A look at HaloPSA's ChatGPT integration with Morgan Aspinall — how to set it up and use it from both a ticketing and an end-user perspective.
hey everyone and welcome back to our halo implementation Series today we're going to actually talk about something that isn't strictly I guess part of the implementation um but it's a really cool new feature in in Halo and this is the chat GPT um integration so this brand this is a brand new feature and by the time this video gets up on onto the air um you will more than likely have it enabled in your database um so with me again today is my favorite person at Halo Mr Halo himself who is actually going to talk us through how to set up the check GPT um you know how we can use it in Halo and um and not just from a ticketing perspective but also from an end user perspective as well so um hey Morgan how you doing and thanks for being with me again today yeah hey Chris thanks for having me um yeah this is an exciting one uh this it came pretty much out of the blue I mean I know there was some internal conversations that we were having about about sort of wanting to integrate with chat gbt and then a week later it was done and now it's like it's on a version version 2.116 to be specific um so yeah keep an eye out for that version in particular um but yeah I thought we we'd sort of give everyone a walk through of how they will be setting up and potentially how they'll be using it um once the once the integration is available so and I just wanted to say you've seen it here first right so that's it that's it I mean this being candid um the first time that I've actually interacted with this was about 10 minutes ago so we're all we're all learning here um but yeah let's take a look good yeah let's do it excuse me cool so um if we head to configuration Integrations we'll now see a new integration at the bottom uh chat EBT so when you click in here the first thing that you'll be prompted with is to provide an API key I just want to quickly highlight where you go to to obtain that API key so you need to ensure that you have an openai account um and then from your account if you click your account in the top right you'll have an API Keys option um and in there you can generate a new key and that will be the key that you you plug into to Halo now once you enable the integration module um what you'll find is a couple of options against your actions so there's two there's two main functions of the integration that you can use at the minute um you can either generate a response or you could you can have chat GPT generate a response or you can have chat gbt sort of improve enhance provide some context to a response that you or your technicians are actually putting into Halo um so if we go to action config what we'll find is against our actions now at the bottom of the details tab there is a chat gbt section now in here we see we have our operations um where you've got generate use response user response or improve agent note so the generate user response will actually send this prompt uh to chat gbt along with the whole thread of the ticket and we'll have chat gbt generate a response and send that back to the customer directly so um she had no more sort of having to actually write a response up yourself especially if it's some sort of generic response um you know we're just waiting for feedback something like that we just need more information what we can do is we can just have chat gbt perform that action for you effectively yeah and I just want to say I mean if you look in the system settings you know how how Morgan's written that um you know we need to at this point tell check GPT um you know how it needs to interact with the with the user so you know you're a helpful service desk technician you can obviously put whatever you want in there um but but you know maybe in this particular instance we wanted check GPT to act as though it was a technician responding to a customer in a in a kind of a friendly manner so that's it that's it yeah um I suppose it's also worth noting here that we've provided an option to input the different models as well so you know as the the new iterations of uh the open AI get released gbt4 and so on um you'll be able to put those updated models in here and get back the response as per the updated model so yeah it's quite it's quite a handy one as well um so we've got this track gbt response action uh if we go to a I'll tell you what let's log a ticket from the the self-service portal so I'll create a new ticket laptop so this is the customer login the ticket yeah this is from The Sound Services yeah yeah of course this would work in the same way if if the customer emailed it in or you know whatever method they want to use we just so happen to be using this but that is yeah that's correct yeah yeah just for the sake of um sort of uh time keeping because I don't have a proper mailbox hooked up to my internal instance um we'll yeah do our interactions via the self-service portal so um I've got this chat gbt response action at the top here click that I haven't typed anything in um as the agent and we'll get our initial response from chat gbt uh sort of saying Hi how can I help blah blah blah just wait to get that back and here we can see a fully fledged message um based on the the context of the the sort of previous Fred and without me as an agent without me typing into the keyboard at all so that's awesome pretty neat yeah yeah it's quite cool yeah yeah yeah I mean and I think this is going to massively um kind of um you know enhance the way the people are interacting with their customers because yeah yeah um you know I don't know if I would necessarily respond like that to be fair you know I'm sorry that you to hear you're having trouble with your laptop um but you know I guess you know I guess now that you've got that response right you can go ahead and just edit it if you wanted to further edit it a little bit more so yeah I mean I I guess just as like an initial demonstration of how the integration Works we've kind of put it together um in in a you know sort of fast and loose way so to speak but but just