It’s been an exciting month at CustomGPT! As you know, we are committed to bringing you The No. 1 platform for building custom ChatGPT chatbots with your own data.
We have listened to your feedback, and our team has been working tirelessly (over 10,000+ man hours) to bring you accurate trusted ChatGPT responses – from your content – without making up facts. All within a secure, privacy-first, business-grade platform.
“I like to Pick Winners – I Picked CustomGPT”— Bill French, OpenAI Thought Leader
???? So buckle up, let’s dive into our mega roundup:
We continue to improve our proprietary anti-hallucination technology and have now battle-tested it across over 2000+ businesses.
This is by far our most popular feature and a core part of our philosophy to create a RAG (retrieval augmented generation) platform with trust and transparency.
We’ve upgraded our proprietary citation calculation algorithm for improved precision. Now, you can clearly trust the sources from which the responses are coming from within your content.
Take a look at the above example. In this case, the chatbot was built with trusted domain-specific papers from Pubmed and curated articles. When the chatbot responds to the user question, the sources are clearly shown so that the user is confident about where the answer was generated from. This combination of anti-hallucination and citations add trust and transparency to the AI answers.
And point to note: the answers are coming from the content of the actual papers and articles – not a summary or snippet (like some other systems). This is the level of trust needed for most business use cases.
We now have 11 free tools to help you build your chatbot. Industry secret: Data ingestion is by far the *toughest* problem in AI, and we’ve made it ultra-simple to ingest your data. To the point where non-technical people are building amazing super-smart chatbots.
Using these free tools, you can quickly ingest your website content. Or even webpages from other web sources or competitors. We have democratized the AI chatbot process so that any person can build a custom chatbot – with zero coding required.
Our platform is now 201 days old and over 3000+ paying business customers have created over 12,000+ chatbots.
The main reason we hear : Ultra-simple no-code setup, even for the most complicated content ingestion cases.
Thanks to our real time indexing, you can now build chatbots with “just in time” learning ability. Create tools and workflows with realtime information and generative AI (aka: no more “ChatGPT knows only till June 2021” problem!)
Customers are even using the API to build “just in time” workflows by ingesting realtime information into the chatbot.
For example: if a news story breaks at 4PM US EST. That very second, it can be pushed to the chatbot using our simple API and then the custom chatbot can be queried for that information.
Customers are building chatbots with billions of words, 100,000 documents and 100,000+ queries/month. The prototyping phase has now transitioned to real-world deployments.
Unlike similar systems where you upload 1-2 PDFs, this system has been designed to build the large-scale chatbots required by businesses. So whether it is 100,000 documents or websites with 100,000 web pages, the system has now been scaled and tested for such use cases.
In the above example, the user has consumed 55 million words out of the 300M words available on the $499 plan. To put that in context, all 7 Harry Potter books combined are about 1 million words. So in essence, you could ingest 2000 Harry Potter books into your chatbot for $499/month.
Using the new Persona feature, customers are controlling exactly how the bot behaves. Savvy customers are even building “super agents” where multiple CustomGPT chatbots are controlling and interacting with each other.
These are next-gen use cases where multiple CustomGPT chatbots (called “agents”) are working in concert to create a “super agent”. This allows expert chatbots trained on their own custom knowledgebases to act in a specified way.
For example: Imagine 8 people in a conference room. Look at their roles and functions. Now imagine their role being enhanced with a custom chatbot trained specifically to behave like an expert trained on domain-specific knowledge.
We’ve invested significant resources to enhance our DevOps and MLOps. These unheralded systems go a long way to have a solid production-grade system (so that you can sleep well at night!)
While creating a prototype might be quite easy these days (“it is a few lines of python code after all”, said the developer), then creating the production-grade systems for deployment is the real challenge.
We see this in play day-in and day-out across our 3000+ paying customers who demand the best from our systems. As customers start deploying the initial proof-of-concepts into their business workflows, the importance of DevOps and MLOps cannot be underestimated.
We now have MFA (multi-factor authentication) ,with Google Authenticator, for businesses that need the enhanced security.
