, and it shows up on every page on your internet. And there's a couple of key pillars here, benefits that the AI Assistant provides. The first one here is trusted answers with sources, so it's based on the content from your internet.

15
00:03:19.220 --> 00:03:34.760
Darren Gibbons: It allows users to use plain language. So a lot of times with searching, you have to be a little bit more aware of getting the right keywords in place, whereas the chatbot, it has the ability to understand human language better, and so it provides that plain language discovery feature.

16
00:03:34.760 --> 00:03:58.599
Darren Gibbons: In addition, the chatbot is great for surfacing information that's embedded inside long content, so it understands large policies and that type of thing. And lastly, we want to try to make sure that we have those enterprise controls in place that respects the site permissions, but also gives administrators controls to, you know, the content that's being used and how it's being rolled out inside that organization.

17
00:03:59.550 --> 00:04:17.080
Darren Gibbons: So, these benefits, they tie back to some friction that we see employees feeling today. So, taking a step back here, you know, employees, they feel information overload, there's so much content that's out there, they don't know where to use it and how to use it. And if you're going to be using search, it requires knowing the right keywords.

18
00:04:17.209 --> 00:04:25.829
Darren Gibbons: To be able to find that content. That content, as I mentioned, is locked into long, complex pages, and people don't necessarily know where that content lives.

19
00:04:26.240 --> 00:04:42.940
Darren Gibbons: And so this is where we see the role of AI in the workplace as a helpful tool to enhance human capability. We want it to help make work clearer, faster, and help users feel more confident in their jobs, while keeping people in control.

20
00:04:42.940 --> 00:04:48.459
Darren Gibbons: So our goal here is that we want to use AI to help make work better, responsibly.

21
00:04:48.750 --> 00:05:11.320
Darren Gibbons: So, I'm going to use this… we've used this term, responsible AI, quite a bit, and I think it's important to just kind of define, like, what we mean by responsible AI. And this, for us, it's clear rules that reduce the risk of using AI tools. And a lot of our customers, you know, they have questions about what these AI… how we've built these AI tools and how they work.

22
00:05:11.320 --> 00:05:25.419
Darren Gibbons: And so that's something that we've spent a lot of time over the past year trying to understand what people are looking for with AI and how we can make that work better. And so we put together a bunch of clear rules. These include things like zero model training.

23
00:05:25.420 --> 00:05:32.789
Darren Gibbons: Private and secure by design, permission-aware access, transparent answers, and deeply integrated into our product. It's not just bolted on.

24
00:05:32.900 --> 00:05:39.440
Darren Gibbons: And so, during this initial beta, when we had just a smaller group of customers using it, I've spoken with a lot of.

25
00:05:39.560 --> 00:06:00.389
Darren Gibbons: technology people, a lot of business people, a lot of security people at organizations, and they have a lot of questions for us about how we've built, built our AI tool. And so, ultimately, you know, we're going to walk through here now some of those questions that we've seen. And this is a good checklist that you can use when you're talking to your other vendors as well, too.

26
00:06:00.390 --> 00:06:22.679
Darren Gibbons: We recognize that a lot of organizations, a lot of software companies these days are adding AI features. We've put a lot of thought into how we build these features and how we integrate them into our product so that our customers can trust them. This is a good vendor checklist questions that you can take and turn around and ask your vendors as well, too, as to how they are integrating AI into their products.

27
00:06:23.330 --> 00:06:26.839
Darren Gibbons: So, I'm gonna run through these questions here now.

28
00:06:27.150 --> 00:06:31.880
Darren Gibbons: First up, will your AI train on our data?

29
00:06:31.880 --> 00:06:55.740
Darren Gibbons: And the answer to this is no. So we don't do any kind of training or reinforcement learning with your data. So this is kind of a question that often comes up with AI models, because AI is trained on information that's on the internet, and a lot of people are concerned that potentially that your private, sensitive corporate information could be used to train the model. We do not do any kind of

30
00:06:55.740 --> 00:06:58.229
Darren Gibbons: Reinforcement learning or training with that data.

31
00:07:00.220 --> 00:07:19.099
Darren Gibbons: Next question here is just around, logging of data. So we are using… ThoughtFarm is hosted on AWS, and we use the AWS AI models. They provide a framework called Bedrock, AWS Bedrock, and that's what, those are the models that we're using, and within that

32
00:07:19.770 --> 00:07:36.610
Darren Gibbons: AWS Bedrock, we're using the Anthropic models currently. We may change that in the future, but those are the ones that we found have worked best for us. So a question that we often get is, are we… does AWS perform any logging or persisting of any of the customer data that's sent to these models?

33
00:07:36.610 --> 00:08:01.380
Darren Gibbons: And the answer to that is no. All requests to the AI models are processed in memory only, they're not logged by AWS. Having said that, ThoughtFarmer itself does still log user requests and the chatbot responses, and it's stored with your customer data, because we find that's really useful for customers to be able to take a look at and see what are the questions that people are asking in the chatbot, how is it responding, is it accurate, how can

34
00:08:01.380 --> 00:08:07.589
Darren Gibbons: we make it smarter. Caroline's gonna demo some of that logging feature, later on in the live demo.

35
00:08:08.640 --> 00:08:25.810
Darren Gibbons: Next question here is, can the requests to the AI models be intercepted? And the answer is no. All the requests to the AI models are made across private AWS subnets. They're encrypted in transit, so there's no risk of a man-in-the-middle style attack there.

36
00:08:26.990 --> 00:08:29.630
Darren Gibbons: Next up, our…

37
00:08:29.680 --> 00:08:38.260
Darren Gibbons: customer… multiple customers' data stored together, and the answer to this is no. So the way we've architected Daw Farmer is we keep all of our customers' data.

38
00:08:38.260 --> 00:09:02.159
Darren Gibbons: separate in separate databases, database instances, and separate Elasticsearch indexes. So Elasticsearch is what we use for the search portion of both the main search as well as the chatbot. We don't co-mingle customer data, so your data is kept separately, logically separated, at rest and in our systems, and so that gives you some confidence that the risk of information

39
00:09:02.160 --> 00:09:04.270
Darren Gibbons: Leakage or sharing is minimized.

40
00:09:05.480 --> 00:09:11.429
Darren Gibbons: The next question here is around data processing and storage.

41
00:09:11.430 --> 00:09:28.120
Darren Gibbons: So this one here, I'm gonna walk through this here. There's a lot of data here on the screen. So we maintain 3 different data centers, and all of them are identical, except for a slight variation around the AI processing, and I'll walk through that now.

42
00:09:28.120 --> 00:09:32.749
Darren Gibbons: So our first, our data centers are in US, Canada, and EU.

43
00:09:33.310 --> 00:09:57.720
Darren Gibbons: And if you take a look there at the table, you can see storage at rest, this is where your data is stored. If you're in the US, data is stored at rest in the US. Same for EU, same for Canada. For the system processing, this is, like, the core ThoughtFarmer functionality, that whole processing is also done in each of those data centers in US, EU, and Canada, so that's exactly the same.