just like with uh chat gbt itself people will um you know be putting different messages into those actions right so if we go back to the configuration of our action I I certainly expect people will be playing about with the the system message prompting here um and I imagine sort of on the you know on the various forums uh the NSP Community will sort of share their ideas about what they've found is a successful system message versus what isn't and over time people will find that the optimal uh text to put in here such that chat gbt can provide a sort of um a convincing and um human response let's say yeah and I guess something else now that you know gbt4 is out which you can provide much more context for um we'll probably start looking at sort of feeding it information about Halo itself so it's not just giving you a generic response about a you know um if your laptop won't turn on can you plug it in but actually sort of giving you knowledge based articles and stuff like that but that's that's further on down the road for now we'll just leave this leave this yeah yeah no I I mean there's a huge amount of um potential with where this can go you know because obviously yeah it's it's responding at this point to um you know to the customer in in the way that we asked it to in a very friendly friendly way um but yeah I can see this going you know write me a knowledge based article on on you know how to solve this issue or you know write some customer documentation on you know the customers asked about you know how they set up their phone system or how they add email to their phone you know use chat GPT to to send them that information and you know write documentation so yeah I can I can definitely see this is gonna This is Gonna Go Far and yeah already this is something that that very few other PSA tools are are doing already so it's really really good to see that yeah yeah yeah we were we were quick to the quick to the party with this one I must admit um yeah yeah which leads me on to the the next feature available um when using the chat GPT integration um so we'll go back to our action configuration and we can see that uh not only can the chat gbt integration sort of generate responses autonomously um but but it can also take a an input from an agent as an initial sort of context for the message and then um tweak it fixed grammatical or spelling mistakes or anything like that um and you would do that via the improve agent note option instead of generate user response um now when you have your improve agent note what you can do is you can provide the chat gbt with a prompt and in that prompt we can use the dollar Rich action note variable which will be replaced with the text that the agent puts in at the point of Performing the action and then we can provide it some wider context here so just as an example I've said you know keypad keep our HTML tags preserve any sort of HTML formatting and but make sure that they're spelling and grammar is corrected and make it polite as well so if I go back to my ticket uh and I click enhance no and I say something like I have no idea what is going on and you are really annoying me so I've put that in as well sometimes we always we do want to say that to our customers sometimes don't we yeah oh well this this could um yeah this could be this could be dangerous this now now we've got chat gbt sort of interpreting agent responses people are going to be putting in whatever they want um yeah and you can see here that uh my rather rude message has been replaced with something uh far more palatable yeah so yeah another really cool feature that's really good and I think sorry if I can understand I think what's really nice about that as you say is it's it's it's going to correct grammatical errors and it's also going to correct um uh you know yeah when when somebody must say something slightly less polite or something less um something that you wouldn't be wanting to say to your customer that's right obviously um and you know calling a customer a or something you may not want to necessarily say that although you might think that um you know and check GPT is going to be able to help you to fix that stuff yeah I see two uses for this at the minute um one is sort of providing certain tonality as we just saw there um so you know improving the uh how polite or how approachable a message may seem um the the other that I see could be useful is uh making a response a bit more verbose so we might want to change our keep doing that we might want to change our prompt in the action to say something along the lines of you know take the details from this note and format it in you know long form paragraph manner that way the agent could just put in jot down some notes on the ticket did this did that fix this fix that and then track gbt could load that into a wider message that you know hi Mr customer uh just wanted to give you an update blah blah blah and all of that stuff that um you know is not really useful um but it all aids to the customer experience that can all be filled out by by the AI now so yeah those are the two uses I see in a minute absolutely and I think one of the things I can see here you know you've mentioned about kind of spelling and grammar um errors you know to fix those now you know obviously as Halo is a is a product that um is sold worldwide you know we all we all spell things differently you you can actually tell chat gbt at this point to you know do this in British English or American English or you know so what it'll then do is take those spelling mistakes um you know and spell color with you and and we'll color without you or whatever the case might be um you know so you can really kind of enhance it um to to use the um uh you know the way in which you would speak in that particular country rather than it just kind of um you know putting something in a in a particular way so yeah no I think it's really good there's this really really cool stuff you can do with this yeah yeah I'm looking forward to sort of hearing about some of the use cases that people actually have um we're setting this up so so yeah um and finally