We are committed to putting in more features related to the privacy and security of your data. And this feature, along with our overall commitment to security and privacy should go a long way.
With 3 lines of code, you get the entire power of the CustomGPT platform. So just imagine doing RAG (retrieval augmented generation) with trust and transparency, within minutes.
We have now enhanced our cookbook with more examples and updates. Any developer should now be able to create a custom chatbot within minutes. Side note: If you don’t have a developer and need one, Upwork is a good resource for such projects. (See example)
Best part: As soon as you finish the proof-of-concept, you are ready to deploy at scale.
If you need help to get started, have questions, scale up or begin deployment, please book a quick call.
We’ve staffed up our customer success – and are vested in your success. So whether it is questions about our platform, questions about whether we are a good choice for your use case – or scaling and deployment questions, we are available to help.
Not everyone can do credit cards. If your procurement department needs a signed service order or PO, we now support procurement processes. We also support ACH bank transfers and custom plans. Book a call.
Ask The Bot
Our chatbot knows pretty much all our own content. Please feel free to ask the bot anything. So instead of waiting for a response from customer service, there is a very good chance that the chatbot knows the answer to your questions.
Yes – we use our own chatbot for our customer interactions. And at this time, our chatbot gets 10,000+ questions each month. And it works 24/7 trained on all our content.
Ninja tip: By analyzing the chat logs, you can understand what content is missing from your website.
We have scaled up our customer service and are now able to respond within 1-2 business days.
We have now enhanced our step-by-step guides and documentation. We now have over 58 step-by-step guides for your every need. Each guide is a clear 1-2-3 style guide with handy images.
You can now export the full chat logs for your chatbot (from both the Dashboard and API). This shows you exactly what the chatbot is answering, and more importantly, what your users are asking it.
Ninja Tip: Analyze the chat logs with “Code Interpreter” to dig into customer intelligence or create an audit trail. For example: What content is your website missing?
We’ve now implemented a proprietary “smart agent” that can interpret the multi-turn conversation for much better responses.
Imagine a conversation with the chatbot where the user is having a long conversation. Unlike other systems, we have implemented a proprietary “smart agent” to understand the conversation and send the correct context, from your content, to ChatGPT for the responses. This has greatly improved the quality of our responses.
In the above example (for a rare disease diagnosis tool), the user is having a diagnosis conversation with the chatbot to gain awareness. Notice how the chatbot is able to understand the conversation and provide trusted responses from the domain-specific training data. See live demo.
Convert your images, faxes and scanned PDFs into a smart chatbot and get instant answers from it. No more painfully reading scanned images.
Fun side story: A 130-year old newspaper is digitizing its entire archive and making it available with a CustomGPT chatbot.
Fine-grained control on your pages. You can now refresh or delete individual pages from the index to stay consistent with your knowledge base.
For example: If your FAQ page changes, you can quickly re-index it and make sure the chatbot is consistent with your content.
We continue to devote an extreme amount of resources to improve our data parsing systems. From Youtube videos, Audio transcriptions (using OpenAI Whisper), CSV handlings, PDF Parsing, etc — these systems have key implications to the answers you get from your chatbot.
As anyone who has worked on AI knows, data integration is the most painful part about dealing with AI systems – and at CustomGPT we have made it simple to ingest your data.
So simple that even non-technical people are able to use the no-code system to ingest websites, helpdesks, Youtube videos and even their audio recordings and podcasts.
Allows multi-user access. Add team members to help you build chatbots. Or just query them privately. Imagine 3 lawyers collectively working on a case.
Imagine a completely private chatbot available only to the 6 (or 600!) people on your team. You can build this chatbot with your private documents and them embed the chatbot on your intranet – accessible only by your authorized users.
This would take employee efficiency and team collaboration to the next level.
We are building trusted partnerships to help you work with our platform. From large consultancies, to agencies to individual developers, you can get the support you need.
Even though our no-code platform is ultra simple to use, sometimes businesses just need some help.
Got questions? Talk To Us !