44
00:09:58.130 --> 00:10:04.200
Darren Gibbons: Now, for AI processing, this is where things are a little bit different. So AWS provides

45
00:10:05.750 --> 00:10:28.770
Darren Gibbons: the way that they make their AI models available is through something called cross-region inferencing. And what that means is… is that AI processing, if you are in our U.S. data center, it will be… the AI model will live somewhere in the U.S, they don't specify where, they load balance across different regions inside the U.S. for doing that AI processing.

46
00:10:28.960 --> 00:10:51.809
Darren Gibbons: In the EU, it's the same thing. So they have a cross-region geographic inferencing models available in the EU. In Canada, they do not currently have a geographic cross-region inferencing model available for us to use. So in Canada, we use the global model. And so that means that the requests, they may actually, they could potentially be handled in Canada.

47
00:10:51.810 --> 00:11:04.929
Darren Gibbons: They'll most likely be handled in the US, but we use this global model here. But we do still have all those other protections in place around no information as logging, everything's done in memory.

48
00:11:04.930 --> 00:11:06.889
Darren Gibbons: And so you can feel more confident with that.

49
00:11:07.180 --> 00:11:22.569
Darren Gibbons: Now, one aspect of this is because of our contracts, the way that we write our agreements is we state in the contract where your data is going to be processed. So if you are a Canadian customer, you will need to sign a contract amendment. It's the AI addendum.

50
00:11:22.790 --> 00:11:46.029
Darren Gibbons: We have sent that out, and you should have it in your inbox. If you have any questions about it, or if you'd like to review it, or you'd like to talk to us about it, please feel free to reach out. I know a number of you have already received that and signed it, and thank you. So if you are hosted in our Canadian data center, which would be our Canadian and Oceania customers, you will need to sign that

51
00:11:46.030 --> 00:11:48.500
Darren Gibbons: Addendum before this feature will be made available.

52
00:11:48.740 --> 00:11:51.850
Darren Gibbons: So this is something that we're,

53
00:11:52.110 --> 00:12:09.009
Darren Gibbons: is not something specific to AWS. The other providers also face this, both if we're hosting with Azure or Google or one of the other systems. So pretty much all vendors that have Canadian data centers are going to be facing the same issue. So it's something we just have to work with there.

54
00:12:09.680 --> 00:12:27.619
Darren Gibbons: All right, moving along. Can the assistant leak private data? And, the concern here is that the assistant could answer questions about content on your internet that is private. The answer to here is no. The assistant will only answer with content that the user has access to.

55
00:12:29.220 --> 00:12:44.320
Darren Gibbons: Next up here, can the assistant hallucinate answers? So, what do I mean by hallucinate? So, when we talk about AI models, they are trained on massive amounts of content available from the public internet.

56
00:12:44.950 --> 00:12:49.949
Darren Gibbons: And, sometimes when you ask them questions specific about your business,

57
00:12:50.340 --> 00:13:13.360
Darren Gibbons: if they… they may, have a… they may hallucinate or answer that question based on content that is in their training data, which is not related to your… to your intranet. This is something that we've worked a lot to try to eliminate during the development of our AI assistant. We're currently, with our current approach and what we're rolling out here to beta.

58
00:13:13.360 --> 00:13:19.960
Darren Gibbons: We've not been able to reproduce any kind of hallucinations. Having said that, it is possible that it could occur.

59
00:13:20.460 --> 00:13:44.259
Darren Gibbons: So how do we prevent these? We try to make sure that we ground our AI models with the most accurate data from your intranet. So if we feed the model with that accurate data, the answers are going to be much better. And similarly, as time has gone on, you know, as you follow the news, it seems like every week there's a new, more advanced AI model that's available.

60
00:13:44.390 --> 00:13:57.259
Darren Gibbons: And these newer models are much, much better at preventing hallucinations, and we've seen that in our testing there as well. So the bottom line is, is that it shouldn't, but you should always make your users aware that

61
00:13:57.260 --> 00:14:11.160
Darren Gibbons: there's a potential for incorrect answers, and they should always… users should always check and verify. It's easy to do that with the chatbot, because we actually provide links back to the sources of that content, so users can double-check that.

62
00:14:12.770 --> 00:14:22.350
Darren Gibbons: Next up here, can we disable all AI features? And the answer is yes. The AI features are disabled by default, and you'll need to opt in to using them.

63
00:14:24.630 --> 00:14:39.950
Darren Gibbons: Next question, can we control who has access to AI features? And yes, you can. AI Create and Enhance, as well as the AI Assistant, can be enabled on a user-by-user and a group basis. You're in control over who has access to the AI features on your internet.

64
00:14:41.750 --> 00:15:04.180
Darren Gibbons: Next up here, can we limit which content is used for answers? So I know that one of the things we saw, again, in our early customer beta was that people were like, you know, some of our content, our internet is, you know, we know we need to clean up some of this content, it's maybe not as fresh, or other information that, you know, maybe not relevant, for the chatbot, for the AI assistant, and so that's something where, you know, can we

65
00:15:04.180 --> 00:15:07.859
Darren Gibbons: Hide that from the assistant so it doesn't answer questions based on that content.

66
00:15:07.860 --> 00:15:32.129
Darren Gibbons: And the answer is yes. So the AI Assistant only uses content that's been configured by an administrator for use by the assistant. So the way this is handled is when you're setting up the AI Assistant, you can create, topics, and topics is going to be, you can give them a name, and you can add pages to those topics. You can create up to 20 topics, and add up to 1,000 pages of

67
00:15:32.130 --> 00:15:42.970
Darren Gibbons: content. Caroline's going to show you what that process looks like in the live demo. Now, when we talk about pages, that includes attachments as well. One file attachment counts as one page.

68
00:15:43.060 --> 00:15:57.350
Darren Gibbons: And we take the text content from Word, PDF, Excel, and PowerPoint, and that's all available to the chatbot to be able to answer questions. So if you've got policy information or procedures that are in Word documents, that information there is available for it to use.

69
00:15:57.790 --> 00:16:15.080
Darren Gibbons: Now, a question that we've had in the early beta there is, it's like, okay, we've got this 1,000-page limit, is there a cost to increasing the page limit? And that gets into the slide question here around the AI assistant pricing, and how much will the AI assistant cost?

70
00:16:15.400 --> 00:16:16.220
Darren Gibbons: Now.

71
00:16:16.850 --> 00:16:31.890
Darren Gibbons: Taking a step back here, we feel that AI has a huge potential to help build amazing internet experiences, and we want to enable it as wide as possible for our customers, which means minimizing your costs. So we recognize that if we do have a high cost, we didn't want to have

72
00:16:31.890 --> 00:16:46.550
Darren Gibbons: AI as a paid add-on, because we knew that that was going to be… would be a limit for organizations to roll it out there. So we want to try to be able to provide, some AI features included in the cost of your subscription.