the the the sort of the final feature um that we've built in natively which is another really cool one is you can now have chat gbt respond via the Halo Halo chat functionality um so I'm aware that we haven't done a video on Halo chat yet I'm not going to go into all of the details of it in today's session um we'll probably schedule that for the coming weeks um but for the sake of the the integration we'll briefly cover how that works so excuse me in configuration chat we can set up chat profiles now chat profiles um kind of work in a similar fashion to our workflows and that we have a flowchart which is our bot flow which can lead customers down a specific path based on user input based on multiple choice selection and so on and again for this demonstration I've just put something fairly basic together it's got an initial message which is composed of in our initial step we have an action the action is send a message which is just hello how can I help um following that initial message we're going to provide the the customer the option to input some text which will then lead us to the final step where we're going to have chat gbt respond in the chat based on the text that was entered um and again we can put our system message in providing some context as to how we would expect to chat gbt to respond to this message um now once we've once we've set our bot flow up again in the details tab just want to make sure you give it a name specify a certain level of access whether it's only on the self-service portal or potentially on external websites as well so you could have this on your own website um and then if we take a step back in our chat configuration we can see that we have our self-service portal chat profile which equates to this little chat bubble being present down bottom right here so I click that let's start a new one actually Morgan when we when we're doing the test before the video you you put something really funny on there and I'd love you love to put the same thing on again and see how it um Howard responds yeah sure sure yeah okay so um we've we've got the first step we've moved to the Second Step but we're just saying please describe it issue slash request I think what I said last time was my laptop is on fire how can I solve this that chart gbt do its thing I really hope it gives the same answer that again when we were doing the test yeah [Laughter] it's so cool it's so cool the way it answers it like oh no that sounds like a very serious problem you should yeah yeah and right like hell yeah yeah yeah exactly um so yeah you can load up track GBC into into the Halo chat itself and another really neat feature yeah that's really cool because it just gives you kind of the um I mean actually and the interesting thing is yeah I was just gonna say can you take this further so if you then said um you know could you you know could you give me some more information on what else I need to do or I've tried that what else can I do I I guess chat gbt will just carry on answering you for you know as long as you're you're going yeah unless you have another step to tell it to kind of eventually go to an engineer or something yeah I mean I guess I haven't actually tested this um but perhaps perhaps you could put it sort of in a feedback loop where you've got a text prompt chat gbt responds to it then we can provide some more text again um this is the first time I've seen this as well so uh let's see let's see what happens and and as Morgan said I mean you know on on one of the future videos we will um uh you know we will spend some more time on how to kind of set up this this live chat um because I think this is something that's pretty awesome so I wonder if it's gonna be able to answer a question like that something a bit more specific be very interesting to see I'm I'm sure knowing what I know about how check GPT works I'm sure it'll be able to um you know that is so cool okay so it's gone back to the prompt yeah so yeah yeah what else can I try yeah that is really cool that's pretty cool I mean and and that's really nice that that whole thing was automated you know rather than um you know an engineer having to type that um indeed sorry just jump to the bottom yeah um yeah and it's and it has it has understood the context from the previous message so I mean again it's a rather crude example here because I've just put it into a feedback loop but what you would probably want is you'd probably want an initial message a chat gbt response that would then lead you into a separate input box and then have a feedback loop in there um but yeah sure sure the concept of it is the same because I guess at some point you're gonna say okay I need to you know would you like me to log a ticket or something and then I can go ahead and Vlog with us a ticket or whatever so um no I mean that's really cool I think there's some there's some really cool stuff yeah and I'd love to um I'd love to get people's feedback on this so um if you you know if you want to let us know what you think about this um in in the comments below um yeah we'd love to hear your thoughts on it and love to hear some of the ways you think you might be using chat GPT in um in inhaler moving forward absolutely yeah yeah I'm looking forward to hearing some of the feedback um from how msps are going to use chat gbt within their business and within Halo so yeah excellent well that was really good that was a a very quick kind of introduction to what this is going to be um you know I'm looking forward to getting this in in my live environment um you know can't wait to see this come out I think it's going to be really really good so you know once again Morgan really really appreciate you um you're jumping on the call really appreciate your time um and I appreciate you know hey Halo letting me use you and borrow you to to do these videos so um you know really um very grateful and really do appreciate it it's always a pleasure Chris really is yeah yeah thank you very much and I'll see you on the next one yep see on the next one cheers take care
Get in touch and we'll talk through how it applies to your MSP.
Let's Talk