KCS & the AI Revolution: Building Powerful Knowledge

welcome to the service management show from the team here at Ficode i'm Gary Blower uh IT service management practice lead and this week we're going to be talking about knowledge centered support or knowledge centered service and I'll let my guest introduce himself hi everyone I'm Jack Graves i'm the uh founder and CEO of Capable um and we're an Atlassian marketplace partner um and I'm here to talk about knowledge centered uh services yeah indeed so um so Jack and I first sort of started chatting a couple of weeks ago actually at um connect in London uh which see previous episode of the service management show uh to do with that um and we were talking so Jack will explain a bit later on but Jack's area of expertise really is around um Atlattian confluence and using it for knowledge and collaboration so we were talking around this topic actually and I thought it's an interesting one to feature on the next service management podcast um if you're not familiar with knowledge centered support or knowledge centered service as it's sometimes referred to uh it's a framework or methodology that's been around for I think 15 20 years now it's it's not it's not a new thing but it's had more of a resurgence recently because of the importance of knowledge to things like large language models and AI so there's much more of a focus on knowledge centered support so in layman's terms uh just before we sort of get into like lessons learned and stuff knowledge centered support or knowledge centered service is about if you're if you're a support team and you're getting questions and you're and you're providing responses is don't just provide a response provide knowledge so you write the knowledge article and then you share the knowledge article so if you get something novel you write a knowledge article first and then you share it with the customer and then next time the customer asks that question you've already got that knowledge likewise if you you you know agents and service agents things like that can do the same thing for more complex things it's write it up first and then you've got it for next time um there is a whole framework you can look up which we're not going down to detail now um but yeah so to to to Jack obviously you know the stuff that you've been doing uh over the last few years what would you say are the kind of the main benefits of using you know a knowledge base to provide that kind of um self-service yeah I think in the past it's been uh it speeds up resolution time as your other support staff can read what the other people have already solved um so I think we've all seen it where we've solved an issue in the past and then it comes up again and then we're back to square one and we have to re understand what the problem is so I think previously it was a bit of an afterthought to kind of document certain bugs and things that have happened in your support desks um but now as you say it's become more and more important as we're moving towards more AI agent models uh to solve it so I think we all know like um it's kind of what you put into AI will dictate what you get out of it right so if you put a load of rubbish into it it's not going to be very effective um so really building that knowledge base has become more and more important for companies now uh as we look to kind of effic more efficiently handle our support yeah and I mean I I've you know I've worked with building knowledge bases for a long time and I can remember um like I had way back I had I worked for a bank and I had a team that I took over there that looked after like all the dev and testing tools for the for the entire bank one of the first things I did when I I joined that team is I said to them that okay when you get something you know an inquiry or there's a problem or something like that write the knowledge article first and then send the knowledge article to the customer and initially I got loads of push back on that it's like "Oh but it takes twice as long." And this and until I proved to them it takes the same amount of time because often they were just writing an email like well don't write the email write the knowledge article and then just copy and paste that into the email it's job done um and uh you know this is like I said something's not something new but as you said because of AI and the whole garbage in garbage out it's something that there's now actually a lot more focus on what you know in your experience of you know particularly working with uh confluence where have you seen people really struggle to um create the right sort of content and and have content that's that's that can be shared and used in that way yeah i think there's you got to recognize there's a difference between internal documentation and documentation you want to share with customers or AI agents um so obviously we've all got intellectual property and and things that we don't necessarily want to share like the technical details of how things work so I think that's been an area I found with developers especially um or support staff um that are more technical is that they might write the article with the audience being internal um and then that can't be shared right you need to rewrite it uh for a more customer centric approach so I think understanding the difference between internal and external is a challenge I've seen people face in the past agreed yeah i've seen plenty where everything's mixed together and so it's oh we can't share this with customers because it's okay um the other thing um and I know this is something that that you guys have done some work with at capable is one of the other challenges is if you're good at doing this and you're good at producing this content you're providing this kind of self-service knowledge or internal you know playbooks and runbooks and things like that is keeping that stuff up to date and relevant and and if anything weeding out the stuff that isn't so again what sort of challenges have you seen with with customers that you've worked with around sort of just maintenance of of knowledge really yeah I mean that's that's a big challenge i it reminds me of I did a job when I was a consultant setting up something similar for a legal company so in finance and their documents had to be reviewed say every quarter or every uh half of the year because the law might change right and then they need to update it so uh the lack of kind of there's not many document management solutions in confluence for finding those pages that are out of date um so I think you do need right now you need to have a a kind of plugin that that provides