Jaya Malhotra

<p>Learn to create a smart loan assistant that handles customer questions, captures leads, and manages complaints automatically</p>
<h2>About this guide</h2>
<p>This guide is based on a webinar led by <a href="https://www.linkedin.com/in/pratiksha-tewari-ba562119a/">Pratiksha</a>, our customer success manager at Tars with over three years of experience in the field. What stood out was her practical approach to building AI Agents. As she put it during the session, &#8220;I build, eat, and drink AI Agents for a living.&#8221; This hands-on experience shows in how she breaks down complex concepts into simple, actionable steps.</p>
<p>In this detailed walkthrough, you’ll learn how to build AI Agents without any coding knowledge. Pratiksha demonstrated a real loan assistant that handles customer questions, captures interested leads, and manages complaints. The best part? Anyone can build something similar using drag-and-drop tools.<br />
<iframe src="//www.youtube.com/embed/FIyIZkybFug" width="560" height="314" allowfullscreen="allowfullscreen"></iframe></p>
<h2>Why this matters for your business</h2>
<p>Most businesses face the same challenge: customers ask the same questions repeatedly. Whether it’s about loan interest rates, eligibility criteria, or application processes, teams spend valuable time answering identical queries. An AI Agent can handle these conversations automatically while you focus on more complex tasks.</p>
<p>But here’s where it gets interesting. A well-built AI Agent doesn’t just answer questions. It can detect when someone wants to apply for a loan and automatically save their information. It can handle complaints and send organized emails to your support team. This level of automation used to require a development team and months of work.</p>
<h2>What we built together</h2>
<p>The loan assistant created in this walkthrough will do three main things:</p>
<p><strong>Handle customer questions</strong>: When someone asks about interest rates or loan tenure, the Agent pulls answers from your knowledge base and responds immediately.</p>
<p><strong>Capture potential customers</strong>: If someone shows interest in applying for a loan, the Agent collects their contact information and saves it to a Google Sheet automatically.</p>
<p><strong>Process complaints</strong>: When customers have issues with their existing loans, the AI gathers details and sends a formatted email to your support team with a ticket number.</p>
<h2>Getting started: Your first Agent</h2>
<p>Start by signing up for a no-code AI platform. Most offer free trials that let you build and test your Agent without paying anything upfront. Once you log in, you’ll see templates for different industries, but this walkthrough builds from scratch to ensure you understand how everything works.</p>
<p>Click &#8220;<a href="https://www.hellotars.com/blog/how-to-build-ai-agents-from-scratch/">Create AI Agent from scratch</a>&#8221; and give it a name. In the webinar example, it was called &#8220;Loanie.&#8221; You’ll land on a blank canvas that looks intimidating at first, but it’s simpler than it appears.</p>
<h2>Understanding the workspace</h2>
<p>The platform organizes everything into three sections:</p>
<p><a href="https://www.hellotars.com/blog/best-ai-agents-for-customer-support-in-2025/"><strong> AI Agent Gambit</strong></a>: This is where you build your AI Agent’s workflow. Think of it as drawing a flowchart of how conversations should go.</p>
<p><a href="https://www.hellotars.com/blog/how-to-create-a-knowledge-base-for-ai-agents-a-complete-guide/"><strong>Knowledge</strong></a>: Your Agent needs information to answer questions accurately. This section stores all the documents, websites, or text that teach your AI Agent about your business.</p>
<p><a href="https://www.hellotars.com/blog/understanding-the-role-of-ai-agent-tools/"><strong>Tools</strong></a>: These are the actions your AI Agent can take, like updating spreadsheets or sending emails.</p>
<h2>Building the brain of your AI Agent</h2>
<p>Every AI Agent starts with a core <strong>AI Agent Gambit</strong> – the main brain that processes what customers say and decides how to respond. Here’s how to set it up:</p>
<h3>Creating the welcome message</h3>
<p>Your Agent needs to introduce itself. Keep it simple and friendly:</p>
<p>&#8220;Hi, I’m Loanie. Nice to meet you! How can I help you today?&#8221;</p>
<h3>Writing the AI Agent prompt</h3>
<p>This is where you tell your Agent how to behave. Think of it as giving instructions to a new employee. Be specific about what you want:</p>
<p>You are a friendly loan assistant. Keep your responses brief and break down large answers into small pieces. If users ask about loan tenure, interest rates, or general questions, always check the knowledge base first.</p>
<p>If you can’t find the answer in the provided information, say &#8220;I’m not sure about that&#8221; and don’t make up answers. Never discuss topics unrelated to loans and stay in character.</p>
<p>The key is being clear about boundaries. You don’t want your loan assistant giving cooking advice or discussing politics.</p>
<h2>Teaching your Agent about loans</h2>
<p>An Agent without knowledge is like hiring someone who knows nothing about your business. You need to feed it information about your loan products.</p>
<h3>Adding a knowledge base</h3>
<p>You have two options here. If your company has a website with loan information, you can simply add the URL and let the platform extract everything automatically. Otherwise, create a PDF with basic information like:</p>
