<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by João Cadidé on Medium]]></title>
        <description><![CDATA[Stories by João Cadidé on Medium]]></description>
        <link>https://medium.com/@joaocadide?source=rss-2c8976c6b99------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*tiuldSeb1prAgYHhrXJF7w.jpeg</url>
            <title>Stories by João Cadidé on Medium</title>
            <link>https://medium.com/@joaocadide?source=rss-2c8976c6b99------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 18:00:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@joaocadide/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Learn how to automatically analyse solved tickets in Zendesk with ChatGPT and Zapier webhooks.]]></title>
            <link>https://medium.com/@joaocadide/learn-how-to-automatically-analyse-solved-tickets-in-zendesk-with-chatgpt-and-zapier-webhooks-fbbb786ec366?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/fbbb786ec366</guid>
            <category><![CDATA[zapier]]></category>
            <category><![CDATA[zendesk]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[reporting]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Wed, 30 Oct 2024 12:27:49 GMT</pubDate>
            <atom:updated>2024-10-30T12:27:49.010Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yZCyCRqPt0Yn3OiWQClxeQ.jpeg" /><figcaption><em>Photo by </em><a href="https://unsplash.com/@campaign_creators?utm_content=creditCopyText&amp;utm_medium=referral&amp;utm_source=unsplash"><em>Campaign Creators</em></a><em> on </em><a href="https://unsplash.com/photos/person-using-macbook-pro-pypeCEaJeZY?utm_content=creditCopyText&amp;utm_medium=referral&amp;utm_source=unsplash"><em>Unsplash</em></a></figcaption></figure><p>In the <a href="https://medium.com/@joaocadide/more-ai-add-ons-for-your-zapier-integration-with-zendesk-and-openai-bfb719777457">previous article</a>, I covered how to analyse new tickets on Zendesk using Zapier and ChatGPT/OpenAI to automatically get the ticket summary, category, sentiment and next steps. But what happens once a ticket is closed? At that point, we have more context — messages between the requester and agent, internal notes, and the resolution itself.</p><p>This added context raises useful questions: <em>Why was this ticket closed? Was it a user error, a bug fix, or maybe a completed request?</em> To answer these questions automatically, we’ll use ChatGPT to analyze closed tickets and add a “Reason for Closing.”</p><p>Let’s once again use ChatGPT to automatically analyse tickets for us once they’re solved to determine the reason for closing.</p><p>We’ll be using a different method to trigger the analysis: webhooks. No idea what they are? Let’s check a quick explanation:</p><h3>What are webhooks?</h3><blockquote><strong>Think of webhooks as a way for two apps to chat — one app triggers an event, and the other app gets notified</strong>.</blockquote><p>Zapier has a more detailed explanation on <a href="https://zapier.com/blog/what-are-webhooks/">their blog</a> if you’re interested.</p><h4><strong>Advantages of webhooks vs the “New Zendesk ticket” trigger</strong></h4><p>Webhooks can be sent from Zendesk following any conditions of your choice. Maybe you want your ticket to be analysed after <em>every new message</em> or when the <em>ticket is escalated</em> or <em>marked as solved</em>. It is a lot more flexible than the simple “New Zendesk ticket” trigger we used before.</p><p>Let’s dive right in!</p><h3>Adding “Reason for Closing”</h3><h4>Step 1: Creating a new Zap for the “Ticket solved” workflow</h4><ol><li>Duplicate the <a href="https://medium.com/@joaocadide/more-ai-add-ons-for-your-zapier-integration-with-zendesk-and-openai-bfb719777457">previous Zap</a> to make things easier.</li><li>Change the first step (your Zap trigger) from Zendesk to <strong>Webhooks by Zapier</strong>.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i_1n4hiiy96Povpp_Tq_PQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H5_EoeX0KCrP2QBEdFIvVQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*omUZNa-0HJ_ujBdPHWXxIg.png" /></figure><p>3. Copy the webhook URL from the Test tab:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A1vQibEzsBguz9WfXl8gnw.png" /></figure><p><em>You can always come back to this step if you need to copy the webhook URL again.</em></p><h4>Step 2: Creating a webhook on Zendesk</h4><p>We’re now going to connect your webhook created on Zapier to your Zendesk.</p><ol><li>On your Zendesk Admin Centre, Go to <em>Actions and webhooks -&gt; Webhooks</em>.</li><li>Click on <em>Create webhook</em>:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*63e1Za7VYOXJByCTLFXM0w.png" /></figure><p>3. On the Create webhook screen, select <em>Trigger or automation</em>.</p><p>4. Give it a name, such as <em>“Zapier — ticket solved analysis”,</em></p><p>5. Past the webhook URL copied from Zapier in the <em>Endpoint URL </em>field.</p><p>6. Keep the other settings as they are. It should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BeqT71AvKlWu0s9d8kB67A.png" /></figure><p>7. Test your webhook. This will send the JSON message to Zapier.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jVp1cOmosbG6u6EUE4BW4A.png" /></figure><p>8. Go back to Zapier and click on Test trigger. The message from the previous “<em>Test webhook</em>” should show up as a new request:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L7Ycfwx6ezumMG2iHgCbRA.png" /></figure><p>9. Save your webhook on Zendesk.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fOF1WT-hwiPf22g6k-oDrQ.png" /></figure><p>If your test message from Zendesk shows up on Zapier, you’re ready for the next step: <strong>send actual ticket content to Zapier</strong>.</p><blockquote>Check out <a href="https://support.zendesk.com/hc/en-us/articles/4408839108378-Creating-webhooks-to-interact-with-third-party-systems">this Zendesk guide</a> for more on Zendesk webhooks.</blockquote><h4>Step 3: Create a trigger on Zendesk to send the webhook message</h4><p>This step ensures that your Zap runs every time your conditions are met. In our case, we want the Zapier to run when any ticket is marked as “Solved”.</p><ol><li>In your <em>Zendesk Admin Centre</em>, find <em>Triggers</em>.</li><li>Click on <em>Create trigger</em>.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3eaLYNo7f9F6I6hPPMBdXQ.png" /></figure><p>3. Give your trigger a name. We’ll choose “Ticket solved — send it to Zapier for analysis”</p><p>4. In <em>Meet ALL the following conditions</em>, select:</p><ul><li><strong>Ticket &gt; Ticket Is Updated</strong></li><li><strong>Ticket &gt; Ticket Status Changed to Solved</strong></li></ul><p>These conditions ensure that this trigger is run only when a ticket has its status changed to Solved. It should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EVYrgNHLY1RXclGd8eIOjg.png" /></figure><p>5. Add actions for your trigger. We want to send a message using a webhook, select <strong><em>Notify by &gt; Active webhook </em></strong>and select the webhook created in the previous step.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iISxGraNR3hcfeBW4sjicA.png" /></figure><p>6. In the JSON body field, add the following:</p><pre>{<br>      &quot;ticket&quot;: {<br>        &quot;id&quot;: &quot;{{ticket.id}}&quot;,<br>        &quot;priority&quot;: &quot;{{ticket.priority}}&quot;,<br>        &quot;agent&quot; : &quot;{{ticket.assignee.name}}&quot;,<br>        &quot;requester&quot; : &quot;{{ticket.requester.name}}&quot;,<br>        &quot;requester_email&quot;: &quot;{{ticket.requester.email}}&quot;,<br>        &quot;organization&quot; : &quot;{{ticket.organization.name}}&quot;,<br>        &quot;subject&quot;: &quot;{{ticket.title}}&quot;,<br>        &quot;ticket_messages&quot;: &quot;{{ticket.comments_formatted}}&quot;<br>      }<br>    }</pre><p><em>If there are other fields you’d like to add to your webhook message, check this </em><a href="https://support.zendesk.com/hc/en-us/articles/4408886858138-Zendesk-Support-placeholders-reference"><em>full list of Zendesk placeholders</em></a><em>.</em></p><p>7. Click on Create and your trigger is saved!</p><p>To test this Zendesk trigger, mark a ticket as solved to see the message go to Zapier.</p><h4>Step 4: Tell ChatGPT what to do</h4><p>In this final step, we’re going to change a few things in our Zap’s ChatGPT prompt so it takes into account all the ticket messages.