the uh I want to call it archiving or um approval which expires when the page is considered maybe it needs re-reviewing or updating um but I think that's a huge problem uh I think we've all been on kind of conflence sites where most of the data is a year or two you know the pages are not actually relevant anymore they're not correct um and as we said earlier like with AI it's not going to make a distinction between that 2-year-old article and something that was last week um so yeah I think keeping your content up to date um you can't think of confluence like I create a page I forget about it um you do need to keep kind of looking in there and and updating the content agrees yeah and um obviously people can tell from my background I come from I'm a big ball gamer and uh one of the things you get in the ball gaming world is this concept of living documents this idea that you have like a living rule set that constantly evolves over time and you get you get you know versions of those rules that are updated on a regular basis you know sometimes you know some some games have got rules that go back 25 years you know so um and they're constantly being updated and I think the same is true for for for knowledge isn't it yeah I think so um you do need to keep it up to date yeah definitely uh but luckily you don't have to pay uh for say Dungeons and Dragons edition number four or whatever i forgot what it is now i think it's like seventh or right thing i can't remember but yeah yeah that's also a paid model yeah you get that as well um right okay um so sort of coming back to the kind of topic of knowledge support and we've talked about AI quite a bit just there um I actually saw a talk on this recently at a service management event and um they were talking about the importance of uh you know centered service coming back um for all the reasons you just mentioned and actually one of the things they raised was AI hallucinations which I think is what you were talking about isn't it where you it's reading old stuff that's out of date um in the Atlassian world obviously we've got um Atlassian intelligence and and rovo I understand if you you've done a little bit of rese research around this haven't you with Robo and looking into how you can use that with knowledge what's what what have you played around with yeah I found so Alassian has two separate well more than two we'll probably get into virtual agents later but um they've currently got Alassian Intelligence which is the AI feature and they've also got Robo which is the more kind of complete feature so the big difference between those two is Alassian Intelligence doesn't know anything about your company it doesn't know anything about your confluence knowledge base it's basically like using chat GPT right you ask it a question in fact it is chat GPT um whereas Robo takes it a step further and actually it can pull on all of your knowledge base it can also look in I think now Google Drive and and other places for data um so for us like I can ask it to write me a blog post about say capable um and it will pull on all this data across our system and and write it so I think especially for what we're talking about here it's it's really good to kick off that confluence page by saying like slash AI uh write how you solve this problem and it will look through your knowledge base figure out what what it's got and then actually give you a good starting point for for that so I think that's a really key thing to understand about the differences between Alassian intelligence and ro um just to note that uh obviously ro is currently a separate charge on top of your Jira license uh whereas I think Alassian intelligence is with premium plans this premium above yeah and and as you mentioned the the key thing um you know in the Atlassian uh ecosystem is if you're using um J service management then Atlassian intelligence is the engine behind the virtual agent and that's where this Norris and support really comes in because um you can actually through like the customer portal or through slack or teams or whatever you're using you can actually ask for something called exlassian answers and exclusively exassian answers basically uses the confluence knowledge base to provide the answer doesn't it so you So that's where you're looking to kind of optimize your your knowledge to to enable you to do that but you know just to reinforce just that's just one platform and I've been speaking to other vendors as well are speaking to the guys at Fresh Service um this week or last week rather uh and also guys from another product called AIO I think I pronounced it correctly like um Polish ITSM vendor and they're all they're working through similar things you know where they're bringing in these large language models to particularly around knowledge and providing knowledge in lots of different ways and they're coming up with the same issues in that you know unless the knowledge is maintained their their AI doesn't do a great job so yeah uh that that reminds me I think so recently intercom actually lets you integrate with conflence now so you can actually use uh all your data from Confluence in intercom uh on your website as well really cool uh right okay so we come let's cover some um let's talk about what capable does so um the reason we got chatting really was that you guys do things or you enhance confidence in a way that kind of lends itself to this kind of knowledge centered uh support approach doesn't it so what what do what does uh what's capable been up to yeah so capable is an all-in-one solution for confluence uh the idea is in the past you'd need say 10 15 apps to do very simple things um we think that's kind of changing the landscape now um is moving towards companies want to streamline they want to reduce the amount platforms they have instead of you know having 50 different products that to do the job so what capable does is it gives you everything you need to manage your knowledge uh in confluence so there's a few key things missing from confluence right now um like we touched on earlier the first one being management of like large knowledge bases um managing thousands of pages in conflence is not very easy um it doesn't have the concept of Jira where it's got like saved searches and like JQL and you can share those with people um so like we said finding outdated pages that need updating um there's currently not really a nice way of doing that at scale um so the first thing cable does is provides a a really uh advanced content management system which lets you search pages and save them as filters and then share that with the rest of your