<ul>
<li>Loan interest rates</li>
<li>Maximum loan amounts</li>
<li>Tenure options</li>
<li>Eligibility criteria</li>
<li>Required documents</li>
</ul>
<p>Upload this PDF to the knowledge section and let the platform train your Agent on it. This process usually takes a few minutes.</p>
<p><img decoding="async" class="alignnone size-large wp-image-9314" src="https://cdn.prod.website-files.com/6756905c4f1eabb644da6afe/6a0f1bf3188670864b2310e9_6a0eb1b89fd01acb9b7fe122_wp-Screenshot-2025-08-20-154710-1024x427.png" alt="Knowledge base in tars" width="1024" height="427" /></p>
<h3>Connecting knowledge to your Agent</h3>
<p>Once the training is complete, go back to your Agent and add the knowledge tool. This connects your Agent’s brain to the information it needs to answer questions accurately.</p>
<p>Remember to click &#8220;Publish Agent&#8221; after making changes. Then test it by asking a few questions about interest rates or loan terms. Your Agent should now respond with accurate information from your knowledge base.</p>
<h2>Capturing interested customers automatically</h2>
<p>Here’s where things get interesting. Your Agent can detect when someone wants to apply for a loan and capture their information without any manual work.</p>
<h3>Updating your Agent’s instructions</h3>
<p>Add this section to your Agent prompt:</p>
<p>If a user shows interest in applying for a loan, collect their name, phone number or email, and preferred time for us to call them. Save this information to our Google Sheet automatically.</p>
<h3>Setting up Google Sheets integration</h3>
<p>Create a new Google Sheet with columns for Name, Phone, Email, and Preferred Time. Copy the sheet ID from the URL (the long string of characters in the middle).</p>
<p>In the Tools section of your platform, find the Google Sheets integration and connect it. You’ll need to give permission for the platform to access your Google account. Once connected, configure it to use your specific sheet ID.</p>
<p><img decoding="async" class="alignnone size-large wp-image-9330" src="https://cdn.prod.website-files.com/6756905c4f1eabb644da6afe/6a0f1c0a6b9519181b541409_6a0eb1eb10c8861c76e7b6a2_wp-Screenshot-2025-08-20-154813-1024x435.png" alt="Tools for ai agents" width="1024" height="435" /></p>
<h3>Testing the lead capture</h3>
<p>Now, when someone says &#8220;I want to apply for a loan,&#8221; your Agent will ask for their details and automatically save everything to your spreadsheet. You can watch this happen in real-time, which feels like magic the first time you see it.</p>
<h2>Handling customer complaints</h2>
<p>The third function of your Agent is processing complaints from existing customers. This requires a bit more setup but creates a professional support system.</p>
<h3>Adding complaint handling to your prompt</h3>
<p>Extend your Agent’s instructions with:</p>
<p>For existing customers with complaints, ask for their email address so we can create a support ticket. After collecting their email and complaint details, send a formatted email to our support team with all the information organized clearly.</p>
<h3>Setting up automated emails</h3>
<p>Connect the Gmail tool from the Tools section. This lets your Agent send emails from your business account. You’ll need to give permission for the platform to send emails on your behalf.</p>
<p>Configure the email template in your prompt itself.</p>
<h3>The complaint workflow in action</h3>
<p>When a customer says &#8220;I have a complaint,&#8221; your Agent will:</p>
<ol>
<li>Ask for their email address</li>
<li>Request details about their complaint</li>
<li>Categorize the issue (EMI problems, documentation issues, etc.)</li>
<li>Generate a unique ticket ID</li>
<li>Send a formatted email to your support team</li>
<li>Confirm with the customer that their complaint was logged</li>
</ol>
<p>This creates a professional support process that ensures no complaint gets lost or forgotten.</p>
<h2>Making your Agent faster and smarter</h2>
<p>Speed matters in customer service. Here are ways to optimize your Agent’s performance:</p>
<h3>Keep prompts focused</h3>
<p>Long, complicated instructions confuse AI models. Break complex tasks into smaller, specific directions. Instead of one giant paragraph, use numbered steps or bullet points.</p>
<h3>Choose the right AI model</h3>
<p>Different AI models have different strengths. GPT-4 is great for complex reasoning but might be slower. Claude excels at following instructions precisely. Test different models to find what works best for your use case.</p>
<h3>Structure your knowledge base well</h3>
<p>Organize your PDF or website content logically. Use clear headings and avoid duplicate information. The cleaner your knowledge base, the better your Agent’s responses.</p>
<h2>Adding voice capabilities</h2>
<p>Modern users expect multiple ways to interact with AI. Your Agent can handle voice conversations too.</p>
<p>The platform automatically includes speech-to-text functionality. Customers can click a microphone button and speak their questions instead of typing. The Agent converts speech to text, processes the question, and can even read the response aloud.</p>
<p>This is particularly helpful for loan applications where customers might be multitasking or prefer speaking over typing.</p>
<h2>Monitoring and improving your Agent</h2>
<p>Building the Agent is just the beginning. You need to track how it’s performing and make improvements based on real conversations.</p>