73
00:16:46.550 --> 00:16:48.970
Darren Gibbons: Having said that,

74
00:16:48.970 --> 00:17:13.089
Darren Gibbons: the way that AI processing works for us, it's very difficult for us to model what the pricing is going to use, because it's based on usage, it's based on tokens. If you've worked with any type of AI APIs, you recognize that it's a bit of a slippery thing to try to get a handle on in terms of pricing. So we need to have some smart limits there on it. So our plan for now, the AI and create an

75
00:17:13.099 --> 00:17:31.299
Darren Gibbons: enhance, are free currently. They'll continue to be free. The AI Assistant, you can go and enable it today. It includes your first 1,000 pages of content to be able to go into there and add, and we know that customers are able to use that to really get value there out of the AI Assistant today.

76
00:17:31.300 --> 00:17:36.869
Darren Gibbons: The next question is, is what'll be the pricing, for additional pages? We will be looking to

77
00:17:36.870 --> 00:17:54.829
Darren Gibbons: to announce that shortly as part of the beta period, when we end the beta period, but it's still TBD at this point. If you would like to go above the thousand page mark with your AI assistant, please reach out to our team. Use either create a support ticket for it, and we can talk to you further about that.

78
00:17:56.020 --> 00:17:57.590
Darren Gibbons: That ends…

79
00:17:57.720 --> 00:18:07.709
Darren Gibbons: my wall of text, and so I'm going to now hand it over to Caroline, who's going to be able to show you a demo of the AI Assistant.

80
00:18:08.750 --> 00:18:09.360
Carolien Dekeersmaeker: Awesome.

81
00:18:09.760 --> 00:18:27.869
Carolien Dekeersmaeker: Great. Thank you, Darren. And so, yes, I will pull up a demo site and start showing the AI Assistant momentarily. Before I do, I just wanted to reiterate some use cases or some ways in which, you know, you and your staff can benefit from the AI Assistant, and that goes back to some of the points that

82
00:18:27.870 --> 00:18:41.929
Carolien Dekeersmaeker: Darren made earlier around some pain points, right? Where there's overload of information, and people sometimes struggle to find correct answers to questions they have, amidst all of the content that's available on your internet, among other places.

83
00:18:41.930 --> 00:18:54.780
Carolien Dekeersmaeker: So again, like, how can… how can people use the AI Assistant on your Thought Farmer internet? It's like finding information or finding answers to questions they might have around, for example, onboarding. If you have new hires.

84
00:18:54.780 --> 00:19:05.690
Carolien Dekeersmaeker: that are joining your company, you have onboarding information and workflows available on your internet, the AI assistant can provide those new employees with quick questions, as quick answers to their questions.

85
00:19:05.690 --> 00:19:23.070
Carolien Dekeersmaeker: Anything that relates to your policies, HR questions that people might have, benefits, you know, customer-facing procedures and policies, all of those are really great to feed into your AI assistant to provide those quick answers.

86
00:19:23.070 --> 00:19:35.880
Carolien Dekeersmaeker: As well as the, the IT, information, or, you know, a lot of organizations have information on their internet around password management, security policies.

87
00:19:35.910 --> 00:19:43.460
Carolien Dekeersmaeker: you know, setting up the VPN, all of those, like, information technology, steps and workflows.

88
00:19:43.870 --> 00:19:54.599
Carolien Dekeersmaeker: can now be answered by the AI assistant, and maybe your IT help desk will, you know, have a few fewer tickets coming into them if people are able to find those answers quickly with the AI assistant.

89
00:19:55.210 --> 00:20:11.639
Carolien Dekeersmaeker: Lastly here, even if you, if you use ThoughtFarmer to, keep meeting minutes or, project information, the AI assistant can also help provide people with updates around those. Like, what's the latest status on project ABC, for example? So,

90
00:20:11.970 --> 00:20:18.319
Carolien Dekeersmaeker: Ultimately, really, what it does, and what it can do, is give your staff and yourself

91
00:20:18.470 --> 00:20:36.239
Carolien Dekeersmaeker: really quick answers to questions they have every day in the workplace, even if the answers of those questions are pulled from, you know, maybe a really long PDF file that is somewhere on the internet, right? It, like, synthesizes all of the pages and files on your internet and provides that really

92
00:20:36.930 --> 00:20:38.950
Carolien Dekeersmaeker: Short and quick answer.

93
00:20:39.130 --> 00:20:57.940
Carolien Dekeersmaeker: What are you going to see now, before I head into the demo? It's obviously a fake internet, it'll be a sample internet environment. And one thing I wanted to iterate, I know Darren mentioned that earlier as well, but, when you, or when a user uses the AI assistant, it will only ever give them information.

94
00:20:58.060 --> 00:21:16.200
Carolien Dekeersmaeker: from content that they have few permission on on the internet. So it's extremely secure. If you have content on your internet that is more secure and only accessible for a few people, rest assured that nobody who isn't supposed to see those will get answers from that content in the AI system.

95
00:21:17.040 --> 00:21:33.060
Carolien Dekeersmaeker: Okay, so without further ado, we're here on our fictitious internet side, and as you may have noticed, we have this cute little ninja robot guy sitting in the lower right corner. That is our AI assistant. First time I click on it, it's giving me a little welcome message.

96
00:21:33.120 --> 00:21:58.089
Carolien Dekeersmaeker: This is something that you can customize for your internet if you want to give your staff specific guidance on how to use the AI Assistant. Once I click on continue, it's going to allow me to start asking it questions. Again, there are, and I will demo this in a little bit, there's different… these are the pages or sections of information that Darren talked about, so you can kind of guide your users in telling them, like, hey, these are the types

97
00:21:58.790 --> 00:22:07.100
Carolien Dekeersmaeker: information sources that I can provide questions on, right? So we'll take a look at that in a moment. So for now, I'm going to ask our assistant…

98
00:22:07.220 --> 00:22:11.300
Carolien Dekeersmaeker: How much time do we get for parental leave?

99
00:22:14.720 --> 00:22:28.230
Carolien Dekeersmaeker: So it's, again, conversational, interactions with the AI assistant. It's gonna think and, like, look at all of that content that we have indexed, and then it comes up with a really helpful short,

100
00:22:28.230 --> 00:22:34.980
Carolien Dekeersmaeker: Sir, so illegible employees may take up to 16 weeks of parental leave. You'll see here these footnotes.

101
00:22:34.980 --> 00:22:52.609
Carolien Dekeersmaeker: That's what Darren was referring to as well. The AI assistant will always tell people where it is getting its information from. If you scroll down to the bottom of the answer, you… it will always cite what the sources are for its answer. So if I click on parental leave here.

102
00:22:53.030 --> 00:22:59.319
Carolien Dekeersmaeker: you'll see it leads me directly to that specific policy here on the internet. Okay.

103
00:22:59.570 --> 00:23:10.659
Carolien Dekeersmaeker: Again, this is just a disclaimer and goes back to something Darren said as well. We are… yeah, Assistant works really well and, like, shouldn't hallucinate at any time.