</p><ol><li>On your first Zap step, <em>Catch Hook</em>, select “Find new records” in the Test tab. Select a ticket you’d like to use as an example.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lTbO3csp-DTw2Ra10eEedQ.png" /></figure><p>2. On the ChatGPT step of your Zap, change the variables of the Unstructured Text field to match those of the webhook:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J3kHXn_JBWKjDSK2IfP7YQ.png" /></figure><p>3. In <em>Values to Extract</em>, remove unnecessary variables and add reason for closing</p><p>4. In What is the description of “reason for closing”, add a prompt for ChatGPT to explain what it should output for this variable. Here’s an example:</p><p><em>“Read the full ticket thread now that it’s solved and determine the reason for solving it. A reason for closing a ticket could be “Bug fixed by engineering”, “End-user error”, “Refund issue” and so on — it should explain the outcome or what we did to solve it. Write it in a short sentence.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NktB3jDtEq_ppWQGqnlvag.png" /></figure><p>5. Test the new prompt:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DEoR8uJ0NN0_HqlrTS_WdA.png" /></figure><h4>Step 5: Update the “Reason for closing” Zendesk field automatically</h4><p>In this final step, we’re going to update the Zendesk ticket with the output we’ve just got from ChatGPT.</p><ol><li>Go to the Update ticket step on your Zap and ensure that the <em>Ticket </em>field has the ticket ID variable from your webhook step:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AtdMi5w8aiie-GGvsjZmFA.png" /></figure><p>2. Find your <em>Reason for Closing</em> variable and add the <em>Reason for Closing</em> response from your ChatGPT step:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fh5_WrdWXcbvxAuLyFecog.png" /></figure><p>3. Test the step and check what’s changed on your ticket:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VlmZDShFlXHj62EhQipSSw.png" /></figure><p>All done! Now all you need to do is publish the Zap which will run automatically when a ticket is marked as solved.</p><h3>Other considerations</h3><ul><li>Zapier does not support authenticated webhooks at the time of this writing (October 2024). This means that your data could be potentially intercepted by bad actors. Make sure to check with your colleagues to ensure there are no security concerns</li><li>If you have many ticket groups and queues, you can add a new condition such as “Ticket group &gt; Support L1” in your Zendesk. That way you can be more granular about what Zapier should analyse or not.</li></ul><h3>Conclusion</h3><p>And there you have it! By setting up automated ticket analysis using Zapier, webhooks, and ChatGPT, you’ve just added a powerful layer of insight to your Zendesk reports.</p><blockquote>Now, instead of manually sifting through closed tickets, you get clear, actionable insights on why issues were resolved, all updated in real time. This setup can save countless hours for support teams and make tracking resolutions effortless.</blockquote><p>Whether you’re looking to enhance customer service metrics or identify patterns in your ticket data, automating with AI gives you an edge.</p><p>Feel free to <a href="https://www.linkedin.com/in/joaocadide">contact me on LinkedIn</a> if you’d like to discuss similar ideas.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fbbb786ec366" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[More AI add-ons for your Zapier integration with Zendesk and OpenAI]]></title>
            <link>https://medium.com/@joaocadide/more-ai-add-ons-for-your-zapier-integration-with-zendesk-and-openai-bfb719777457?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/bfb719777457</guid>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[zendesk]]></category>
            <category><![CDATA[customer-support]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Mon, 14 Oct 2024 14:16:28 GMT</pubDate>
            <atom:updated>2024-10-14T14:16:28.104Z</atom:updated>
            <content:encoded><![CDATA[<h3>More AI add-ons for your Zapier integration with Zendesk and ChatGPT</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yq7IlY2v1CMRa0xq" /><figcaption>Photo by <a href="https://unsplash.com/@boliviainteligente?utm_source=medium&amp;utm_medium=referral">BoliviaInteligente</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Imagine you’re a product or support manager and would like to visualise why your customers have been contacting your company in the easiest way possible. Suppose you’ve done enough ticket categorisation and followed my previous guide, “<a href="https://medium.com/@joaocadide/using-zapier-and-chatgpt-to-categorise-zendesk-tickets-and-save-you-money-and-time-b380233798a1">Using Zapier and ChatGPT to categorise Zendesk tickets</a>”. In that case, you already have high-level categories in a field — which is very helpful.</p><p><em>But that’s not enough.</em> It is still painful to manually review tickets to gain more context, as it takes a long time to read them individually. It would help if you had an easy way to view what happened in each of your thousands of tickets without having to read long threads or rely on inaccurate ticket subjects.</p><p>Following up on my <a href="https://medium.com/@joaocadide/using-zapier-and-chatgpt-to-categorise-zendesk-tickets-and-save-you-money-and-time-b380233798a1">previous article</a>, let’s expand on other Zendesk + Zapier + OpenAI integration capabilities to achieve new and exciting use cases:</p><ol><li>Summarise the initial message;</li><li>Analyse customer sentiment;</li><li>Suggest the next steps to the agent.</li></ol><h3>Benefits</h3><ul><li>More granular data in your reports.</li><li>Reports become easier to read and share.</li><li>Anyone can quickly understand the context of a ticket, including customer sentiment.</li><li>Time saved! 🧘</li></ul><p>Follow the steps today to enhance your Zendesk reporting experience.</p><h3>Creating new AI power-ups ✨</h3><h4>Step 1: Creating new Zendesk fields</h4><p>If you want to easily visualise the information you need in a<a href="https://support.zendesk.com/hc/en-us/articles/4408831710618-Getting-started-with-Zendesk-Explore-for-reporting-and-analytics"> Zendesk Explore</a> report, you should create fields for each of the categories:</p><ul><li>Summary</li><li>Customer sentiment</li><li>Reason for closing</li></ul><p><em>(For “Next steps”, I don’t see the benefit of having it as a field as, from my experience, it’s not information needed for reporting. We’ll use internal notes instead.)</em></p><ol><li>Go to your Zendesk Admin Centre</li><li>Find the <em>Fields</em> section and click on <em>Add field:</em></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nIO8TD7GYTJp18XJQ49XEQ.png" /></figure><p>3. Add the text fields for <em>Summary</em>, <em>Customer sentiment</em> and <em>Reason for closing</em>. Make sure that the Permission selected is <strong>“Agents can edit”</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EUUbNSt2dUu2mICOYiu6MA.png" /></figure><p>4. Add the newly created fields to your ticket form.</p><p>4.1. Find the section <em>Forms </em>and select your ticket form</p><p>4.2. Drag and drop the new ticket forms from the right-hand side into the middle box until it looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4GhLIJWrkq_6YpoO4BI9dQ.png" /></figure><p>Your tickets are now ready to receive more analysis from ChatGPT.</p><h4>Step 2: Adding new prompts on Zapier</h4><p>We will use the same Zap we created in the <a href="https://medium.com/@joaocadide/using-zapier-and-chatgpt-to-categorise-zendesk-tickets-and-save-you-money-and-time-b380233798a1">previous article</a>. This will save us time and money as we will reuse the same ChatGPT prompt as before.</p><ol><li>In your Zap, go to the <em>Extract Structure Data</em> step and open the <em>Configure</em> tab.</li><li>Add new Values to Extract:</li></ol><ul><li>Summary</li><li>Customer sentiment</li><li>Next steps</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xMbNOsCdoAJ0nw_sEzEoRA.png" /></figure><p>3. Configure prompts for the new fields. <em>What is the description for “[FIELD]”?