technical writer team for example um to say Gary here's the pages you need to update today for example um I think the second big thing that's missing is an approval and publishing system so at the moment uh conflence can publish publicly um using conflence you could also use scroll viewport or refined wiki to kind of publish a website um and the the other one is Jira Service Desk right someone's raising a ticket they're typing something in it finds articles that are related but there's currently no way to kind of approve those pages to go in there they're either in there or they're not um so I think if you're writing this stuff for public consumption you really need to have an approval system um so what capable approvals feature does is allows you to write your page put it through an approval process automatically uh where it will ping say your manager to say you need to approve this page um that could be on email or Slack um so we get really quick feedback loop going um and the moment that page is approved it'll be published and publicly available for your customers so it provides that kind of missing link between writing the content and the review step that's missing and actually having that being used by your AI being used by service desk or scroll viewport for example um so that's just a subset of the features we have but I think those are the most relevant for this uh discussion yeah definitely definitely yeah and and I think that you know those two in particular where you highlighted earlier is the fact that you need to keep everything up to date um is key and also that you don't want to publish let's use that word stuff that's like you said not ready or is draft or being reviewed and all that kind of stuff i think those are those are essential ingredients really to any kind of successful knowledge center support or or knowledge base that is going to be exposed to uh kind of customers in some way now the other thing I did want to talk about and it probably is only partially relevant but I'm a bit of a nerd when it comes to this stuff and I I absolutely love markdown now I appreciate it's kind of niche but it has relevance because particularly if you're if you're working with say know center support where you're where you've got a product that's been developed within the business developers are generally very good with writing documentation in markdown things like GitHub and GitLab and that support it natively anyway um and I got super excited when we were chatting when you said that it had Markdown support and it does all these clever things like you can use markdown to create diagrams and things do you want to just elaborate on that a bit just for the the nerds that are out there like me yeah um so I think conflence out the box you can generally kind of write Markdown but uh it doesn't have full support of it and you can't do things like diagrams and and things with it um and other solutions for markdown don't have any kind of templating solution so it's kind of like each time you add markdown it's a separate instance it's kind of starting from scratch um so what capable does is it lets you create a template library for markdown um where you can create say how-to articles so that your your team your developers can just start writing markdown um and as you said you can just put in a a diagram block and put in say flowcharts and and other things um to help understand what's going on um that in addition to say including markdown from a URL so uh from a GitHub repo say uh for example um and I think our like secret source is the fact that we've got this live markdown editor so as you're typing in uh it's giving you a live preview of images diagrams tables everything there um and it just makes it easy say for your development team or your support team to write those articles because I think we've all seen that chat GPT has got this canvas feature now which makes it really easy to write markdown so um you can use that to kind of give you again a good starting point for your uh doc pages um yeah it's it's very seamless very streamlined and uh gives you that kind of template library to share as well yeah I I think it's awesome i said I'm I'm a big fan of it i remember um I think I told you this that we many years ago now I did a a really massive knowledge base for a for a scientific institute in the UK quite a famous one and they use um they actually used Confluence to for all the documentation for things like um electron microscopes and and various other things you know so they're basically their their manuals were effectively all digitized and obviously being scientists and and and kind of nerdy everything was done markdown so um it's kind of uh Oh yeah i just like anything that supports it basically um so yeah it's kind of a a thing that that tickles me um right okay so let's kind of wrap things up so kind of closing thoughts do you have any of general tips for folk in terms of using any kind of knowledge base but obviously specifically confidence because that's what you guys work with um you know if you're looking to provide a kind of knowledge base what are the top tips uh I think the first one would be consistency so you know if you're going to do how-to articles or guides or references you should really have templates for those so that it's all consistent across the knowledge base um so if you've got six seven eight like support agents you don't want each of them to be writing a different style of of of article really um so that's the first one I'd say is consistency is is king right um I also think organizing those spaces so in confluence as you know spaces are the logical kind of container for all your pages so if you're going to have draft content or internal content and external content don't put them in the same space um this is especially important with Jira Service Desk where you kind of give it a space to index um I mean that said you can use labels to do it but I kind of I suggest not doing that because now you're just not sure which space is internal and which one's external um so I think the second point is logical separation of internal and external pieces um I think uh the third one would probably be you really need to have that review process for this uh content as well um especially if you don't want uh someone to unilaterally be able to publish something publicly that could cause say some reputational damage if there's something inappropriate in there um or just the wrong tone um so I think that's the that's the next one sort of a having a a rigid approval process that people can't bypass um as well very good anymore is that That's Yeah that