<h3>Reviewing conversation logs</h3>
<p>The platform stores every conversation your Agent has. Review these regularly to identify:</p>
<ul>
<li>Questions your Agent couldn’t answer (gaps in knowledge base)</li>
<li>Common user frustrations or confusion points</li>
<li>Successful lead captures and their patterns</li>
<li>Complaint types and resolution times</li>
</ul>
<h3>Tracking key metrics</h3>
<p>Monitor these numbers to measure success:</p>
<ul>
<li><strong>Response accuracy</strong>: How often does your Agent give correct answers?</li>
<li><strong>Lead capture rate</strong>: What percentage of interested visitors provide their information?</li>
<li><strong>Customer satisfaction</strong>: Include rating requests at the end of conversations</li>
<li><strong>Time saved</strong>: Estimate how many hours of human work your Agent handles</li>
</ul>
<h3>Making iterative improvements</h3>
<p>Based on your analysis, update your knowledge base with missing information, refine your prompts for clearer instructions, and add new capabilities as you identify needs.</p>
<h2>Advanced features and integrations</h2>
<p>Once your basic Agent is working well, consider these enhancements:</p>
<h3>Multi-language support</h3>
<p>AI models understand multiple languages automatically. If you serve customers who speak different languages, your Agent will detect their language and respond appropriately. You might need to translate your knowledge base content for the best results.</p>
<h3>CRM integration</h3>
<p>Instead of Google Sheets, connect directly to your CRM system like HubSpot, Salesforce, or Bitrix. This keeps all customer information in one place and triggers your existing follow-up processes.</p>
<h3>Complex decision trees</h3>
<p>Use the platform’s logic features to create different conversation paths based on customer type, loan amount, or other criteria. For example, business loan inquiries might follow different processes than personal loans.</p>
<h2>Common challenges and solutions</h2>
<p>Every AI Agent builder faces similar issues. Here’s how to handle them:</p>
<h3>AI Agent gives wrong answers</h3>
<p>This usually means your knowledge base needs work. Add more comprehensive information, remove contradictory content, and be more specific in your prompts about when to say &#8220;I don’t know.&#8221;</p>
<h3>Slow response times</h3>
<p>Review your Agent’s complexity. Too many tools or overly complicated prompts slow things down. Simplify where possible.</p>
<h3>Integration problems</h3>
<p>Double-check permissions for Google Sheets, Gmail, or other connected services. Make sure your sheet IDs and email addresses are correct. Test integrations individually before combining them.</p>
<h3>Customers get confused</h3>
<p>Review your conversation logs to see where confusion happens. Often, the Agent’s language is too technical or the conversation flow isn’t intuitive. Adjust your prompts to be more conversational and clear.</p>
<h2>Scaling your AI Agent</h2>
<p>As your Agent proves its value, you’ll want to expand its capabilities:</p>
<h3>Multiple deployment channels</h3>
<p>Beyond your website, deploy your Agent on social media platforms, mobile apps, or as a WhatsApp business bot. Most platforms support multiple deployment options.</p>
<h3>Team collaboration</h3>
<p>Larger businesses might need multiple people managing the Agent. Set up proper access controls and documentation so team members can make updates safely.</p>
<h3>Performance monitoring</h3>
<p>Implement more sophisticated analytics as your Agent handles more conversations. Track conversion rates, customer journey completion, and ROI measurements.</p>
<h2>The business impact</h2>
<p>Companies using AI Agents report significant improvements in customer service efficiency. Instead of hiring additional support staff as you grow, your Agent scales automatically. Customer questions get answered instantly, even outside business hours.</p>
<p>The lead capture functionality is particularly valuable. Instead of hoping interested visitors fill out contact forms, your Agent engages them in conversation and gathers information naturally. This typically improves conversion rates compared to static forms.</p>
<p>For complaint handling, the automated email system ensures consistency and prevents issues from being overlooked. Support teams receive organized information instead of scattered customer messages.</p>
<h2>Getting started today</h2>
<p>The technology for building sophisticated AI Agents is available now, and the learning curve is manageable for most business professionals. Start with a simple use case like FAQ handling, then gradually add more complex features as you become comfortable with the platform.</p>
<p>Most platforms offer free trials, so you can experiment without financial commitment. The time investment is minimal compared to traditional software development, and you can see results within hours rather than months.</p>
<p>The loan assistant example from Pratiksha’s webinar demonstrates what’s possible, but the same principles apply to any industry. Whether you’re in insurance, real estate, healthcare, or retail, you can build an Agent that handles your specific customer interactions.</p>
<p>The question isn’t whether AI Agents will become standard business tools – they already are. The question is whether you’ll build yours now or wait for competitors to gain the advantage.</p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