</em></p><p>I used the simple prompts below, but you can personalise them to be more appropriate to your use case:</p><ul><li><strong>Customer sentiment</strong>:<br>Prompt: “<em>Analyse the customer sentiment from the message and assign one or more sentiments from the possible values.</em>”</li><li><strong>Summary</strong>:<br>Prompt: <em>“Summarise the ticket in one sentence to include the key points such as who they are, why they contacted us and any relevant keywords.”</em></li><li><strong>Next steps:</strong><br>Prompt: <em>“Read the content and give the support agent suggestions on what to do next so that the ticket is resolved as soon as possible. Write them in a numbered list.”</em></li></ul><p>4. Add possible values for <em>Customer sentiment</em>. You can use the list below or personalise it:</p><p><em>Frustration, Discontent, Impatience, Confusion, Anger, Disappointment, Concern, Gratitude, Hopefulness, Trust, Indifference, Urgency</em></p><p>Your Configure tab should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NnMOxkDat486MHcwiv-wAA.png" /><figcaption>Configuration for “Next steps”</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FywqY4a-sJgLPN4bYacTqA.png" /><figcaption>Configuration for “Customer sentiment”</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DVRNPneeK5dEWruWYAw_4g.png" /><figcaption>Configuration for “Summary”</figcaption></figure><p><strong>Let’s test everything in the test tab by clicking on the <em>Test step</em>:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TAWlSwFazg-ev-j97NpEFw.png" /></figure><h4>Step 3: Updating the Zendesk ticket</h4><p>Now that we have new data generated from ChatGPT in our Zap, we can skip to the <em>Update Ticket</em> step.</p><ol><li>Go to the <em>Update Ticket</em> step on your Zap and open the <em>Configure</em> tab.</li><li>Find the ticket fields you created previously. They should be at the bottom of the <em>Configure</em> tab.</li><li>Add the variables for <em>Summary</em> and <em>Customer Sentiment</em> from your <em>Extract Structured Data</em> step.</li></ol><p><em>Using variables from the previous step ensures that the text for Summary and Customer Sentiment are always dynamic and represents the context of the current ticket being analysed.</em></p><p>For Summary:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IFsY_sdY6ukr-SnkfeUR3g.png" /></figure><p>For Customer Sentiment:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PigwXaJwFWQkCCPTZiD4Eg.png" /></figure><p>5. Add <em>Summary</em> and <em>Next steps</em> as an internal note in your ticket:<br>5.1.In <em>New Comment to add to the Ticket</em>, add the comment below:</p><blockquote>*Here’s a short summary for this ticket*:<br>Summary from ChatGPT</blockquote><blockquote>*Suggested next steps*:<br>Next steps from ChatGPT</blockquote><p>5.2. Set<em> New Comment forma</em>t as <em>Plaintext/Markdown</em></p><p>5.3. ⚠️ Set <em>Should the new Comment be public</em> to <em>No</em>. This ensures that the requester will not see the internal comment. It should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HH84aR1N6o2_a24ZyZlK3w.png" /></figure><p>6. Test the step.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-xk1DEHWCxOCWnrmv0bKRg.png" /></figure><p>7. Check the ticket to confirm that the fields have been updated and the internal note is shown:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a4F98-kWzKMQdXsoa8PXQQ.png" /></figure><p>8. Publish the Zap so it runs automatically whenever a new ticket arrives.</p><p><strong>Success</strong>! 🎉 Now you’re getting even more useful data that can be easily viewed in a report and shared with colleagues. And the best part is: it’s all done automatically for you!</p><h3>Conclusion</h3><p>Now, it’s very easy to create a reusable<a href="https://support.zendesk.com/hc/en-us/articles/4408821589530-Creating-reports"> Zendesk Explore report</a> that clearly shows the data in a CSV document, for example.</p><p>These are just some examples of what’s possible when you integrate Zendesk and ChatGPT/OpenAI using Zapier. The limit here is just your imagination!</p><p>—</p><p>Feel free to <a href="https://www.linkedin.com/in/joaocadide">contact me on LinkedIn</a> if you’d like to discuss similar ideas.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bfb719777457" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Using Zapier and ChatGPT to categorise Zendesk tickets (and save you money and time!)]]></title>
            <link>https://medium.com/@joaocadide/using-zapier-and-chatgpt-to-categorise-zendesk-tickets-and-save-you-money-and-time-b380233798a1?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/b380233798a1</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[zapier]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[zendesk]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Wed, 02 Oct 2024 11:15:31 GMT</pubDate>
            <atom:updated>2024-10-02T11:22:33.005Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EsdIapY0Re0l0no5-aVP5g.png" /></figure><p>As a Product Manager in the Merchant Support squad at BVNK, my team and I always tried to find opportunities to improve the Merchant Support team’s workflows to <strong>drive efficiency</strong>, <strong>reduce time to resolve tickets</strong>, <strong>improve reporting</strong> and <strong>increase customer satisfaction</strong>.</p><p>See, Support teams are a great source of feedback for Product teams in tech companies as they provide a massive dataset of customer feedback and can <strong>ultimately help Product teams focus on improvements that customers are reporting.</strong></p><p>However, categorising tickets is still a manual process in ticketing systems like Zendesk: support agents will deal with the query (which can be, at times, very complex) and have to remember to set the right category every time, which can be time-consuming and inaccurate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x0msgVhVQDLCNV2hdkUBhw.png" /><figcaption>“Views” on Zendesk allows agents to see all threads.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QfsoIsyBeOOO-Wc2lCyz0A.png" /><figcaption>The ticket view on Zendesk. On the left-hand side, tickets can be categorised using custom fields.</figcaption></figure><blockquote>We live in the world of AI, so… What if we could leverage LLMs (Large Language Models) to free up the agents’ time for what matters: ticket resolution? As a bonus, Product teams get more accurate reports!</blockquote><h3>Why wasn’t Zendesk’s AI solution ideal?</h3><p>Zendesk offers an AI add-on that <a href="https://support.zendesk.com/hc/en-us/articles/5524125586330-About-Zendesk-Advanced-AI">analyses the customer’s intent, language, and sentiment amongst other features</a>. This sounds interesting, but…</p><p>The solution costs a whopping <a href="https://www.zendesk.co.uk/pricing/">$50/£40 per agent per month</a>! And <strong>it did not auto-categorise tickets using custom ticket fields</strong>, which, in my opinion, is a very useful use case.</p><h3>Here’s Zapier</h3><p>I love no-coding tools. I studied Computer Science and enjoyed learning computational logic, conditions, etc. Zapier brings all of that to life without having to do (much) code and it helps automate real-life workflows.</p><p>Using the power of Zapier and its API connections, we can (fairly easily, with the guide below) <strong>connect Zendesk to OpenAI’s AI models (that power ChatGPT) to automatically analyse and categorise support tickets.</strong></p><p>Zapier opens up a whole new world to your Zendesk tickets, so here are some ideas of what you can do:</p><ol><li><strong>Auto-categorise tickets using a custom field</strong></li><li>Summarise the initial message</li><li>Analyse customer sentiment</li><li>Suggest next steps to the agent</li><li>Notify Slack channels of new tickets</li><li>And more! The options are endless!</li></ol><p>We’re going to be focusing on the first use case here but I’ll also give some suggestions for the other points.</p><h3>How much will it cost?</h3><p>The Zapier Professional plan starts at $29.99 a month; the OpenAI tokens may vary, but for my small examples in this guide, each ticket analysis using GPT-4o-mini cost me $0.01. Bargain!</p><blockquote>In a company with 20 agents on Zendesk dealing with 200 small tickets a month, using Zapier and ChatGPT is 96.7% cheaper versus the Zendesk AI solution ($31.99/month versus $1000/month)!!!