Airbnb Demo

This is a brief demo of an app I built for Airbnb.

And this is the interface.

It's called Airbnb Strategist.

And let me jump into the tickets.

So this is, this app connects to a couple of platforms.

One is Jira for ticket management and the second one is Confluence for the knowledge articles.

So this is a list of tickets.

in Jira, there's a section here called Knowledge Gap Detected.

I'll talk about that in a moment.

But here is your typical list of tickets.

A couple of new columns here that I've added is that one's called Type, which is going to separate M2 types of tickets.

AI is looking at the context of the content of the tickets and looking for whether it's either relational related meaning an upset, customer, maybe somebody, there's a hidden camera somewhere and the response needs to be more personal with some more empathy versus more technical tickets.

For example my, I can't connect to WI Fi, I can't log into my, my web portal or I can't upload a photograph.

The things that are, that require a solution article that are more is more procedural and repeatable by nature.

So at least that's an indication of what type of ticket is the second.

Second column, new column here is called resolution Likelihood.

And what that is is that during when it's loading the tickets it's pre searching the, the knowledge base for is there a solution already in place that can solve this?

But the reason is because when there is a, a ticket with a high probability that there's a solution in the knowledge base that's already available, it's a quick win, right?

An ambassador can open the ticket and quickly resolve it because we already have the solution and the ability to open a ticket and close it quickly that impacts our TTR our time to resolution, metric.

So this is one cool way of looking as an ambassador to look at the queue and, and be able to identify very easy quick wins.

So let's just open up a ticket here that is typical.

let's see, connect to property WI Fi.

So let me just open one of these.

So here's the ticket to name.

Here's the description.

So of course what it's trying to do is it wants me to search for an article.

So here's searching for an article.

There's no hits.

So the next it's prompting me to do next prompt is to generate a troubleshoot shooting playbook.

This is the first introduction of AI.

So real AI.

So let me click this button.

What it's doing AI and in this case I'm using Grok.

It's going into look at the ticket details and it's going out on the Internet and it's trying to come up with a, a an approach, a plan for me as an ambassador to resolve the issue.

So this is not the resolution, it's not the salt to how to solve it.

It's, it's how I can solve it.

So it's and it's in a decision tree format.

So for example if this condition happens, do this, if this and that for example is the password verified network reset, router power cycle.

So it gives me the instructions as an ambassador itself to resolve it.

So let's just say I did all that and then me as, as the ambassador, I solved it.

Great.

So I'll say you know what the, the issue was?

The issue was the router needed to be reset once customer reset router everything the thing worked right?

all right.

And then I click resolve and close ticket.

So what this is doing of course is going to jira.

It's closing the ticket and it's including my resolution notes.

Now what is also being presented is called create the solution article.

So in true KCS form or processes we should as a support agency or as a help organization, we should try to capture every time we learn something.

So in this case we did not have a solution already existing.

So we are being flagged or being suggested that we create the solution so that the next time we have a ticket that says I can't connect my wi fi, there's a solution already in the knowledge base.

So let me go ahead and click that.

Now this is the second time we're introducing AI.

So it's taking a moment to load because what it's doing AI is going into the ticket details, it's reading the ticket, what was wrong and it's looking at my resolution notes.

Hopefully I'd have a lot of resolution notes but it's looking at my resolution notes for how I solved it.

So you can see here it's can't connect To WI fi.

Here are the symptoms, here's the environment that pulled from the ticket issue, ticket details, and here's the cause.

Sometimes the router can get stuck or experience a temporary glitch.

Resetting the router clears these issues.

So obviously this is a rich text editor.

I, as the ambassador, can make editorial decisions.

I can make changes to it.

but what's also noteworthy is here I have what's called a voice, score.

This is again another algorithm that's going on behind the scenes.

It's reading the article that the draft and it's measuring it against some criteria.

These, what I have these five dimensions for my content standard.

So I've defined a content standard and that it must meet these dimensions and it must be very good.

AI generated that.

Of course it's going to be a good score.

But if I, generated this from scratch, I probably wouldn't have that high of a score.

So it would recommend fixes.

So here's a button called Optimize.

I click this and then the panel pops out and it's going to make, it's going to make some suggestions for how I can improve my scoring.