How to build an AI Agent without coding: Decoding the AI Agent Builder Canvas

Learn how to build a customer support AI Agent with no code. This guide shows you how to capture leads and manage support tickets.

How to build an AI Agent without coding: Tars AI Agent Builder

<p>Meet Scott from Devonport City Council in Australia, who just walked us through his incredible journey of creating Mersey, a Travel AI Agent.</p>
<h2>Meet the builder behind the lighthouse</h2>
<p><iframe src="//www.youtube.com/embed/jqaPkLmTKMA" width="560" height="314" allowfullscreen="allowfullscreen"></iframe><br />
Scott is an enterprise solutions architect who&#8217;s worked at Devonport City Council for almost seven years, with dual degrees in criminology, criminal justice, and business management.</p>
<p>He built Mersey to help users landing on the website, and this blog is about his learnings and journey.</p>
<h2>Version 1: The struggle was real</h2>
<p>Scott started with what seemed like a simple setup. Mersey version 1 used Tars&#8217; web knowledge base tool. The flow was straightforward: the user greets Mersey, asks a question, and Mersey pulls information from the knowledge base. Simple, right?</p>
<p>&#8220;The web crawler had to be manually activated and it was a bit I found it a bit tiresome to keep activating when we added new content,&#8221; Scott revealed.</p>
<p>Imagine having to manually retrain your AI Agent every time someone added a new business listing. &#8220;By introducing a self sort of help feature for businesses and events, it just wasn&#8217;t feasible to go in every time a new listing happened and click a button to sort of retrain it.&#8221;</p>
<p>But the real problems ran deeper:</p>
<p><strong>The data wasn&#8217;t live and was completely mixed up.</strong> &#8220;If I were to ask a question about an event, and let&#8217;s say I&#8217;ve got an event page and then I&#8217;ve also got an article that mentions an event. It started getting confused between the two because there was no clear separation of the data. it was just sort of all thrown together, and that just added to sort of like a poor user experience.&#8221;</p>
<h2>Version 2: The rebuild that changed everything</h2>
<p>Instead of patching up the mess, Scott did something bold – he rebuilt most of the structure of the AI Agent. This time, he separated everything into clean databases: places, events, travel guides, and reviews each got their own home.</p>
<p>Here&#8217;s where it gets clever. When someone asks Mersey, &#8220;Is Coles open?&#8221; (that&#8217;s a supermarket chain for non-Aussies), here&#8217;s what happens:</p>
<ol>
<li>Mersey figures out this is about a place, not an event</li>
<li>It calls a custom place tool</li>
<li>That tool checks the current date and time</li>
<li>It searches the places database for Coles</li>
<li>It comes back with &#8220;Is Cole currently open,&#8221; plus contact details and opening hours</li>
</ol>
<p>If Coles were closed, Mersey would even tell you when it opens next. That&#8217;s the kind of practical help tourists actually need.</p>
<p>This rebuilding involved the creation of custom tools for Mersey that improved the functioning of the AI Agent.</p>
<h3>The 100-hour reality check</h3>
<p>This wasn&#8217;t a weekend project. &#8220;It was actually about over 100 hours of work gone into rebuilding the entire system, building the databases, building the website and platform. So it is a big time investment that you sort of need to big time investment if you&#8217;re going to go through that.&#8221;</p>
<h2>The data structure revolution</h2>
<p>The key insight that changed everything? Data quality.</p>
<p>&#8220;If you don&#8217;t get your databases correct and if you don&#8217;t have your data structured, I mean, it&#8217;s bad data, and if you&#8217;re putting bad data into an AI Agent, the output&#8217;s just going to be bad data. So, you really need to get your data sorted.&#8221;</p>
<p>Scott contrasted this with his previous setup: &#8220;When I was on the previous website, the data structure was terrible. It was a drag-and-drop page builder, so all the data was stored in the same sort of column in a table.&#8221;</p>
<p>The new system processes data before sending it to the AI Agent: &#8220;The data is pre-processed before being provided to tar. So I know that the data going to t is actually good quality data, which I couldn&#8217;t before.&#8221;</p>
<h3>Self-service business portal</h3>