</blockquote><h3>Integrating Zendesk with ChatGPT/OpenAI using Zapier</h3><p>Finally, the guide! This will only need to be set up once and, when published, Zapier will do the rest for you.</p><h4><strong>Requirements</strong></h4><ul><li><a href="http://zapier.com/">Zapier</a> account (Pro version is required for Zendesk)</li><li><a href="http://zendesk.com/">Zendesk account</a> with a custom field used for ticket categories.</li><li><a href="https://platform.openai.com/api-keys">OpenAI API key</a> (with credits — this is different from a ChatGPT account. You can top up your account with as little as $5)</li></ul><h4><strong>Step 1: Connect Zendesk to Zapier</strong></h4><ol><li>Create a <a href="https://zapier.com/app/zaps">new Zap on Zapier</a>.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7btRk6D4uOl6Eus0MnnbNA.png" /></figure><p>2. Add Zendesk as your first trigger (the first block in your flow)</p><p>3. In <em>Trigger event</em>, select <em>New ticket</em>. This means that this Zap will run every time a new ticket arrives on your Zendesk instance (this is verified every 15 minutes).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qb4Zn9ckgysxFwbIwZCsYg.png" /></figure><p>4. Connect your Zendesk account, if not already connected. <em>(Here’s how you can </em><a href="https://support.zendesk.com/hc/en-us/articles/4408889192858-Managing-access-to-the-Zendesk-API#topic_bsw_lfg_mmb"><em>generate a Zendesk API token</em></a><em>)</em></p><p>5. In the “Organization” tab, select the organization your Zendesk tickets are assigned.</p><p>6. Test your trigger on the <em>Test</em> tab. You’ll be able to see all the details from the ticket there. Select one to use as an example.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*axB6-N41OFWdjG72pZdafw.png" /></figure><h4>Step 2: Get ticket categories</h4><p>The second step involves using the Zendesk API to get all the ticket categories so they’re always dynamic and up-to-date. <strong>You can skip this and hardcode your categories but this is not recommended if your categories change over time.</strong></p><p>To configure the Zendesk API call, we’re going to use Zendesk’s<a href="https://developer.zendesk.com/api-reference/ticketing/tickets/ticket_fields/">ticket_fields endpoint</a>.</p><ol><li>Find the ID of your custom field on Zendesk. Go to your Zendesk Admin -&gt; Objects and Rules -&gt; Fields and copy the Field ID.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*foedty9HHTMHnxCBXOXlyw.png" /></figure><p>2. Back to Zapier, in step 2, select Zendesk again as the app.</p><p>3. In the <em>Action event</em> dropdown, select <em>API Request (Beta):</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*km2kPyThaW06dcPYq6dfKA.png" /></figure><p>4. Add the URL to the <em>URL </em>field containing your Zendesk instance and the field ID.<br>Example: <a href="https://yourcompany.zendesk.com/api/v2/ticket_fields/CUSTOM_FIELD_ID"><em>https://YOURCOMPANY.zendesk.com/api/v2/ticket_fields/CUSTOM_FIELD_ID</em></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NXaN-jh6h00w5g2HjoL7ng.png" /></figure><p>5. Test this step to make sure everything works fine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ckdaFc7lSefle_zhpfaoJQ.png" /><figcaption>If your test works fine, you’ll get the custom field values in the response</figcaption></figure><h4>Step 3: Analyse the ticket using ChatGPT/OpenAI</h4><p>Now we’ve come to the fun part! 🥳 Let’s create another step in our Zap to send the ticket content and categories from the previous steps to ChatGPT/OpenAI.</p><ol><li>Select ChatGPT as the app for the 3rd step.</li><li>Choose <em>Extract Structure Data </em>as the Action event. This event ensures that ChatGPT outputs exactly what we want, in the right format.</li><li>Connect your OpenAI account. Make sure it is <a href="https://platform.openai.com/settings/organization/billing/overview">funded</a> and an <a href="https://platform.openai.com/api-keys">API key has been created</a>.</li></ol><p>It should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QOkNa_XvlDtX9bMxc7AWSA.png" /></figure><p>4. Let’s set up our OpenAI prompt in the <em>Configure</em> tab:</p><ul><li><strong>Unstructured Text:</strong> add the relevant ticket information that will be analysed. I recommend the following: Requester Name, Requester Email, Subject and Description (the first message in the ticket).</li><li><strong>Model</strong>: gpt-4o-mini does the job fine and it <a href="https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/">costs less</a>.</li><li><strong>Temperature</strong>: leave it at 0.0 for more accuracy.</li><li><strong>Description</strong>: here’s where you can describe this GPT’s persona and what your company does. It’s good to give it some context here so it can understand your requests better (e.g.: if you work for a financial company, explaining some acronyms like POP for Proof of Payment, can help).</li><li><strong>Values to extract</strong>: this is where you can ask GPT to generate as an output. Let’s type category for this example.</li><li><strong>What is the description for “category”?</strong>: if your categories are particularly complex, give some details to assist with categorisation here.</li><li><strong>What are the possible values for “category”?</strong>: select the Response Data from Step 2. You should be able to see your categories in plain grey text in there. If you decide to hard code your categories, this is where you’ll manually input the values.</li></ul><p>Your Step 3’s <em>Configure</em> tab should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s1vGVuWmT98I76Uhwdsfyg.png" /></figure><p>Let’s test the flow!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-JXB3m3B1a-YVFIqV4EG8g.png" /></figure><p><strong>Now your ticket is automatically categorised by ChatGPT! </strong>Next, we’re going to make Zapier save this category in the ticket.</p><h4>Step 4: Mapping the category value with the label</h4><p>Now that we have the category chosen by ChatGPT, we’re going to create a step so that it is also set in the ticket. Before we do that, <strong>we need to map the category in plain text to its tag.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4XN7v9LUKdGM3hPXMOkuTw.png" /><figcaption>On Zendesk, each field value has a tag</figcaption></figure><p>This step is needed because Zendesk will not accept plain text such as Missing payment as the input on Zapier; instead, we need to use its matching tag.</p><ol><li>In Step 4, select <em>Code by Zapier</em> as the app and <em>Run Python</em> as the <em>Action event</em>.</li><li>In Input Data, let’s add two values:</li></ol><ul><li>categories : this is the Response Body from Step 2 that contains all category values in plain text and their tags.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y3mY96mWT2EasW7esm5oQg.png" /></figure><ul><li>category_selected: the category ChatGPT has chosen.</li></ul><p>3. In the <em>Code</em> area, copy and paste the Python code below:</p><pre>import json<br><br># Parse the input data<br>categories_json = input_data.get(&#39;categories&#39;, &#39;{}&#39;)<br>category_selected = input_data.get(&#39;category_selected&#39;, &#39;&#39;)<br><br># Load the categories from the JSON string<br>categories_data = json.loads(categories_json)<br><br># Initialize the output variable<br>output = {}<br><br># Iterate through the custom field options to find the matching raw_name<br>for option in categories_data.get(&#39;ticket_field&#39;, {}).get(&#39;custom_field_options&#39;, []):<br>    if option.get(&#39;raw_name&#39;) == category_selected:<br>        output[&#39;value&#39;] = option.get(&#39;value&#39;)<br>        break<br>else:<br>    output[&#39;value&#39;] = None  # If no match is found, set value to None</pre><p>Your Configure tab should look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V5kYtDD9uJ7cuVn8X5O9yQ.png" /></figure><p>4. Test the step. The <em>Data out</em> tab should give you a simple <em>Value</em> which is the label that matches the category (in my example, my category was Order update and the value is order, which is the Zendesk label).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xpZ7B2LBPqHYGno3gwWmvA.png" /></figure><h4>Step 5: Updating the ticket</h4><p>Now you’ve got everything you need to update the category field on Zendesk.