104
00:23:10.660 --> 00:23:22.549
Carolien Dekeersmaeker: But it… we do want to make sure that people know, you know, it is an AI functionality, it is an AI feature, and it can mistake… make mistakes, so always check the responses, always take a look at those sources.

105
00:23:22.610 --> 00:23:42.390
Carolien Dekeersmaeker: Every interaction that someone has with the AI assistant, they can let it know if its response was helpful or not. So, I'm gonna say, yes, that's exactly what I needed to know here. This feedback will become helpful for you as an internet manager, in terms of sort of reviewing how things are going with the AI assistant. We'll look at that in a little bit.

106
00:23:43.010 --> 00:23:49.190
Carolien Dekeersmaeker: Maybe I'll ask it another question. What do we use for password management?

107
00:23:52.430 --> 00:23:58.590
Carolien Dekeersmaeker: Again, the little robot is thinking, and is going to,

108
00:23:58.590 --> 00:24:13.699
Carolien Dekeersmaeker: now, in this case, it's looking at all of that information technology content on this internet, right? So you can see here, it says, SlotFarmer uses 1Password. And in this case, again, you see now it's actually surfacing multiple sources.

109
00:24:13.700 --> 00:24:27.389
Carolien Dekeersmaeker: Right? We have these footnotes here. If we go to the bottom, we can see it's referencing two pages on the internet. Our password and identity management page, our policy, but then also an instructional manual about how setting up

110
00:24:27.650 --> 00:24:41.310
Carolien Dekeersmaeker: 1Password. So again, it's pulling multiple content sources together to provide the user with the most helpful answer in what that person may need to accomplish in this moment in time.

111
00:24:42.630 --> 00:25:00.230
Carolien Dekeersmaeker: Okay, just going back here, one thing that I wanted to talk about, sometimes people are used to not conversing with an internet search or an AI feature, but they might just type in, just a keyword. If that happens, the AI assistant, even though it's a conversational,

112
00:25:00.230 --> 00:25:13.060
Carolien Dekeersmaeker: assistant, it will… it will ask qualifying questions. It will ask the user, what do you mean exactly, or what do you need exactly? So, just to give you an example, if I need to find our VPN information, connection information, and I just typed in

113
00:25:13.070 --> 00:25:14.360
Carolien Dekeersmaeker: VPN.

114
00:25:14.590 --> 00:25:32.319
Carolien Dekeersmaeker: And again, this is a little exciting, because it's a live demo, but okay, it did what I was expecting, and the AI assistant doesn't really know what to do. It's just like, okay, what would you like to know about the VPN? For example, are you looking for setup instructions or access policies, right? And so now I can say, setup.

115
00:25:33.850 --> 00:25:40.260
Carolien Dekeersmaeker: And hopefully that is gonna give it, more information on exactly what this person is hoping to get out of it.

116
00:25:40.640 --> 00:25:42.579
Carolien Dekeersmaeker: So, it's thinking…

117
00:25:43.490 --> 00:26:01.619
Carolien Dekeersmaeker: And here we go. Now, with that additional context, the AI system knows, okay, here are the steps to set up the VPN and how to accomplish that. So all the steps, this is, again, a really good example, right? The AI system looks at multiple larger sources of information, but

118
00:26:01.620 --> 00:26:03.569
Carolien Dekeersmaeker: it intelligently

119
00:26:03.570 --> 00:26:13.260
Carolien Dekeersmaeker: will condense the information to exactly what that user needs in that moment, a checklist of how to set things up, right? So, that's where the value lies.

120
00:26:13.500 --> 00:26:23.860
Carolien Dekeersmaeker: Okay, so let's take a look at how you set up the AI Assistant first time, right? So I'm gonna head into our admin panel here.

121
00:26:25.010 --> 00:26:28.510
Carolien Dekeersmaeker: And, you will now see a new

122
00:26:29.100 --> 00:26:41.139
Carolien Dekeersmaeker: link under integrations. Again, we are focusing on AI Assistant here for the demo today. AI Create and Enhance is available in a different page. That's not something we'll be talking about today.

123
00:26:41.140 --> 00:26:59.550
Carolien Dekeersmaeker: So this is the AI Assistant Admin page. When you first hit this page, it will be… the AI Assistant will be disabled. So that is, of course, the big checkbox you have to… or the big toggle you have to turn on. My recommendation is to leave it disabled while you go through all the setup steps, and then once you're done with that.

124
00:26:59.750 --> 00:27:03.959
Carolien Dekeersmaeker: Toggle it to enabled, and then it will start surfacing to those people you've given permission.

125
00:27:04.290 --> 00:27:23.849
Carolien Dekeersmaeker: All right, so the settings tab, just really quick, a few things you can do to customize your AI assistant. You can give it a custom name, and you can also change the, the icon or the avatar that it uses. You can really have a lot of fun with this. I know some of our customers have,

126
00:27:25.260 --> 00:27:46.210
Carolien Dekeersmaeker: a very specific name for their internets, oftentimes with a mascot. I know there's… there's otters and birds and, you know, 19th century company… century company founders out there that are mascots for their intranets, so you can, update the avatar with their likeness, if you want, and give them a specific name.

127
00:27:46.320 --> 00:27:56.299
Carolien Dekeersmaeker: And then in the welcome message here, that's that first screen people are going to see. You can customize this message entirely, to better serve your staff.

128
00:27:57.760 --> 00:28:10.050
Carolien Dekeersmaeker: In terms of organization terms, this is a helpful feature, again, to… it's something that you can do to help your AI system provide even better answers. What it is…

129
00:28:10.150 --> 00:28:21.859
Carolien Dekeersmaeker: a way to provide context for your AI assistant around your organization. So, the organization name is something that's really important. Obviously, you want to make sure that your AI assistant

130
00:28:21.860 --> 00:28:43.699
Carolien Dekeersmaeker: knows what your company name is, because people will reference that in their questions. So for us, of course, it's Thought Farmer, but we also included TF, because internally, we usually, you know, shorten it to TF. So if that's something you do, just feed it all the possible names that your staff might refer to your organization, and then if you use any acronyms internally.

131
00:28:43.710 --> 00:28:55.909
Carolien Dekeersmaeker: That might not be, that might not be used on pages of content itself, or things that people might use an acronym to refer to a longer term, like PTO, paid time off.

132
00:28:55.910 --> 00:29:07.150
Carolien Dekeersmaeker: Add that here as well, so that the AI assistant can make those connections. And if somebody asks, what's my PTO, and the page only talks about paid time off, it will make that connection and provide the answers.

133
00:29:07.150 --> 00:29:13.469
Carolien Dekeersmaeker: So again, you can add as many, specific terms here as you feel will help your AI assistant.

134
00:29:14.960 --> 00:29:28.680
Carolien Dekeersmaeker: Okay, great. So that is the settings. In terms of permissions, like Darren mentioned, you can decide who should have access to asking your AI assistant questions, right?