<p>Scott built something clever – a business portal where companies can manage their own listings. &#8220;I&#8217;ve actually got a business portal. And this is a self-help business portal for businesses to actually submit their own events and sort of submit their own listings, and then they can edit listings as well.&#8221;</p>
<p>But he learned something about human nature: &#8220;I know from experience and especially even like big tourism data warehouses like we got in Australia, businesses just don&#8217;t want to update the data. They sign up and then they forget about it.&#8221;</p>
<h2>The prompt engineering journey</h2>
<p>Scott emphasized that building the AI Agent was an ongoing process: &#8220;Build a prompt with the knowledge that you&#8217;ll regularly need to fine-tune. I think I told you this before, six months ago, when I said that I regularly read the chat logs and I&#8217;m regularly changing that prompt so I can give it more sort of like detailed information.&#8221;</p>
<h3>Real-world prompt challenges</h3>
<p>One perfect example: &#8220;I had an issue with you, where do I where can I park with my caravan, and it would be goin,g telling them where to go to a caravan park versus an actual parking spot where they could park their caravan in the CBD. So, I had to provide more prompting.&#8221;</p>
<p>These kinds of misunderstandings only surface through real user interactions, which is why Scott stressed the importance of monitoring chat logs and continuously refining the prompts.</p>
<h2>The personality factor that everyone forgets</h2>
<p>He didn&#8217;t just throw together a random image: &#8220;I hired a designer to do my icon for my AI Agents. I had him design a lighthouse icon. I wanted to give it a bit more personality.&#8221;</p>
<p>The result? &#8220;I think a lot of people have sort of really liked that visit my site, that it&#8217;s a little lighthouse with a face.&#8221;</p>
<h2>Measuring success (the manual way)</h2>
<p>When asked about metrics and evaluation, Scott was refreshingly honest about his manual approach: &#8220;When it comes to actually working out how successful like version one all the it&#8217;s just as you read the chat logs but if I read something and I can&#8217;t I&#8217;m not getting the answer which I think the user wants and the questions are usually pretty direct even though there are some really weird ones that come through like that I don&#8217;t consider that successful.&#8221;</p>
<p>He also noted a gap in the platform: &#8220;There isn&#8217;t really a I don&#8217;t believe there&#8217;s way to view up votes and down votes for responses, which would be helpful.&#8221;</p>
<p>But Scott had ideas for scaling this: &#8220;I download hundreds of chat logs, and I use AI because I don&#8217;t want to read through all of them. I use I actually score the AI Agent in its accuracy and a certain number of conditions that I knew. So I think there is room to use AI to actually give a sort of generalized success score.&#8221;</p>
<h2>The custom tools experience</h2>
<p>Scott&#8217;s experience with <a href="https://www.hellotars.com/blog/how-can-you-add-tooling-capabilities-to-ai-agents/">Tars&#8217; custom tools feature</a> was smooth: &#8220;Once I had the once had the IP pass up in the actual schema generated in the format you needed. I mean, it was pretty easy just to paste it in, and yeah, and just connect it up, and it worked.&#8221;</p>
<p>For generating the API schema, Scott used a practical approach: &#8220;To get the schema, I used a bit of AI to work that out, and then I hadn&#8217;t used that schema format before. So I just for my API documentation, I made it into AI to actually get it to generate it.&#8221;</p>
<h2>Lessons from the trenches</h2>
<p>Scott&#8217;s key takeaways:</p>
<ul>
<li>Start simple, then build up complexity</li>
<li>Clean, structured data beats fancy features</li>
<li>Regular <a href="https://www.hellotars.com/blog/how-to-master-prompting-and-build-super-cool-ai-agents/">prompt</a> fine-tuning is essential</li>
<li>Give your Agent personality</li>
<li>APIs work better than web scraping for complex data</li>
</ul>
<h2>The results</h2>
<p>Today, Mersey provides actual, useful information. Ask about opening hours, and get the current status. Ask about weekend events, and get events actually happening this weekend. It is helping tourists navigate Australia&#8217;s largest city tourism site (eat your heart out, Sydney).</p>
<p>Scott&#8217;s journey shows that building truly useful AI Agents isn&#8217;t about having the fanciest AI – it&#8217;s about having the right data structure, clear purpose, and attention to user experience. Plus, a lighthouse with a face doesn&#8217;t hurt either.</p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

