1 00:00:02,160 --> 00:00:02,720 Hi guys. 2 00:00:02,720 --> 00:00:03,760 Welcome to this session. 3 00:00:03,760 --> 00:00:10,480 So in this session we'll talk about the retrieval augmented generation rag key to custom business chatbot. 4 00:00:10,480 --> 00:00:13,480 So these are going to be chatbots which are empowered with AI. 5 00:00:13,800 --> 00:00:21,200 So usually when we are using a basic chatbot to communicate with our customers, a lot of times the 6 00:00:21,200 --> 00:00:24,000 answers which we are providing might not be consistent. 7 00:00:24,240 --> 00:00:30,240 And there's a lot of manual time and effort involved in responding, communicating with our customers. 8 00:00:30,240 --> 00:00:36,280 So in such a case, you can replace it with with an AI powered chatbot, which can help you in this 9 00:00:36,320 --> 00:00:38,000 such in this particular scenario. 10 00:00:38,200 --> 00:00:41,640 So how it works is the Rag specifically retrieves data. 11 00:00:41,640 --> 00:00:48,640 So what we do here is we upload the AI with completely all our information, specifically uh, resource 12 00:00:48,640 --> 00:00:51,160 materials, files, documents. 13 00:00:51,320 --> 00:00:54,120 Everything about our business is uploaded over here. 14 00:00:54,160 --> 00:00:56,240 It's like a chat, like custom GPT. 15 00:00:56,520 --> 00:01:03,280 And now the AI responds to our customers by going through all this information, which we have uploaded. 16 00:01:03,840 --> 00:01:07,360 So ideally it it will going to look at. 17 00:01:07,400 --> 00:01:10,610 So we are going to generate build a knowledge base. 18 00:01:10,730 --> 00:01:18,570 So what it does is it goes ahead and converts the FAQs, manuals, documents into a searchable database. 19 00:01:18,570 --> 00:01:20,850 So your whole documents which you have provided. 20 00:01:20,930 --> 00:01:24,290 Becomes like a database which is searchable in nature. 21 00:01:24,290 --> 00:01:29,010 And now the AI can scan through it to find out the right answer for the customer. 22 00:01:29,330 --> 00:01:32,410 It allows chatbots to access the relevant information. 23 00:01:32,530 --> 00:01:38,770 Also, it will matches the user's questions with the best content which is available in the database. 24 00:01:39,090 --> 00:01:45,170 It is going to also look at blending retrieved information with AI's conversational skills for accurate 25 00:01:45,170 --> 00:01:46,010 responses. 26 00:01:46,410 --> 00:01:49,010 So how does it matter is because it. 27 00:01:49,050 --> 00:01:55,490 This specifically ensures accuracy because now the answers it reduces AI hallucinations. 28 00:01:55,530 --> 00:01:58,130 The answers are not inconsistent anymore. 29 00:01:58,570 --> 00:01:59,530 These are delivered. 30 00:01:59,730 --> 00:02:01,130 Deliveries delivers. 31 00:02:01,130 --> 00:02:04,610 Consistent fact based answers are given to our customers. 32 00:02:04,970 --> 00:02:07,450 Now you can also scale this further. 33 00:02:07,490 --> 00:02:09,210 It can handle thousands of pages. 34 00:02:09,210 --> 00:02:15,930 So if the if you upload huge amount of files documents, the AI is capable of scanning all those and 35 00:02:15,930 --> 00:02:18,770 give very effective answers to our customers. 36 00:02:19,130 --> 00:02:26,570 It can also optimize the chatbot design if data belongs, belongs in the knowledge base or in instructions, 37 00:02:26,570 --> 00:02:31,050 and it empowers you to troubleshoot much more in an effective manner. 38 00:02:31,250 --> 00:02:34,050 So let's see this in action how this would be happening. 39 00:02:34,210 --> 00:02:40,690 So we'll make use of let's say the previous custom GPT which we had created for Google Ads optimization. 