135
00:29:28.680 --> 00:29:36.900
Carolien Dekeersmaeker: And so you can get very granular in here. On this side here specifically, we've chosen to give access to everyone, all registered users.

136
00:29:36.900 --> 00:29:49.589
Carolien Dekeersmaeker: But you can customize this if you want to, say, start with a smaller group, or a specific location or department to test out the assistant before rolling it out, you can do that as well. So if I type in

137
00:29:49.720 --> 00:30:01.739
Carolien Dekeersmaeker: human resources, for example, I can check that one, so only people in our HR team have access to it. I can also give specific, people access as well, so it's super flexible.

138
00:30:02.400 --> 00:30:21.669
Carolien Dekeersmaeker: Click on save. And then the next step would be during your setup is to add topics, right? That's, like, essential, of course. You want to make sure that you add pieces or sections of content and information to the AI assistance, and those are the areas that it's going to be able to provide answers about.

139
00:30:21.850 --> 00:30:27.620
Carolien Dekeersmaeker: So in this one, in this case, we already have a few items set up, a few topics, as you can see.

140
00:30:27.620 --> 00:30:41.979
Carolien Dekeersmaeker: right here, at the top, like Darren mentioned, you can index, up to 1,000 pages of internet content into the AI system currently, and it'll give you an overview of where you're at. So right now, we have

141
00:30:41.980 --> 00:30:48.900
Carolien Dekeersmaeker: 79 pages that are indexed, and we have 921 available. Please note that every…

142
00:30:49.160 --> 00:31:04.099
Carolien Dekeersmaeker: page, every attachment is considered a page. So if you have a page that has 20 PDFs attached to it, that section would take up 21 pages, right? So each file constitutes a page as well.

143
00:31:05.140 --> 00:31:22.020
Carolien Dekeersmaeker: Okay, and so let's see how you can add a topic, right? So I can… maybe I want to add all of our HR information and policies to our AI system. So I'm going to click on add topic. I'm going to give it a description, so let me say HR information.

144
00:31:22.560 --> 00:31:39.049
Carolien Dekeersmaeker: And then, much as, like, when you use a news card and you add a new source, this works the same way as a find as you type field. So, I'm going to type HR, and then I'm going to find our HR department here. In this case, I'm just going to feed everything under that

145
00:31:39.180 --> 00:31:56.720
Carolien Dekeersmaeker: group page, make sure that when you add a new topic source, to click include include subpages, right? We don't only want the main landing page that says HR, we want all of the information underneath it. And so you can see it goes from one page to 75 as soon as I click that.

146
00:31:58.100 --> 00:32:03.539
Carolien Dekeersmaeker: Click on Save, it is now listed here in our list of topics.

147
00:32:04.390 --> 00:32:13.849
Carolien Dekeersmaeker: Right? So, that is now added, so if we would go back to our HR Assistant.

148
00:32:14.520 --> 00:32:21.130
Carolien Dekeersmaeker: There we go, and we go all the way back, we're now seeing HR information is a topic that is added here.

149
00:32:22.010 --> 00:32:38.290
Carolien Dekeersmaeker: The last thing I'm going to talk about, and this is sort of after you have launched your AI system and have been using it for a time, it's… it's really helpful to go back and, and look at how people are using it, both in terms of learning about

150
00:32:38.740 --> 00:32:43.269
Carolien Dekeersmaeker: your users? Like, what are they looking for? What type of information might they be looking for that

151
00:32:43.270 --> 00:33:01.469
Carolien Dekeersmaeker: maybe isn't available on the internet. But also kind of troubleshooting, right? Especially as you go and as you start to experiment with adding different areas of information into the AI assistant, you might discover maybe people are searching for information and not finding the answers because I didn't add that specific section as a topic.

152
00:33:01.710 --> 00:33:04.560
Carolien Dekeersmaeker: Right? So,

153
00:33:04.560 --> 00:33:22.689
Carolien Dekeersmaeker: An example here could be, and I'm just referring to… you see this? This is, like, this news article here on the main carousel. It talks about one of our executives who won the Executive of the Year award, right? Maybe it's something that people are interested in, so I'm going to

154
00:33:22.690 --> 00:33:29.310
Carolien Dekeersmaeker: Go ahead here and ask our AI assistant, who One… sorry, who won?

155
00:33:29.460 --> 00:33:31.880
Carolien Dekeersmaeker: Executive of the Year.

156
00:33:33.330 --> 00:33:40.060
Carolien Dekeersmaeker: It's gonna start thinking, it's gonna scour all of that information in the topics that we've added.

157
00:33:40.150 --> 00:33:46.890
Carolien Dekeersmaeker: And it says, I don't know, I don't have information about that, right? In this case.

158
00:33:46.960 --> 00:34:01.140
Carolien Dekeersmaeker: you know, the person would probably say, no, this was not helpful. Why? Right? Every time somebody gives a negative feedback on the AI system, they can leave a comment as to what they were expecting, or how it can be improved. So,

159
00:34:03.480 --> 00:34:11.110
Carolien Dekeersmaeker: So maybe I'll say it… didn't know… About Sylvia's…

160
00:34:14.000 --> 00:34:19.900
Carolien Dekeersmaeker: Award, even though there is a news announcement on that topic, right?

161
00:34:21.080 --> 00:34:27.909
Carolien Dekeersmaeker: Okay, I'm gonna go ahead and submit it. Now, on my end, if I am the,

162
00:34:27.920 --> 00:34:37.580
Carolien Dekeersmaeker: the, internet administrator, I will see this feedback in the AI Assistant logs. So if I go back into my page here.

163
00:34:37.580 --> 00:34:59.749
Carolien Dekeersmaeker: We went over all of these setup items, but the log one is one that you will want to revisit, regularly. So this is going to list every single interaction that people have with the AI assistant. And you can see here, for example, my… this is the one I just did, right? Who won Executive of the Year? At a glance, you can see

164
00:34:59.750 --> 00:35:12.909
Carolien Dekeersmaeker: all of the thumbs ups or thumbs down that have been added, and so if you want to sort of see, okay, why did Caroline give a thumbs down here? I can click on the request number, and it's going to give a…

165
00:35:12.910 --> 00:35:24.209
Carolien Dekeersmaeker: verbatim transcript of what happened, right? So it shows me what the user typed in, what the response was, and then what the feedback was, right? Now, in this case.

166
00:35:24.310 --> 00:35:28.239
Carolien Dekeersmaeker: What happened is those news announcements are not actually

167
00:35:28.410 --> 00:35:35.700
Carolien Dekeersmaeker: indexed in the AI topics, AI system topics, so I can do that here. I can say, okay, add a topic.

168
00:35:36.440 --> 00:35:38.950
Carolien Dekeersmaeker: I'll say HR Announcements.

169
00:35:39.930 --> 00:35:49.939
Carolien Dekeersmaeker: And then, I'll search for that HR news feed, where that, that announcement was made. Again, make sure to include subpages.