Learn how Scott Newman built Mersey, the Travel AI Agent that never sleeps

This blogs details the journey of how Scott Newman built Mersey, the Travel AI Agent that helps users 24/7.

Learn how Scott Newman built Mersey, the Travel AI Agent

<div class="last-updated" style="background: #f0fdf4; border-left: 3px solid #16a34a; padding: 8px 12px; margin-bottom: 20px; font-size: 14px; color: #166534;">&#10003; <strong>Last Updated:</strong> August 22, 2025</div><h2>The problem we solved (and you probably have it too)</h2>
<p>Your team gets hit with the same questions every single day. &#8220;Where are my pay slips?&#8221; &#8220;How many vacation days do I have?&#8221; &#8220;What&#8217;s our remote work policy?&#8221; It&#8217;s like Groundhog Day, but with more spreadsheets and less Bill Murray.</p>
<blockquote><p>&#8220;There&#8217;s a lot of the same questions being asked and a lot of the same answers being given. And that&#8217;s why this Agent becomes handy.&#8221;</p></blockquote>
<p>We built OrgAnswers AI, an Agent that handles these repetitive questions so your HR team can focus on actual human problems instead of playing broken record.</p>
<p><iframe src="//www.youtube.com/embed/suNkWEfJ3Ww" width="560" height="314" allowfullscreen="allowfullscreen"></iframe></p>
<h2>What actually happened in our live session</h2>
<p><strong>The bold promise:</strong> Build a working AI Agent in 40 minutes with zero coding knowledge. <strong>The audience:</strong> People who&#8217;d never built an Agent before. <strong>The result:</strong> A functional employee Q&amp;A system that actually works.</p>
<p>No, seriously. We went from a blank screen to a working Agent while people watched. Here&#8217;s how it went down.</p>
<h2>The anatomy of an AI Agent</h2>
<p>Think of an AI Agent like hiring someone really smart but incredibly literal:</p>
<ul>
<li><strong>Trigger:</strong> The thing that wakes them up (usually someone asking a question)</li>
<li><strong>Brain:</strong> Where all the thinking happens</li>
<li><strong>Knowledge:</strong> Everything they know about your company</li>
<li><strong>Tools:</strong> The stuff they can actually do (like sending emails)</li>
</ul>
<blockquote><p>&#8220;You can think of like different use cases and make sure that the AI knows how to handle each one of those.&#8221;</p></blockquote>
<p>The trick is teaching them well enough that they don&#8217;t embarrass you at the company party.</p>
<h2>Our live build breakdown (the good, the messy, and the &#8220;why won&#8217;t this work&#8221;)</h2>
<h3>Step 1: Setting up the foundation</h3>
<p>We started with the AI Agent builder and created our Agent&#8217;s personality. Without a personality, an AI Agent is purposeless.</p>
<p>The welcome message we crafted was simple:</p>
<ul>
<li>&#8220;Hi, I&#8217;m OrgAnswers, your AI assistant for employee questions&#8221;</li>
<li>&#8220;How can I help you today?&#8221;</li>
</ul>
<p>Nothing fancy, but it works.</p>
<h3>Step 2: Feeding it the right knowledge</h3>
<p>This is where things get interesting. We uploaded different company documents:</p>
<ul>
<li>Leave policies</li>
<li>Remote work guidelines</li>
<li>Employee benefits</li>
<li>Payroll procedures</li>
</ul>
<blockquote><p>&#8220;The key of writing really good prompts so the AI works well is to really structure your prompt well and also keep it very specific.&#8221;</p></blockquote>
<p>We kept each document separate instead of dumping everything into one giant file. Why? Because even AI Agents need good organizational skills.</p>
<h3>Step 3: The moment of truth &#8211; testing our creation</h3>
<p><strong>First test:</strong> &#8220;How many leaves can I take?&#8221;</p>
<p>The Agent pulled information straight from our uploaded policy document. It worked, but the answer was basically a copy-paste of the entire document. Not great.</p>
<p><strong>The fix:</strong> We added specific instructions about how to format answers. More examples, better structure, shorter responses.</p>
<blockquote><p>&#8220;The more examples you add, the better your AI will perform. It&#8217;s able to learn more from examples than it is if you just directly give it instructions.&#8221;</p></blockquote>
<h3>Step 4: Adding the escalation safety net</h3>
<p>Here&#8217;s where we got fancy. We connected Gmail, so when the Agent hits a wall, it doesn&#8217;t just shrug and give up.</p>
<p>The workflow:</p>
<ol>
<li>Agent realizes human intervention is required</li>
<li>Asks the user if they want to escalate to HR</li>
<li>Collects details (name, email, problem description)</li>
<li>Sends a proper email to HR automatically</li>
<li>Tells the user help is on the way</li>
</ol>
<h2>The questions people actually asked</h2>
<p><strong>&#8220;Are there ready-to-use templates?&#8221;</strong> Yes! Tars has 10 templates you can customize instead of starting from scratch. We just like the dramatic flair of building from nothing.</p>
<p><strong>&#8220;Can I use website content as knowledge?&#8221;</strong> Absolutely. Just paste the URL and it&#8217;ll crawl everything.</p>
<p><strong>&#8220;What about email conversations as knowledge?&#8221;</strong> Turn them into PDFs first, then upload.</p>
<p><strong>&#8220;Does my data stay secure?&#8221; </strong>All data you upload as knowledge sources is only accessible by you, and the information stays confidential.</p>
<h2>What you should do next (if you&#8217;re actually going to build this)</h2>
<h3>Before you start:</h3>
<ul>
<li>Round up your most common employee questions</li>
<li>Gather your company documents (employee handbook, policies, etc.)</li>
<li>Pick someone to be the escalation contact</li>
</ul>
<h3>During the build:</h3>
<ul>
<li>Test constantly with real questions your people ask</li>
<li>Start simple, get fancy later</li>
<li>Save your work frequently (learned this the hard way)</li>
</ul>
<h3>After you launch:</h3>
<ul>
<li>Check the conversation logs to see what people are actually asking</li>
<li>Update knowledge bases based on gaps you find</li>
<li>Don&#8217;t expect perfection &#8211; plan to iterate</li>
</ul>
<h2>The bigger picture</h2>
<p>We&#8217;re not trying to replace humans here. We&#8217;re trying to stop humans from answering &#8220;Where do I find my W2?&#8221; for the 47th time this month.</p>
<p>Instead of always seeking out a human for something that can be answered, people don&#8217;t really go through that employee handbook with all the answers. Now they can just directly ask that question to the Agent, without shuffling through documents.</p>
<p>Your HR team has bigger problems to solve than being a human search engine.</p>
<h2>Why this matters right now</h2>
<p>Every company is dealing with the same repetitive question fatigue. The tools to fix it are here, they&#8217;re accessible, and they don&#8217;t require a computer science degree.</p>
<p><strong>The best part?</strong> The Agent we built is already handling questions better than most company intranets. And it took less time than your last team meeting.</p>
<h2>Ready to stop being the office search engine?</h2>
<p>All you need to do is sign in, and you can choose to build from scratch, with knowledge, tools, or there are templates if you want to skip the dramatic live-building experience.</p>
<p>Just jump in and start building. The worst that happens is you learn something new about how AI actually works.</p>
<p><strong>Got questions about your specific use case?</strong> The Tars <a href="http://hellotars.com/discord">community</a> is pretty helpful, and honestly, half the fun is figuring it out as you go.</p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

[Webinar] OrgAnswers AI: Building your first employee Q&A AI Agent

Read the blog and learn how you can build OrgAnswers AI Agent that answers all the repetitive questions asked by your employees.