</p><ol><li>In Step 5, select <em>Zendesk</em> as the app and <em>Update Ticket</em> as the <em>Action Event</em>.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TMLdWlIpZFpQeaadVFN2eQ.png" /></figure><p>2. On Ticket, select the ticket ID from Step 1. <strong>This is crucial and should not be manually input.</strong> It ensures that we’re updating the same ticket we started the flow with, not a random one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tT93klkj_9TpcLcCUAfwjw.png" /></figure><p>3. Find your custom field in this screen, in my case, <em>Ticket category</em>. Select the value from Step 4 as the input:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hi-FoquA68eu-dmFcQjCPw.png" /></figure><p>4. Test the step. If successful, this should update your custom field with the value from ChatGPT:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X9b0-WFZJTO3o9mBMW0g8A.png" /></figure><p>Here’s what the ticket looked like before:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r3TGEknOeOrJs4fsdoZr7w.png" /></figure><p>And after:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XWvHJRvkG_-2Hto6Z7iuxw.png" /></figure><blockquote>That’s it! Your flow is done and now you can publish your Zap to ensure new tickets are automatically categorised.</blockquote><h3>Other considerations</h3><h4>Play with prompts</h4><p>This example was a fairly simple one to illustrate what you can do. Every use case is different and you can always adjust your prompts on step 3 to improve the categorisation.</p><h4>More efficient trigger</h4><p>If you have a large amount of tickets coming in, consider changing the trigger to a webhook instead. Webhooks can be sent on Zendesk via triggers and these allow for a lot more customisation (e.g.: you can create rules so only <em>some</em> tickets are sent to Zapier).</p><h4>Security considerations</h4><p>The content of your support queries might contain sensitive information. Make sure the Security team is involved and this flow is approved by them. Show them the productivity and money savings to get their buy-in.</p><h4>Other use cases</h4><p>Now that the flow is done, you can easily add other Values to Extract to step 3 to make your flow even more powerful. Here are some examples:</p><ul><li>Ticket summary</li><li>Customer sentiment</li><li>Suggest next steps to the agent</li><li>Notify Slack channels of new tickets</li></ul><h3>Conclusion</h3><p>If you’ve read up until here, thank you! Let me know in the comments if you have any feedback or if you’re interested in me covering other use cases for this flow.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b380233798a1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What’s Level 0 support, and why does it matter?]]></title>
            <link>https://medium.com/@joaocadide/whats-level-0-support-and-why-does-it-matter-ca393d7c0a73?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/ca393d7c0a73</guid>
            <category><![CDATA[kc]]></category>
            <category><![CDATA[best-practices]]></category>
            <category><![CDATA[help-centre]]></category>
            <category><![CDATA[knowledge-management]]></category>
            <category><![CDATA[knowledge-sharing]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Mon, 01 Nov 2021 14:31:37 GMT</pubDate>
            <atom:updated>2021-11-01T14:31:37.109Z</atom:updated>
            <content:encoded><![CDATA[<h4>Giving your customers the option to self-serve can be easier than you think. And everybody wins.</h4><p>I love self-service. It dreads me when I think I have to call my mobile carrier to upgrade my plan; or that I need to go to a physical branch to set up my bank account. Solving issues by myself is not only <em>more fun</em> but also more rewarding.</p><p>And I’m not the only one here! A recent <a href="https://www.salesforce.com/eu/blog/2020/01/state-of-the-connected-customers-report.html">report</a> from Salesforce shows that 68% of the customers prefer to use self-service channels for simple questions and issues. Furthermore, 56% of them expect to find whatever they need in 3 clicks or less.</p><p>One of my favourite and easiest ways to provide a self-service system is by having a <strong>Help Centre that answers most asked questions (FAQs) and provides quick guides and solutions to issues. This is what some would like to call “Level 0 support”.</strong></p><p>Granted, not every query can be solved by a Help Centre article: empathetic and personalised support also play a big part in giving a great experience. For everything else, having an always-available public knowledge base will provide your customers with constant guidance and independence.</p><p>And that’s not all! Having a robust help centre full of useful articles bring many benefits, such as:</p><ul><li>Repetitive questions can be easily avoided (faster resolution times)</li><li>Single source of truth and consistent answers</li><li>Support agents have more time to up-skill and work on exciting problems.</li></ul><p>There are benefits for everyone: the users, support agents and the business!</p><p>Here are my <strong><em>quick tips</em></strong> for creating useful articles (or <em>capturing knowledge</em>) in your Help Centre. These were heavily influenced by <a href="https://www.thekcsacademy.net/kcs/">KCS</a> (Knowledge Centered Service) and I take them to heart when creating (or capturing) knowledge at work.</p><h3>Create new articles (FAQs) by demand ✍️</h3><p>Imagine that a new article or feature is coming soon and you, one of the Help Centre authors, decide to write as many FAQs as possible. Then… no one reads them!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vYIbMjXUy3KUtzt7KeXGYA.jpeg" /><figcaption>“I spent so much time writing those articles that solve nothing and no one reads.” Photo by <a href="https://unsplash.com/@raduflorin?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Radu Florin</a> on <a href="https://unsplash.com/s/photos/sad-writer?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>This is why creating new FAQs <em>by demand</em> is more effective than just preempting or imagining questions.</p><p>Have you got a new question that has no documented answer? That’s fine! Find the solution and provide it to your user, as you would normally do. Then, <strong>capture the knowledge</strong>. Next time, users facing the same issue will have the solution available straight away.</p><p>Note: There might be cases where you’ll still want to create some content to support a new feature or product. If that’s the case, do a little testing or focus group and take note of the questions that might come up. <strong>This is the best way to emulate a real-life scenario.</strong> You can use those to start your knowledge base with some successful content.</p><h3>Replicate the requester’s original question ⁉️</h3><p>You’ve got a solution that’s now documented in your Help Centre, but users can’t find it. The internet has come a long way with searching capabilities (thanks Google, Bing and DuckDuckGo). However, putting yourself in the users’ shoes still helps articles have better findability.</p><p>Make your articles easy to find. <strong>Use the original requester’s question as to the title of your article.</strong></p><p>Example: <em>Instead of “Mobile coverage in the UK” try “Do you cover my town?” or “What cities are covered in the UK?”.</em></p><p>If your Help Centre tool also supports it, <strong><em>tags</em></strong><em> </em>are a great way to add synonyms and keywords. For the example above, “coverage”, “roadmap” and “availability” would be good ideas for tags.</p><h3>If it’s enough to solve, publish it! ▶️</h3><p>Does your article have only one line? Does it solve the question in the title? That’s fine. You can go ahead and make it public!</p><p>Remember that you can always iterate and add more information to it over time. Your content, however small, will already add value to your users.</p><h3>Make the content simple. 🥘</h3><p>Help centre articles <strong>are not manuals or contracts</strong>! Think of them as structured recipes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RFy6PtVgrO7MQ9sl9NAVFw.jpeg" /><figcaption>You know exactly where to find ingredients in a recipe, thanks to its structure — Photo by <a href="https://unsplash.com/@picsbyjameslee?