170
00:35:50.440 --> 00:36:01.899
Carolien Dekeersmaeker: Click and save, and now I can see this here in the list of topics. This happens instantaneously, so if I go back and open my AI assistant here, and I ask the same question…

171
00:36:02.180 --> 00:36:05.640
Carolien Dekeersmaeker: Who won Executive of the Year?

172
00:36:07.320 --> 00:36:21.269
Carolien Dekeersmaeker: It should now find that answer, because I've indexed those, that news feed. There we go, Sylvia George won Executive of the Year, and it's now citing, that news article as,

173
00:36:21.450 --> 00:36:23.509
Carolien Dekeersmaeker: As… as the main source for that.

174
00:36:23.630 --> 00:36:26.820
Carolien Dekeersmaeker: Right? So this is something that,

175
00:36:26.910 --> 00:36:42.819
Carolien Dekeersmaeker: we really recommend is, like, go into your AI assistant logs regularly, review the feedback that you're getting from staff, and, if something is not indexed correctly, or you're finding that people really are interested in

176
00:36:42.820 --> 00:36:50.119
Carolien Dekeersmaeker: for example, in this case, HR News Announcements, and it's not indexed, you can add that, and kind of, like, make the AI system better and better that way.

177
00:36:50.120 --> 00:37:07.189
Carolien Dekeersmaeker: Similarly, you might run into situations where, you know, everybody is asking about the mailing address for, your, Toronto location, and that's not actually on the internet, so it can inform you on which content to add to your internet as well.

178
00:37:09.400 --> 00:37:18.289
Carolien Dekeersmaeker: Alright, so that concludes the demo portion of, of what I wanted to show you guys today. Just hop in…

179
00:37:18.310 --> 00:37:27.149
Carolien Dekeersmaeker: back to our slide deck here, and really quickly to wrap up in terms of what is next. I mean, we are…

180
00:37:27.150 --> 00:37:42.750
Carolien Dekeersmaeker: really committed to our AI assistant and AI features and making it as valuable as possible. So we are continuing to make improvements and enhancements to it. In terms of what's coming up in the, in the short term.

181
00:37:43.110 --> 00:37:57.169
Carolien Dekeersmaeker: Right now, the AI assistant can answer questions about any type of sections or pages on the internet, but not information about people. That's going to be something that we add. So in one of our next versions, we're going to have an integration with our people directory.

182
00:37:57.170 --> 00:38:03.900
Carolien Dekeersmaeker: So that you can ask the AI system questions specifically about staff. For example, what's Jasper's phone number? Or…

183
00:38:03.900 --> 00:38:11.889
Carolien Dekeersmaeker: who are all the members of the marketing team, or who reports to Preston, right? So that is coming up in the very near future.

184
00:38:11.950 --> 00:38:36.640
Carolien Dekeersmaeker: And then something that we are starting to investigate is, and we're investigating this because we've heard the question come up from a number of customers, is we're going to look into Microsoft Copilot, and how that could potentially integrate with our ThoughtFarmer AI Assistant. We've heard a lot of organizations are starting to use Microsoft Copilot out there in terms of

185
00:38:37.040 --> 00:38:56.779
Carolien Dekeersmaeker: gathering different sources of truth across the organization, and, you know, wanting to integrate ThoughtFarmer into that as a potential, addition, and source of information. So, if you are on the call today, and you use Microsoft Copilot, or you're interested in it.

186
00:38:56.890 --> 00:39:12.059
Carolien Dekeersmaeker: please let us know, reach out to the help desk or to your customer success manager. We would love to chat with you if you do, just to kind of see, what requirements would be, what you would be hoping for with that type of, integration.

187
00:39:12.590 --> 00:39:13.550
Carolien Dekeersmaeker: Okay.

188
00:39:13.660 --> 00:39:19.030
Carolien Dekeersmaeker: In terms of getting started, yeah, Darren kind of went over that in terms of

189
00:39:19.030 --> 00:39:35.260
Carolien Dekeersmaeker: You know, if you are a Canadian customer there, or an oceanic, customer, there would be sort of the data processing agreement to kind of sign off on, but in terms of recommendations for getting started with the AI Assistant.

190
00:39:35.260 --> 00:39:47.800
Carolien Dekeersmaeker: honestly, obviously, the value it brings all goes back to, to your content, your information, and, and the governance around it, right? Like, how up-to-date is your content, right?

191
00:39:47.800 --> 00:39:55.549
Carolien Dekeersmaeker: So, a few tips. Pick a safe, high-volume use case. We've talked about a few on here, so those

192
00:39:55.550 --> 00:40:08.309
Carolien Dekeersmaeker: IT help questions, any type of onboarding policies and procedures, we all have them. If you have them documented on the internet, those are really great candidates to start testing out the AI assistant with.

193
00:40:08.310 --> 00:40:21.190
Carolien Dekeersmaeker: Make sure that you confirm your content permissions. So, again, the AI system does respect, any page permissions in ThoughtFarmer, but make sure that you verify that people have access to the right information.

194
00:40:21.190 --> 00:40:28.639
Carolien Dekeersmaeker: And, maintain clean source content. So, again, like, make sure your content is up to date.

195
00:40:28.640 --> 00:40:45.909
Carolien Dekeersmaeker: it's accurate, it's not duplicated. Again, if this is something that you, need to revisit, if you've had your intern for a few years and there's certain areas that might need a little bit of cleanup, do that before you start out with the AI assistant. Incidentally.

196
00:40:45.910 --> 00:40:56.650
Carolien Dekeersmaeker: If you want to chat with anybody in our CS team, we have a lot of tips and tricks about going about updating your content and cleaning it up, so feel free to reach out if you want to chat about that.

197
00:40:57.670 --> 00:41:06.969
Carolien Dekeersmaeker: Define your internal AI policy. This is something we really recommend. We've done this at Thought Farmer as well, just sort of, like, as an organization, think about

198
00:41:07.630 --> 00:41:15.609
Carolien Dekeersmaeker: what AI means to you and what policies are around AI that you want to roll out across the company. That will definitely be helpful as you roll out something like this.

199
00:41:15.760 --> 00:41:30.440
Carolien Dekeersmaeker: And then lastly, we recommend starting with a little pilot. So, enable the AI assistant, but maybe test that out with a smaller group first, whether that is, you know, if you have an intranet committee, a team of internet administrators, start out there, perhaps

200
00:41:30.470 --> 00:41:50.489
Carolien Dekeersmaeker: expand it to one specific team, or maybe all your content owners, and then learn from that. Like I said, look at the logs, look at the feedback, update and, improve how… what AI system knows about, and improve your content, and then scale it out to, the entire organization.

201
00:41:51.810 --> 00:41:54.679
Carolien Dekeersmaeker: Alright, that is…

202
00:41:55.140 --> 00:42:03.519
Carolien Dekeersmaeker: the end of my spiel. I think we're at the end of the, the webinar here. I do think we have a few minutes for questions and answers.

203
00:42:03.630 --> 00:42:08.080
Darren Gibbons: And I'm just opening the panel here, and there's a lot of them.