OrgAnswers AI: Learn how to build your employee Q&A AI Agent

Bani

<div class="last-updated" style="background: #f0fdf4; border-left: 3px solid #16a34a; padding: 8px 12px; margin-bottom: 20px; font-size: 14px; color: #166534;">&#10003; <strong>Last Updated:</strong> August 29, 2024</div><p id="ember2224" class="ember-view reader-text-block__paragraph">Vector databases have become a hot topic in the tech world. The appeal lies in their ability to quickly find similar information in unstructured data. However, this hype overlooks its limitations in understanding complex data and connecting the dots in the given information.</p>
<p>Once known as the only solution to overcome AI challenges, <a href="https://www.linkedin.com/pulse/all-vector-databases-must-die-teaching-ai-speak-less-groot-more-rocket-seiac/?trackingId=i4j14bFsOw4ljBkWBBrtLw%3D%3D">vector databases are falling behind</a> in the race for accuracy and reliability.</p>
<p class="ember-view reader-text-block__paragraph">We set out on a journey to find a better solution, and the results are thought-provoking.</p>
<blockquote><p>&#8220;Which architecture makes more intelligent AI agents: Graph RAG or Vector RAG?&#8221;</p></blockquote>
<p>How does Vector RAG fare against Graph RAG for AI accuracy and reliability? The Tars and <a href="https://www.cogniswitch.ai/">CogniSwitch Inc. Team</a> practitioners <strong>pitted two <a href="https://hellotars.com/blog/webinar-recording-the-tale-of-two-non-techies-and-their-ai-agents" target="_blank" rel="noopener">AI Agents</a> against each other for complex customer support scenarios in a Show-and-Tell session.</strong></p>
<p>During the webinar, we <strong>demonstrated how graph-based AI better modeled complex relationships and provided better responses than traditional vector-based approaches.</strong></p>
<p>We also uncovered some often-asked questions around Graph RAG in real world scenarios:</p>
<ul style="list-style-type: circle;">
<li>How can Graph RAG provide more accurate and contextually relevant responses?</li>
<li>How does it address generative AI challenges like hallucinations?</li>
<li>Is purely Graph RAG enough to create the best-in-class AI Agents for customer support?</li>
</ul>
<h3>Key takeaways of the webinar include:</h3>
<ul>
<li><strong>Understanding Limitations:</strong> Learn about the limitations of Vector RAG, including context loss, struggles with relational information, and lack of reasoning capabilities.</li>
<li><strong>Graph RAG Advantages:</strong> The webinar showcases how Graph RAG improved response accuracy by modeling complex relationships and providing contextually rich answers.</li>
<li><strong>Real-World Applications:</strong> The team demonstrated practical examples of Graph RAG outperforms Vector RAG in customer support scenarios, showcasing its reliability and efficiency.</li>
</ul>
<p>Want to see this RAG showdown for yourself? Check it out <a href="https://www.youtube.com/live/gOXvR2b7Og8?feature=shared">here</a> and view our slides on <a href="https://www.slideshare.net/slideshow/why-should-gen-ai-adopters-switch-to-graph-based-ai-agents_-pptx/270839020">Slideshare</a>.</p>
<p><iframe style="height: 430px;" title="YouTube video player" src="//www.youtube.com/embed/gOXvR2b7Og8?si=MGi8ARmrF0sVy1kT" width="560" height="315" frameborder="0" allowfullscreen="allowfullscreen"></iframe></p>
<p>&nbsp;</p>
<p>Check out <a href="https://www.cogniswitch.ai/">CogniSwitch.ai</a> today! To stay up-to-date on this conversation, join our <a href="https://discord.gg/ZVWD6HVv">discord channel.</a></p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

[Webinar] Why Should Gen AI Adopters Switch to Graph RAG Based AI Agents?

Which architecture is better: Graph RAG or Vector RAG? In this webinar, we compare the performance of Graph vs Vector RAG for complex customer support scenarios.

[Webinar] Why Should Gen AI Adopters Switch to Graph RAG Based AI Agents? - Tars Blog

Arnav Patel

<div class="last-updated" style="background: #f0fdf4; border-left: 3px solid #16a34a; padding: 8px 12px; margin-bottom: 20px; font-size: 14px; color: #166534;">&#10003; <strong>Last Updated:</strong> March 18, 2020</div>
<p>Chatbots and push notifications go together like peanut butter and jelly. Both technologies are pretty great on their own, but together they are downright phenomenal. Think about it like this. Lead generation consists of two stages:<br></p>



<ol><li> Pre-click optimization</li><li>Post-click optimization.  </li></ol>



<p>On the pre-click side, you need to worry about getting your landing pages in front of the right customers. If implemented correctly, push notifications are the best way of achieving this. Unlike Google ads, Social Media Marketing or SEO which give you commoditized, low-attention &#8216;traffic&#8217; a well-executed push notification campaign allows you to re-engage visitors and to<a href="https://www.izooto.com/"> own an ‘audience’</a> who is more willing to give you their attention and click on to your landing pages. <br></p>



<p>Of course, this is only half of the picture. Getting your landing page in front of a focused audience does not guarantee conversions. Modern buyers are discerning with their attention and they will punish you with a high drop rate if you show them a poorly designed landing page. This is where post-click optimization comes into play. As a marketer, it&#8217;s on you to deliver an engaging experience to your prospect that puts them in their mind space to convert. For a lot of industries (mainly B2C Services), <a href="https://www.hellotars.com/blog/ppc-conversion-rate-conversational-landing-pages/">Chatbots are the best way to achieve this</a>. Conversational Landing Pages give your audience a personalized lead generation experience that converts at rates that are <a href="https://www.hellotars.com/blog/chatbots-conversion-rate-increase/">2-3x higher than traditional landing pages.</a><br><br>We were so excited by the potential of a chatbot+ push notification combination that we teamed up with the push notification wizards at <a href="http://izooto.com">iZooto</a> to create the ultimate webinar on how you can use chatbots and push notifications to supercharge your conversions and take your marketing to the next level. </p>