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">James Lee</a> on <a href="https://unsplash.com/s/photos/recipe?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>There’s no need to overcomplicate your articles by adding loads of information that no one will read. <strong>People go to a help centre to get quick answers.</strong></p><p><strong>Do: </strong>✅</p><ul><li>Use headings to split sections.</li><li>Use tables to organise complex data</li><li>Add bullet points for general lists</li><li>Use numbered lists for steps (or if you want the user to refer back to a specific point on the list</li><li>Create one-line articles that solve the main issue</li></ul><p><strong>Don’t</strong>: ⛔️</p><ul><li>Write long paragraphs that make the content hard to read. Users go to your help centre to find quick solutions to their problems, not to read a big chunk of text. <em>(Ironically, this is a long paragraph and I thank you I’ve you’re still with me!)</em></li></ul><h3>Add multimedia 📸</h3><p>People have different <a href="https://tracyharringtonatkinson.com/barbes-vak-learning-style/">learning styles</a>, and although you can’t capture all of them in a help centre, there’s a lot of them that can be used and can make your articles richer:</p><ul><li>Diagrams (Lucidchart, Draw.io)</li><li>Pictures</li><li>Videos (How to embed <a href="https://support.google.com/youtube/answer/171780?hl=en-GB">YouTube</a> and <a href="https://vimeo.zendesk.com/hc/en-us/articles/224969968-Embedding-videos-overview">Vimeo</a> videos)</li><li>Quizzes (<a href="https://quizlet.com/en-gb">Quizlet</a>, <a href="https://quizizz.com/">Quizizz</a>)</li></ul><p>And whatever else your imagination and creativity allow.</p><h3>More</h3><p>If you’re also passionate about knowledge management, I highly recommend <a href="https://www.thekcsacademy.net/kcs/">KCS — Knowledge Centered Service</a>.</p><p>I also recommend reading on Information Architecture, which is super relevant for creating organised support/help centres.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ca393d7c0a73" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Making employee training fun and engaging with CELTA techniques]]></title>
            <link>https://medium.com/@joaocadide/making-employee-training-fun-and-engaging-with-celta-techniques-119a62ce8fe8?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/119a62ce8fe8</guid>
            <category><![CDATA[employee-training]]></category>
            <category><![CDATA[employee-engagement]]></category>
            <category><![CDATA[celta]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Tue, 28 Apr 2020 19:56:37 GMT</pubDate>
            <atom:updated>2020-04-28T19:56:37.661Z</atom:updated>
            <content:encoded><![CDATA[<p>As a former language teacher and Computer Science graduate, I’ve always loved both sides of the coin: teaching a language made me feel alive and it was super fun (cheesy, but true); technology has always fascinated me from breaking computers to fixing them in the dead of night. I now have the opportunity to put both skills to use and deliver presentations and training sessions at TrueLayer using a lot I’ve learnt from my teaching years.</p><p><em>CELTA stands for </em><strong><em>Certificate in Teaching English to Speakers of Other Languages</em></strong><em> and it’s offered by Cambridge English. You can find more information </em><a href="https://www.cambridgeenglish.org/teaching-english/teaching-qualifications/celta/"><em>here</em></a><em>. I highly recommend it!</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/433/1*670CJ0uKZ-nm37uqM7NDqA.jpeg" /><figcaption>That’s me last March 2020 introducing a new tool at TrueLayer. Had a blast!</figcaption></figure><p>I plan to write about some of the techniques I learned during my CELTA course and how I’ve been applying them in employee training in my past two jobs. These concepts can be used not only in formal <em>onboarding training</em> but also in common presentations. <strong>They aim to keep the audience engaged and motivated throughout the whole process, as well as allow trainers to check the efficacy of the knowledge-sharing process.</strong></p><p><em>Throughout this series of posts, I will use the words “teacher” and “trainer” interchangeably as well as “trainee” and “student”.</em></p><p>Let’s dive into the first one.</p><h3>Contextualising</h3><p>It is tempting to introduce a new topic and just start talking about it. Learning is proven to be more effective when a context is put into place as it <em>“activates areas of the brain which relate to learners’ experience in relevant areas, known as schema”</em> (<a href="https://hongkongtesol.com/blog/2019/05/importance-context-english-teaching-experience-situation-and-surrounding-language">source</a>). <strong>Context makes the content more memorable and relatable as it allows learners to refer back to the topic in the future quickly.</strong></p><h4>Don’t:</h4><ul><li>Present new concepts in an isolated form</li><li>Assume that the audience will understand all the acronyms and technical terms</li></ul><h4>Do:</h4><ul><li>Introduce a story that can be revisited and connect all topics within a presentation</li><li>Make analogies</li><li>Use pictures and other media to aid with your explanation</li></ul><h4>Example</h4><p><strong>You want to explain what an <em>API</em> is</strong></p><p>A trainer could just throw the definition of “API”, some technical examples and move on. Quick and dirty.</p><p>However, check the story below:</p><p><em>“When you go to a restaurant (assuming you’re a regular person and the place has a good reputation), how do you order the food? Do you care who and how the food is being made? You would probably just select the food and drinks of a menu, the waiter/waitress would take the order and you would receive what you ordered, right?</em></p><p><em>Similarly, an API provides developers with a set of functions (menu) that they can use to make requests (order food) and get responses (finally getting the food delivered).”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vdV4-gxI3cG5uwOVAeD_bw.jpeg" /><figcaption>Everyone’s so happy with the <strong><em>responses</em></strong> they got that they look like they’ll leave a big tip. Photo by <a href="https://unsplash.com/@priscilladupreez?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Priscilla Du Preez</a> on <a href="https://unsplash.com/s/photos/storytelling?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>And this is just the introduction. Throughout the training session, you can always refer back to the restaurant context.</p><p><em>“What would be a</em><strong><em> successful request</em></strong><em> in a restaurant? Ordering something that is on the menu!</em></p><p><em>How about an </em><strong><em>unsuccessful response</em></strong><em>? If you order sushi and the waiter gives you nothing, that’s probably not successful!”</em></p><p>And so on.</p><p>Contextualising topics and giving typical analogies work very well on non-technical students or people from other areas that might not be familiar to your topic.</p><p><strong>Have you ever used creativity and thought outside the box to give a training session or presentation? Feel free to leave your comments below.</strong></p><h4>References</h4><p>These are some of the websites and videos I used to create this post:</p><p><a href="https://www.howtogeek.com/343877/what-is-an-api/">What’s an API — How to Geek</a></p><p><a href="https://www.teachingenglish.org.uk/article/contextualisation">Contextualisation — Teaching English — BBC</a></p><p><a href="https://youtu.be/s7wmiS2mSXY">What’s an API — Mullesoft Videos</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=119a62ce8fe8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Client Care at TrueLayer: where Sherlocks of the Fintech world find their Holmes]]></title>
            <link>https://medium.com/@joaocadide/client-care-at-truelayer-where-sherlocks-of-the-fintech-world-find-their-holmes-27f3b43087ad?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/27f3b43087ad</guid>
            <category><![CDATA[hiring]]></category>
            <category><![CDATA[customer-support]]></category>
            <category><![CDATA[teamwork]]></category>