204
00:42:09.190 --> 00:42:13.800
Carolien Dekeersmaeker: Sorry, Darren, did you… is there anyone that, any question that you've seen that you wanted to, yeah.

205
00:42:14.030 --> 00:42:17.599
Darren Gibbons: Yeah, I'm happy to jump in here and help out.

206
00:42:17.600 --> 00:42:42.160
Darren Gibbons: So yeah, so just going through here, a couple of questions here. So first one up here from Donna. If the AI assistant is on every page, yet accesses 1,000 pages with internet with more pages than that, how does it respond with an answer outside of that limit? It was unable to ingest, a hallucination or a preset answer. So I think we saw that, actually, in Caroline's demo there. You'll see the response back where it says, I'm sorry, I don't have any. It should come back with, like, I'm not sure about that.

207
00:42:42.500 --> 00:42:52.980
Darren Gibbons: That answer there, and that's a good signal for you as an internet administrator, if that content exists, to go ahead and add that to a topic so that the assistant can help with that.

208
00:42:53.760 --> 00:43:02.660
Darren Gibbons: Next question here from Jade. Can the AI assistant be configured in another language? Our internet is French, and we have some laws in Quebec about use of English in the workplace.

209
00:43:02.660 --> 00:43:18.470
Darren Gibbons: Right now, in the beta, the chatbot is English only, and it is pulling content from English pages only. We have done some internal testing for multilingual, and it is something that's definitely on our roadmap, because we know we've got customers that need to have that multilingual feature there.

210
00:43:18.470 --> 00:43:19.679
Darren Gibbons: There is some…

211
00:43:19.870 --> 00:43:27.459
Darren Gibbons: it actually works quite well. The AI tools that we're using can do the translation quite good, and can actually take

212
00:43:27.810 --> 00:43:35.910
Darren Gibbons: answer questions, pull from English content there as well, too, so there'll probably need to be some additional settings there, just to kind of give you a little bit more control over what

213
00:43:35.910 --> 00:43:48.659
Darren Gibbons: what we should do if the answer exists in a… in English content, how to best respond with that information, but it is something that we are… we're working on currently, and will be… it's on our backlog.

214
00:43:48.660 --> 00:43:56.279
Carolien Dekeersmaeker: just, like, jump in there, because I was recently talking to a customer who's also located in Quebec and has a lot of French content on their site.

215
00:43:56.350 --> 00:44:08.260
Carolien Dekeersmaeker: And even though we haven't officially just sort of enabled that multilingual functionality, if you have French, for example, pages on your internet, and a user would ask a question in French through the AI assistant.

216
00:44:08.260 --> 00:44:20.300
Carolien Dekeersmaeker: it will respond in French. We weren't expecting that, but for this customer, we tested it out, and it worked quite well. So, again, if the source content that you have there is in another language.

217
00:44:20.300 --> 00:44:22.829
Carolien Dekeersmaeker: It should be able to handle,

218
00:44:22.830 --> 00:44:30.139
Carolien Dekeersmaeker: questions in that language as well, is what we found. But again, yes, to Darren's point, further improvements might be on the horizon there as well.

219
00:44:30.140 --> 00:44:31.999
Darren Gibbons: Yeah, yeah, there's gonna need to be some polishing.

220
00:44:34.500 --> 00:44:59.059
Darren Gibbons: what happens if you're close to a 1,000-page limit, and then a subpage is created by a non-admin that pushes over the limit? That page will still get indexed, it will still be made available by the chatbot. You will see a warning message in the log screen, but it's not, like, a hard limit there on the backend there for it. So, it'll continue to work, and you're able to go in there and, you know, remove some pages or archive some content. I'm not sure who was exactly asked there, but just to be clear there as well, too.

221
00:44:59.060 --> 00:45:01.789
Darren Gibbons: Is, pages that are… content that is archived.

222
00:45:01.790 --> 00:45:06.350
Darren Gibbons: If not used by the chatbot, will be, used by the,

223
00:45:06.500 --> 00:45:13.710
Darren Gibbons: It to answer questions, so if you're cleaning up any areas of the internet, using that archive feature is a great way to reduce that number of pages that are in the limit.

224
00:45:14.540 --> 00:45:33.979
Carolien Dekeersmaeker: I just saw a question, sorry, Darren, from Heather, and it's something that I actually meant to show in the demo, but I didn't, so I want to quickly tackle that one. So, Heather asked, is there a way to get more granular about which subpages are searched without having to create a new topic? And so, yes, you can actually, on an individual basis, view whether or not a page is currently

225
00:45:34.040 --> 00:45:36.700
Carolien Dekeersmaeker: indexed in the AI Assistant or not.

226
00:45:36.720 --> 00:45:53.659
Carolien Dekeersmaeker: And for that, I'll go back to my example here, for the, the Sylvia announcement, right, which I ended up adding. So if you're ever on a page and you're wondering, okay, is this information actually… does the AI system know about that? You can open up this page information toggle here.

227
00:45:53.660 --> 00:45:58.979
Carolien Dekeersmaeker: And you'll see, you'll see right here under AI Assistant Topics.

228
00:45:59.030 --> 00:46:03.370
Carolien Dekeersmaeker: We're… that… yes, so this page is actually

229
00:46:03.390 --> 00:46:22.649
Carolien Dekeersmaeker: indexed by the AI assistant under HR announcements, right? So that is a good indication to you that you can rest assured this is indeed something that people can search. A page that, I don't actually know if I can… Jessica, actually, do you know if a page is not indexed, that field would just be empty, or would that…

230
00:46:23.440 --> 00:46:27.539
Jessica Dill: it will show, and it'll say not by AI Assistant.

231
00:46:27.740 --> 00:46:35.629
Carolien Dekeersmaeker: Yeah, so that's another way, Heather, if you wanted to get really granular and, you know, review specific pages, you can double check that way.

232
00:46:36.720 --> 00:46:43.439
Darren Gibbons: you can add multiple pages… you can add additional pages to a topic as well, too, it's not just a one-to-one mapping there as well.

233
00:46:43.440 --> 00:46:44.040
Carolien Dekeersmaeker: Yes.

234
00:46:46.000 --> 00:46:47.070
Darren Gibbons: Great.

235
00:46:49.090 --> 00:47:13.860
Darren Gibbons: So I think we answered that one. Yeah, I answered that one as well, too. If content is cross-linked to a section, will it be included if that section is selected as a topic? So if you're using a link page, for example, so if you're in a policy page and you're linking to something that's in another section, that link page will get indexed, but the actual content that it's pointing to will not. So you would actually need to go in and create a new topic

236
00:47:13.930 --> 00:47:24.560
Darren Gibbons: or add that other section to that topic as well, too, so it doesn't automatically follow linked pages for the chatbot, for the AI assistant.

237
00:47:25.310 --> 00:47:26.710
Darren Gibbons: Yeah.