<script src="https://fast.wistia.com/embed/medias/f39gzrs2lj.jsonp" async=""></script><script src="https://fast.wistia.com/assets/external/E-v1.js" async=""></script><div class="wistia_responsive_padding" style="padding:51.25% 0 0 0;position:relative;"><div class="wistia_responsive_wrapper" style="height:100%;left:0;position:absolute;top:0;width:100%;"><div class="wistia_embed wistia_async_f39gzrs2lj videoFoam=true" style="height:100%;position:relative;width:100%"><div class="wistia_swatch" style="height:100%;left:0;opacity:0;overflow:hidden;position:absolute;top:0;transition:opacity 200ms;width:100%;"><img decoding="async" src="https://fast.wistia.com/embed/medias/f39gzrs2lj/swatch" style="filter:blur(5px);height:100%;object-fit:contain;width:100%;" alt="" onload="this.parentNode.style.opacity=1;"></div></div></div></div>



<p></p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

[Webinar] How Chatbots And Push Notifications Can Supercharge Your Website Conversions

[Webinar] How Chatbots And Push Notifications Can Supercharge Your Website Conversions - Tars Blog

<div class="last-updated" style="background: #f0fdf4; border-left: 3px solid #16a34a; padding: 8px 12px; margin-bottom: 20px; font-size: 14px; color: #166534;">&#10003; <strong>Last Updated:</strong> November 19, 2021</div>
<p class="has-text-align-left">Conversion rates throughout the digital marketing world are at an all-time low. As attention spans become shorter and prospects get desensitized from conventional marketing channels, marketers face an uphill battle while generating leads through digital channels. To get around this problem we propose a simple, but radical solution. </p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow"><p><strong>Make your landing pages more conversational.</strong></p></blockquote>



<p class="has-text-align-left"><a href="http://hellotars.com"> Conversational Landing Pages</a> provide an experience that feels more human than a traditional landing page and consequently allow you to capture a greater share of your prospects&#8217; attention. Put simply,  if you are having trouble converting your traffic into high-quality leads, conversational landing pages could be the customer service and CRO tool that <a href="https://www.hellotars.com/blog/ppc-conversion-rate-conversational-landing-pages/">increases your conversion rate by 2x.</a> </p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow"><p><strong>But there is a catch! And it is a pretty big catch as far as catches go!</strong></p></blockquote>



<p class="has-text-align-left">Like with all new marketing technologies, <strong>the effectiveness of a conversational landing page is wholly contingent on how well you implement it. </strong>This poses marketers with a unique problem. </p>



<p>Conversational landing pages can help you increase your conversion rate, but designing them requires a highly specialized <g class="gr_ gr_5 gr-alert gr_spell gr_inline_cards gr_disable_anim_appear ContextualSpelling ins-del" id="5" data-gr-id="5">skillset</g> that most marketers don&#8217;t possess.</p>



<p class="has-text-align-left">To help you get around this problem <a href="https://vwo.com/resources/webinars/how-to-make-post-click-experience-engaging-conversational-landing-pages/">we conducted a webinar with the folks over at </a><g class="gr_ gr_3 gr-alert gr_spell gr_inline_cards gr_run_anim ContextualSpelling ins-del multiReplace" id="3" data-gr-id="3"><a href="https://vwo.com/resources/webinars/how-to-make-post-click-experience-engaging-conversational-landing-pages/">VWO</a></g> to get you caught up with everything you need to know about conversational landing pages before you get started with them. </p>



<p class="has-text-align-left"><strong>Check it out below:</strong></p>



<div style="position: relative;
padding-bottom: 56.25%;
height: 0;
overflow: hidden;
max-width: 100%;">
<iframe style="position: absolute;top: 0;left: 0;width: 100%;height: 100%;min-height:auto"
      src="https://www.youtube.com/embed/S7-cN2jNLu8"
      title="YouTube video player"
      frameborder="0"
      allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
      allowfullscreen
    ></iframe></div>



<p></p>



<p></p>

<div style="font-size: 0px; height: 0px; line-height: 0px; margin: 0; padding: 0; clear: both;"></div>

Webinar

Top stories

Webinar Blogs

How to build an AI Agent without coding: Decoding the AI Agent Builder Canvas

Learn how Scott Newman built Mersey, the Travel AI Agent that never sleeps

[Webinar] OrgAnswers AI: Building your first employee Q&A AI Agent

[Webinar] Why Should Gen AI Adopters Switch to Graph RAG Based AI Agents?

[Webinar] How Chatbots And Push Notifications Can Supercharge Your Website Conversions

[Webinar] How To Make Post-Click Experience Engaging With Conversational Landing Pages

Our journey in a few numbers

Still scrolling? We both know you're interested.

GET IN TOUCH