So it'll, whether it's related to accuracy or structure or clarity, it's going to make those.

And I can click the apply button.

I won't do that, but it'll, it'll update the, the content.

So let's just say I'm happy with this.

There's a couple of toggles here.

One is called Flag for training.

So I can flag this article for, for training.

Let's say this is significant.

I, as a, as a strategist want to train all my ambassadors on this so that they know about it.

Or I could just toggle it for required reading, which that means it'll push this to everybody's dashboard.

So when they log in, hey, read this.

new solution.

and now I can just publish.

It's doing its magic.

It's actually going to Confluence and it's publishing this article in Confluence.

Meanwhile, it's also linking this solution to the ticket that I just closed.

So now that, that, that's closing the loop there.

So that's pretty cool.

So let me go back to tickets and I'm going to introduce you to what's the knowledge gap?

which I think is a really interesting way of looking at these tickets.

So this knowledge gap detected.

AI looked through the ticket queue and it's looking for a, pattern.

It's detecting and recognizing patterns.

So here's A few in this case there's eight opportunities.

So here's 11 hosts or 11 tickets are related to that.

The property is not clean.

So what, what this opportunity is, it's saying look, we have 11 people who have talked about that their properties are not clean.

We should have a standard response generated for this.

So here is a trigger for that.

So let me click that again.

Here's another introduction for AI.

It's doing the same thing.

It's actually reading all those 11 tickets.

It's digesting that and it's And let me expand this show issue details which is doing the same thing.

It's going through all those 11 tick.

It's, it's understanding what's going on so that I as an ambassador can really understand what these 11 tickets are about.

So here's the issue.

There's the root causes, here's the proposed solution.

So this is just for me as the as the ambassador I now know what this common issue is.

I'll collapse that and, and then it's also drafting again AI is drafting a standard response draft.

So and the standard responses need to be.

Are very different than solution articles of course because they are personal, they are they show empathy.

It's more of a conversation.

So you can see that over here the voice score, the dimensions are different.

So we have tone, we have clarity, actionability, editorial consistency and of course inclusivity and accessibility.

So you can read this.

I'm so sorry to hear the property wasn't clean when you arrived.

So it's a very different response that the.

This is generating.

So I.

And then so this is a temp.

This is a template that to go into our, our knowledge base.

So and again I can flag it for training and let's just say I am happy with that.

I want to push, push it to my knowledge base.

This also going to Confluence is putting this into our, into our repository.

So let's just look what that is.

So here's our standard.

Let me just go to standard responses here on the left property.

Not clean.

That's the article we just wrote.

So if I click on that there it is.

I can always edit it.

I can improve it and let me go to the solution articles.

Let's go to there can't connect to wi Fi.

That's the article we created earlier as well.

you may also see that periodically there's these Solve improve or solve reuse or solve capture.

These are just teachable moments learning moments where to indicate where we are following KCS practices.

So this is an interesting way for not only is it a tool that solves, but it's also a tool that's teaching.

here on voice score standards is a lot more information about how the scoring is, is calculated.

The that is following KCS is following some brand and voice guidance.

why score matters.

So you can read into this.

Of course there's a calculation, the calculation is there as well.

You can read about this app.

So I use a product called Replit.

Replit is the app that I used to create the apps.

I cost me $147 and about two full days to develop this.

But I also would have probably would have cost just this iteration of this app, about $24,000 if we went through the traditional developer route.

So that's a pretty significant impact in savings and not to mention how fast I was able to pull this off.

A little bit about me, a little bit about the technology.

The tech stack here of course is connecting to Jira Confluence and it's using Grok for its AI.

So there's a lot more to this, to this app I would want to love to get into.

You get into it about the things about the dashboard, some of the KPIs that we're tracking I could switch to for example the Strategist.

You'll see that the dashboard refreshes here and it changes.

So as a strategist, as the manager, I have a little bit more insight about my agents, how they're performing, how many articles are we writing, things like that.

So this is an interesting way of managing not just our incoming tickets, but using AI to assist the troubleshooting, assist with the drafting of knowledge articles as well as our standard responses.

I know this is a long video but if you want a one on one demo you're welcome to let me know and I'll walk you through it.

I'll share the link below.

Have a good day.