238
00:47:26.710 --> 00:47:44.870
Carolien Dekeersmaeker: There's one here, the question from Lorraine. I think it's similar to the question you answered previously. So, Lorraine is asking, if more pages are added into a section that was selected as a source, will it include new pages? Alternatively, if one is deleted, is it removed from the AI system? So, Darren, that's… the answer is yes.

239
00:47:44.870 --> 00:47:46.709
Darren Gibbons: To both those questions, right?

240
00:47:47.050 --> 00:47:53.770
Darren Gibbons: That's correct, yeah. If you delete a page, if you archive a page, it gets removed. If you add a page, or move a page, it'll get included, so…

241
00:47:55.070 --> 00:47:55.690
Carolien Dekeersmaeker: Right.

242
00:47:57.660 --> 00:48:16.039
Carolien Dekeersmaeker: All right, question from Anu at ACCA. You said earlier the AI Assistant will appear on every page. Is there any thought to restrict this to only certain areas? For example, does the AI Assistant only pop up on HR pages where it can pick up HR policy?

243
00:48:17.520 --> 00:48:25.990
Darren Gibbons: Yeah, we currently don't have the ability to restrict it in that way. One of the things there, Caroline, you've got the chatbot there open, is you can actually select a topic.

244
00:48:26.390 --> 00:48:29.700
Carolien Dekeersmaeker: Oops, sorry, what did I do there? Yeah. Yes.

245
00:48:30.200 --> 00:48:36.240
Darren Gibbons: Yeah, if you are going in here and you're creating a chat, you can actually limit your

246
00:48:36.480 --> 00:48:40.359
Darren Gibbons: your question to only being answered by

247
00:48:40.360 --> 00:49:04.340
Darren Gibbons: you know, information that's available there in that topic. By default, we choose all topics, but if you do want to restrict it, we can do that as well. We don't yet have the ability to, to be able to limit it to just… so if I'm on, like, an HR section, to just limit the chatbot only to that HR information, we don't have that ability to tie it together. But if you think that that's something useful, that's really useful feedback there for us, and something we can definitely consider for the roadmap.

248
00:49:06.260 --> 00:49:06.860
Carolien Dekeersmaeker: Great.

249
00:49:07.230 --> 00:49:16.059
Carolien Dekeersmaeker: Alright, great. So, Teresa just had a question. Similarly, if a content source is archived, will the AI tool find the information?

250
00:49:16.060 --> 00:49:27.959
Carolien Dekeersmaeker: and serve up the answers. So, no, it does not, look at archived or deleted content to formulate an answer, so you'd have to go and unarchive or restore that section if…

251
00:49:28.210 --> 00:49:32.179
Carolien Dekeersmaeker: If you wanted that included in its potential answers.

252
00:49:33.230 --> 00:49:37.799
Carolien Dekeersmaeker: Alright, we're almost there.

253
00:49:37.800 --> 00:49:55.639
Darren Gibbons: I found a few more questions. I saw some in the chat here as well, too, so I'll run through these as well. So, Copilot, there's a question here about Copilot, and the answer is yes, we are investigating Copilot Search there, integration there as well. That would probably be an add-on for the feature, because it would be a separate

254
00:49:55.950 --> 00:50:20.159
Darren Gibbons: from our actual chatbot, but it's something that we're investigating, so if that's important to you, please let us know. We're reaching out to customers that are using Microsoft Copilot and want to get a better understanding of your use case and your expectations around that. Will the AI assistant have access to any Microsoft 365 content? So this is kind of the opposite. Copilot is, we've got Copilot AI, will it be able to read and answer questions based on content in ThoughtFarmer?

255
00:50:20.350 --> 00:50:34.410
Darren Gibbons: And then the access to 365 content is kind of the inverse of that. Will the AI assistant be able to have access to Microsoft 365 content? That is something we're also investigating there as well, too. The idea being there is we'd be able to add

256
00:50:34.510 --> 00:50:42.569
Darren Gibbons: Microsoft 365 content to a topic so that information that's there can be answered by the chatbot there as well.

257
00:50:42.780 --> 00:51:00.570
Darren Gibbons: question here about the beta period. Any insights? We're hoping to… we'd like to get it to a point where we can drop the beta title from there and provide that finalized pricing for people. Hoping, in the next 6 months, we don't yet have a fixed date there for that, though, however.

258
00:51:01.720 --> 00:51:13.020
Darren Gibbons: Number of people using Copilot. Question here about HR as a topic. Select a source of the HR page, check include subpages. Came with 491 pages, tells me the page limit has been reached.

259
00:51:13.020 --> 00:51:29.940
Darren Gibbons: So I guess, we… it sounds like we might need to follow up there with just a support ticket on there. Should be, 1,000 pages is the limit there. It's possible… in the early beta, we had a limit of 500 pages, so it's possible there that, we just need to do a quick update there to increase the limit there for,

260
00:51:29.940 --> 00:51:30.930
Darren Gibbons: For your internet.

261
00:51:31.250 --> 00:51:37.869
Carolien Dekeersmaeker: Yeah, John, so if you want to follow up with just a quick email, or even, like, a help desk ticket, we can… we can unlock that for you,

262
00:51:38.000 --> 00:51:41.380
Carolien Dekeersmaeker: within a few minutes. That was probably just a setting to update.

263
00:51:43.480 --> 00:51:48.380
Carolien Dekeersmaeker: Okay, I feel like we have probably…

264
00:51:48.380 --> 00:52:11.640
Carolien Dekeersmaeker: covered most questions that were a little bit over. If you have any other questions that you didn't see answered today, or that you think of later, please feel free to get in touch. We are more than happy to chat about… we love chatting about the AI assistance, so let us know. Feel free to reach out to Darren and me directly, your Customer Success Manager, Colton or Corin, if you're currently talking with them.

265
00:52:11.720 --> 00:52:32.950
Carolien Dekeersmaeker: And we'll be happy to chat more. Again, stay tuned for further announcements on improvements to the AI Assistant. I know the people indexing is very close. And yeah, we hope you go out there, test it out, discover all of its potential, and any feedback at all.

266
00:52:32.990 --> 00:52:38.419
Carolien Dekeersmaeker: Is super helpful for us, so don't be shy. Reach out if you have anything you want to share.

267
00:52:38.630 --> 00:52:46.230
Carolien Dekeersmaeker: Other than that, I just want to reiterate that we are, we have recorded the session, we'll follow up in a few days and send that out.

268
00:52:46.380 --> 00:53:00.310
Carolien Dekeersmaeker: And thank you so much for joining everyone today. We really appreciate it. As you can tell, we're really excited about AI Assistant, so we're happy that you guys showed up in droves today and share our enthusiasm.

269
00:53:00.500 --> 00:53:08.470
Carolien Dekeersmaeker: That's it for us in Vancouver. Thanks so much, guys. Have a great rest of your day, and I'm sure that we will chat soon.

270
00:53:09.000 --> 00:53:09.770
Darren Gibbons: Thanks, everyone.

271
00:53:10.030 --> 00:53:10.740
Carolien Dekeersmaeker: Bye!