            <category><![CDATA[client-care]]></category>
            <category><![CDATA[fintech]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Thu, 05 Sep 2019 13:58:50 GMT</pubDate>
            <atom:updated>2019-09-05T14:28:29.057Z</atom:updated>
            <content:encoded><![CDATA[<h3>Client Care at TrueLayer: where the Sherlocks of the Open Banking world find their Holmes</h3><p><em>Do you want to work with smart and energetic people and drive our business forward? TrueLayer’s Client Care is looking for new detectives to join the team and help them solve some unique cases.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*abPUaC8PuR7E4zBF-oMipQ@2x.png" /></figure><p>As you may know, at TrueLayer, we have two main products: <a href="https://truelayer.com/data-api/">Data API</a> and <a href="https://truelayer.com/payments-api/">Payments API</a>. We connect different applications and their users to multiple banks to provide data about their transactions and allow them to initiate payments. We take the complexity out by integrating with many nuanced API endpoints. This means that, on occasion, unexpected things may happen. Fear not, the team of TrueLayer detectives (a.k.a. Client Care) is here to help, answer any questions, and find that special root cause that has been troubling you. As Sherlock once said: “The world is full of obvious things which nobody by any chance ever observes.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-4auEJhI188jTVrX8Ryy_g.jpeg" /></figure><h3>We are diverse and we take work seriously</h3><p>We are a team of three people (soon to be 4) and we want to grow at a pace that the rest of the company is growing at. Much like the famous detective’s encyclopedic knowledge of cigar ashes and London soils, our team also has their passions. : Chris is a DJ, <a href="https://blog.truelayer.com/truetalk-with-naomi-a3734aacc3af">Naomi</a> is passionate about NASA, and Joao used to be a language teacher. We’re that cool! 😎</p><p>And when it comes to the daily grind, we all chip in to help each other. The Sherlocks 🕵 🕵️‍♀️ are constantly thinking of ways of improving how we work and making our clients happy.</p><p>We are a young and thriving company (TrueLayer just turned 3 years old!) which means we’re building new things, creating processes, and learning from our mistakes. This also means that things change and improve fast — which makes working here even more exciting.</p><h3>Our days are never the same</h3><p>It’s 8 am. The kettle has boiled, the tea is brewing, and in our Baker St, ahem, Hardwick Street headquarters, we get ready to throw ourselves upon the next investigation.</p><p>Zendesk is our 221B Baker Street, and much like Holmes’ consultation room, it plays host to a number of visiting clients — or tickets. Some are straightforward and easily solved: these lead to new or existing <a href="http://support.truelayer.com/">FAQs</a>. But what we really love, are the clients whose stories require further investigation.</p><p>Due to TrueLayer’s clever <a href="https://truelayer.com/security/">security model</a>, we don’t store transactions or credentials in a way that we could have access to that data if we wanted to. And this requires some creativity.</p><p>Let’s say we get a request from a client who can’t refresh their tokens (we’ve covered this <a href="https://blog.truelayer.com/managing-your-truelayer-tokens-the-dos-and-don-ts-de22822d335d">here</a>). They send us unique identifiers for the connection, the time and date of the occurrence. Everything seems fine on our side but the error persists. <strong>We have a new mystery to solve!</strong></p><p><strong>We put our detective hats on, grab our (digital) magnifying glass 🔍, and:</strong></p><ul><li>check the logs;</li><li>decode the debug_id;</li><li>try to replicate the issue;</li><li>talk to other teams and brainstorm ideas.</li></ul><p>And finally, we get to the root cause and help our engineering teams build a fix for the issue.</p><p>Mystery solved! “Elementary”, really.</p><p><strong>But that’s not all! We also love to:</strong></p><ul><li>Chat with the clients who reach out using Zendesk chat;</li><li>Create new <a href="https://truelayer.zendesk.com/hc/en-us/sections/360000466514-General-questions">FAQ</a> articles to improve self-service;</li><li>Work on new procedures to make our jobs easier;</li><li>Laugh at good but also at some very bad puns 🙃</li></ul><h3>We love a good challenge</h3><p>One of the best aspects of working with a team of committed people in a company of smart brains is that work <strong>is never boring</strong>. I mean it! We are always full of ideas to build new things to make our lives and our customers’ lives easier.</p><p><strong>Here are some of the things we have been working on:</strong></p><ul><li>Collecting feedback: we implemented an <strong>NPS</strong> (Net Promoter Score) tool to gather feedback and improve our services.</li><li>Holmes had Watson to chronicle his adventures. We have our <strong>Client Care Chronicle</strong>, highlighting all the mysteries we’ve solved, usually with a heavy dose of memes, for our colleagues to appreciate.</li><li>We also meet with new and integrating clients to <strong>build a good relationship</strong> and show them that they can count on us for anything! This helps us shape a service that meets their needs.</li><li>Although we value human interactions, we are constantly building automation to make the processes and resolutions faster. This means more <a href="http://support.truelayer.com/">FAQ</a> articles, scripts and more.</li><li>We <a href="http://www.twitter.com/truelayerstatus">tweet</a> updates to keep everyone in the loop about downtime and resolutions.</li></ul><p><strong>And a lot more!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-4auEJhI188jTVrX8Ryy_g.jpeg" /></figure><h3>Looking forward to the future</h3><p>Big things are coming and we could not be more excited! By the end of the year, we want to be a team of 10 detectives: each one special in their own way with a range of skills that will make our team grow and improve our Client experience.</p><p>With TrueLayer’s expansion to other countries, we want to speak more languages (we speak three so far!), share more knowledge and build new and exciting things. 🎉</p><p><strong>Interested?</strong> <a href="https://truelayer.workable.com/">We’re hiring</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=27f3b43087ad" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Managing your TrueLayer tokens: the do’s and don’ts]]></title>
            <link>https://medium.com/@joaocadide/managing-your-truelayer-tokens-the-dos-and-don-ts-de22822d335d?source=rss-2c8976c6b99------2</link>
            <guid isPermaLink="false">https://medium.com/p/de22822d335d</guid>
            <category><![CDATA[token]]></category>
            <category><![CDATA[data-api]]></category>
            <category><![CDATA[payments-api]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[open-banking]]></category>
            <dc:creator><![CDATA[João Cadidé]]></dc:creator>
            <pubDate>Wed, 12 Jun 2019 08:45:02 GMT</pubDate>