40 00:02:40,770 --> 00:02:48,810 So if you see currently in this if we edit this GPT, this does not have any particular file which we 41 00:02:48,850 --> 00:02:49,770 have provided. 42 00:02:49,770 --> 00:02:57,530 So now currently what we are going to do is we are going to give it a prompt of how I can improve my 43 00:02:57,530 --> 00:02:59,210 quality score in Google Ads. 44 00:02:59,330 --> 00:02:59,730 Okay. 45 00:02:59,770 --> 00:03:06,130 And with this, without any information, any files, any documents given, we'll see how the chatbot 46 00:03:06,170 --> 00:03:06,970 responds. 47 00:03:08,210 --> 00:03:10,570 So it's going to give us the information over here. 48 00:03:10,570 --> 00:03:12,610 It's going to answer the query. 49 00:03:12,810 --> 00:03:13,010 Okay. 50 00:03:13,010 --> 00:03:14,610 So we'll get all the information out here. 51 00:03:14,610 --> 00:03:17,570 As you can see it's providing all the information here. 52 00:03:18,010 --> 00:03:23,450 After this what we are going to do is we are going to provide it a file, a document which contains 53 00:03:23,450 --> 00:03:25,460 that specific answer. 54 00:03:25,700 --> 00:03:30,220 And then we'll again test the tool with the same query. 55 00:03:30,820 --> 00:03:33,900 So right now as you can see it has given us the answer. 56 00:03:33,900 --> 00:03:41,420 And let's say this is the document which we ideally want to upload as one of the files which we give 57 00:03:41,420 --> 00:03:42,620 to the custom GPT. 58 00:03:42,860 --> 00:03:47,780 So we're going to give this and we're going to upload it over here. 59 00:03:53,820 --> 00:03:55,020 So now we have uploaded this. 60 00:03:55,060 --> 00:03:56,620 It's getting uploaded over here. 61 00:04:01,340 --> 00:04:04,780 Once it is uploaded we can update it. 62 00:04:13,540 --> 00:04:17,700 And now we can view the GPT. 63 00:04:17,940 --> 00:04:18,420 Okay. 64 00:04:18,460 --> 00:04:21,740 And now we can give the prompt once again. 65 00:04:21,780 --> 00:04:27,300 So if you see we have this question in this FAQ right now which is how can I improve my quality score 66 00:04:27,300 --> 00:04:28,140 in Google Ads. 67 00:04:28,420 --> 00:04:33,220 So these are the this is the answer given called score is influenced by ad relevance landing page expected 68 00:04:33,260 --> 00:04:33,780 CTR. 69 00:04:34,060 --> 00:04:40,300 To improve it, you can use highly relevant keywords optimize ad copy, ensure fast loading, increase 70 00:04:40,300 --> 00:04:40,820 CTR. 71 00:04:40,860 --> 00:04:43,580 So this is the ideal response which we are looking for. 72 00:04:45,180 --> 00:04:46,940 So we are going to give the same question. 73 00:04:54,100 --> 00:04:55,180 Which is giving us the answer. 74 00:04:55,180 --> 00:04:59,100 So now you can see it is picking up similar information. 75 00:04:59,100 --> 00:04:59,900 AD relevance. 76 00:04:59,900 --> 00:05:03,180 We talked about ad relevance landing page experience, create fast loading. 77 00:05:03,180 --> 00:05:05,500 We had talked about that over here as well. 78 00:05:05,540 --> 00:05:10,460 Fast loading expected CTR okay. 79 00:05:10,500 --> 00:05:12,380 Additional tips to improve okay. 80 00:05:12,420 --> 00:05:16,300 So now it's going to pick up the answer from the document which we have given. 81 00:05:16,300 --> 00:05:19,260 And it's more aligned to that document's answer. 82 00:05:19,420 --> 00:05:22,460 And that is what is being communicated to our customers. 83 00:05:22,460 --> 00:05:30,220 So this is how you can build your own personal business chatbots by empowering them with AI, by giving 84 00:05:30,220 --> 00:05:36,540 all your company documents, files to it in the custom GPT, and then let it do the job of scanning 85 00:05:36,540 --> 00:05:39,900 through all those documents and answering your customer queries.