            <atom:updated>2019-10-10T15:07:43.401Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>Learn the best practices around storing and renewing your tokens and avoid common issues</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RoQadeJzSjsXLYu-kPPgYg.png" /></figure><p>You’re an avid developer and have this amazing idea to create an app that changes the way users manage their money — but your users need to send you their data. No need to worry about connecting to banks anymore: <strong>leave this job to TrueLayer!</strong> Provided your user consents, we provide you with access to bank data and are able to initiate payments on many banks in the <a href="https://truelayer.zendesk.com/hc/en-us/articles/360013178154-What-banks-are-we-currently-integrated-with-">UK</a> and EU. We provide developers with guidelines 📝on how to implement the APIs and they can be found <a href="https://docs.truelayer.com/">here</a>.</p><p>Our Client Care team frequently receive queries about tokens and how to manage them. To make the best out of our Data API, here’s a handy guide of best practices for correctly utilising tokens.</p><h3>What are tokens anyway?</h3><p>One of the most fundamental parts of TrueLayer’s Data and Payments API are <strong>tokens</strong>. Thanks to our proprietary technology, <strong>end-users do not share their bank credentials with third-party applications. Ever. </strong>(Curious about how we do it? Check more <a href="https://truelayer.com/security/">here</a>). Instead, you will obtain <strong>a pair of tokens</strong>, that will unlock the power of access to bank data 🔐.</p><p><strong>Tokens</strong> are encoded strings that contain unreadable information to the human eye. When a user connects their bank account to your app, TrueLayer will provide you with two tokens: <strong><em>access_token</em></strong> and <strong><em>refresh_token</em></strong>. Your <strong><em>access_token</em></strong> may look like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xWxB7bAd65DBDy22fzcEWQ.png" /></figure><p>The <strong><em>access_token</em></strong> used by TrueLayer is created based on an open standard called <a href="https://jwt.io/introduction/">JWT</a>. They contain non-sensitive information about the user such as <strong>consents given, expiry dates, provider, <em>credentials_id</em></strong>, etc.</p><p>Keeping your tokens safe and up-to-date guarantees <strong>continuous access to your user’s data</strong> (where consent has been provided) <strong>and perfect functionality of your application!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xGXEd0j0DWUsebYp.jpg" /></figure><h3>The 🐺 lone wolf dies but the pack survives</h3><p>Once your end-user gives TrueLayer and your application permission to access their data offline, you will get a code that you will need to <a href="https://docs.truelayer.com/#exchange-code-with-access-token">exchange</a> for TrueLayer tokens. Our tokens come in pairs. Let’s see them:</p><pre>{</pre><pre>&quot;access_token&quot;: &quot;JWT-ACCESS-TOKEN-HERE&quot;,</pre><pre>&quot;expires_in&quot;: &quot;JWT-EXPIRY-TIME&quot;,</pre><pre>&quot;token_type&quot;: &quot;Bearer&quot;,</pre><pre>&quot;refresh_token&quot;: &quot;REFRESH-TOKEN-HERE&quot;</pre><pre>}</pre><p>The <strong><em>access_token</em></strong> allows you to access the user’s data on their behalf. They are <strong>short-lived </strong>(up to one hour) and must be used in every request.</p><p>When the <strong><em>access_token</em></strong> expires, the <strong><em>refresh_token</em></strong> is needed to <a href="https://docs.truelayer.com/#renew-the-access_token">renew </a>both tokens. You will get a new pair that will allow you to carry on retrieving data — with the user’s consent.</p><h4><strong>Do</strong></h4><p>✅ <a href="https://docs.truelayer.com/#exchange-code-with-access-token">Exchange</a> the code with a pair of tokens<br>✅ Store your tokens securely in your database, always in the same row</p><h4><strong>Don’t</strong></h4><p>❌ Wait longer than 5 minutes to exchange your code — they expire very quickly!<br>❌ Store the matching <strong><em>access_token</em></strong> and <strong><em>refresh_token</em></strong> separately.<br>❌ Share tokens with anyone.</p><h3>Multiple accounts = Multiple tokens?</h3><p>Take our fictional user John Doe 🕵. He really enjoys testing banks and credit cards and is a member of Natwest (with a Current Account, Savings and a credit card), Monzo (current account and pots), Barclays (Current account), and American Express (!!!).</p><p>Being hard for him to track all of his money, he decides to use our demo application, <a href="https://demo.truelayer.com/">PiggyBank</a> that uses TrueLayer’s Data API as the middle-man. John then connects all his accounts above using TrueLayer’s Data API and leaves PiggyBank to handle all the connections.</p><p>To ensure that the tokens are not mixed up or lost, here’s how PiggyBank handles the relationship between the tokens:</p><p>John Doe’s Natwest identity will generate <strong>one </strong>TrueLayer <strong><em>credentials_id</em></strong> which identifies this account in our system. You can see his <strong><em>credentials_id</em></strong> below:</p><pre>cX5HUQflnkWt5u0iUbSR8PHsV7QtN1aWowsI6TYeaH0=</pre><p><strong>One</strong> <strong><em>credentials_id</em></strong> is associated with <strong>one set of credentials</strong> (username, membership number, password, PIN, etc) which will then have an associated pair of <strong><em>access_token</em></strong> and <strong><em>refresh_token</em></strong>.</p><p>The <strong><em>access_token</em></strong> will give you access to all John’s Natwest accounts (Current, Savings and Credit Card).</p><p>John’s connection with Monzo will give <strong>a different </strong>pair of tokens that will be associated with another <strong><em>credentials_id</em></strong>. This <strong><em>access_token</em></strong> gives then access to John’s Monzo Current Account and Pots.</p><p>Have a look at the diagram below to clarify:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yYC4tSREUA6GzO5p.png" /></figure><h4><strong>Do</strong></h4><p>✅ Associate one user’s connection to a bank with a matching pair of tokens (<strong><em>access_token</em></strong> and <strong><em>refresh_token</em></strong>)<br>✅ Use the same <strong><em>access_token</em></strong> to retrieve data from different accounts and cards within the same connection</p><h4><strong>Don’t</strong></h4><p>❌ Use the same tokens for different connections of the same user. <em>E.g.: John’s Natwest token </em><strong><em>can’t</em></strong><em> be used to retrieve John’s Monzo data.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xGXEd0j0DWUsebYp.jpg" /></figure><h3>One pair to rule them all</h3><p>Access tokens are needed to do all sorts of requests: <strong>balance, transactions, accounts, info, cards</strong> and so on.</p><p>As long as it is still valid, the same <strong><em>access_token</em></strong> gives you access to all requests supported. Therefore, <strong>there is no need to save the same pair of tokens in more than one place</strong>. This creates more complexity and could result in errors such as old tokens being left behind, wrong tokens for the wrong credentials, etc.</p><h4><strong>Do</strong></h4><p>✅ Save your pair of tokens in <strong>one place only<br></strong>✅ Make sure that other parts of your application read your tokens from the same place</p><h4><strong>Don’t</strong></h4><p>❌ Save the same tokens in multiple places</p><h3>Keep them fresh! 🍊</h3><blockquote>❗️<strong>Attention</strong>: as for September 2019, refresh tokens <strong>do not change</strong> between requests. The access_tokens will still change. Read more information in our new blog post: <a href="https://blog.truelayer.com/refresh-tokens-refreshed-e4a1da43e5fd"><strong>Refresh tokens — Refreshed</strong></a>.</blockquote><p>For security reasons, <strong><em>access_tokens</em></strong> expire very shortly (up to one hour, depending on the provider). You will need to <a href="https://docs.truelayer.com/#renew-the-access_token">renew</a> them periodically and make sure to keep <strong>only</strong> the most recent pair in your database.</p><p>Here are some errors that might occur if old tokens are used:</p><ul><li><strong><em>invalid_grant</em></strong></li><li>Token expired</li><li>Internal error</li></ul><p><strong>Tokens that are badly managed might lead in end-users having to re-authenticate with the bank provider </strong>🏦<strong>.</strong></p><h4><strong>Do</strong></h4><p>✅ Replace the old <strong><em>access_token</em></strong> and <strong><em>refresh_token</em></strong> with the new ones after every refresh<br>✅ Renew tokens periodically or prior to a big batch of requests</p><h4><strong>Don’t</strong></h4><p>❌ Replace only one of the tokens after a refresh<br>❌ Keep old tokens in your database<br>❌ Forget to refresh tokens periodically</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xGXEd0j0DWUsebYp.jpg" /></figure><p>These are the best practices recommended by TrueLayer when it comes to managing your tokens. <strong>Have you got any further questions?</strong> Take a look at our FAQs <a href="http://truelayer.zendesk.com/">here</a> and contact our Client Care team; we’re always happy to help you get the best out of our APIs!</p><p><strong>Want to work with us?</strong> Say hi 👋, <a href="https://truelayer.workable.com/">we’re hiring!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=de22822d335d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>