SIGN up
The Easiest Way to Work with Data
Now everyone can leverage the power of AI to grow their business.
Simple chatbots are easier to build than ever before. If you have straightforward needs, like guiding users to a specific section of your site, gaining user feedback, or offering support, tools like Landbot make it easy to get up and running in minutes.
However, if you have more complex needs, like lenders predicting borrowers credit approval or analyzing user sentiment, then it’s a lot harder. Indeed, Landbot raised $2.2 million “for its on-message ‘anti-AI’ chatbot.” There’s typically no (or very little) AI functionality in the simple, drag-and-drop chatbot builders, like Landbot, Joonbot, Wotnot, and so on.
If you want to build an Artificial Intelligence (AI) powered chatbot, then you’re faced with automation options like Google’s DialogFlow or SAP Conversational AI, which are immensely complex and take months and lots of time reading the FAQ to figure out.
In short, there are two ends of the spectrum, without much middleground: Simple, no-code bot builders without AI functionality, and extremely complex AI-powered bot builders.
With Akkio, you can reach a nice middleground, and create feature-rich AI-powered chatbots with simple bot builders. In this hands-on guide, we’ll make an AI-powered chatbot in Landbot in minutes, which will score applicants by likelihood of repayment automatically approve or reject extending credit during the application process.
First, sign up for Akkio. Then, hit “Create New Flow” to get started making an AI flow.
Then, click “Table,” as we’ll be working with this tabular dataset of loan approval information from Kaggle. Note: Each business will want to customize their credit approval model for their business process - to account for their unique customers, credit scores, interest rates, etc.
After uploading the Kaggle dataset, you’ll get an overview of the table, as seen below.
Now that the data is uploaded, hit “Add Step” and then “Predict.”
We then simply select the column we want to predict, which is “Loan_Status.” Optionally, you can click on the “Ignore” tab and select “Loan_ID” and “Gender,” as these fields aren’t desired for our model. Then, click “Create Predictive Model.”
And we did it! We’ve now built an AI model that accurately predicts loan approval.
Our final step in Akkio is to deploy it, so just hit “Add Step” and then “API.” You’ll then see a screen like the below, and you can simply click “Deploy” to finish.
Now, let’s sign up for Landbot, which we’ll use for the chatbot functionality.
Then, click the “loan application” template to get started.
This chatbot template walks users through a loan (or credit) application process, with an option to send that information over to a loan provider for manual review. We can use the Akkio AI model we just created to automate the loan approval process, and immediately give users an answer, saving huge amounts of time and money.
We want to edit the Landbot to ask users for the same information used in our predictive model. That data was:
Here’s what the Landbot looks like with those questions. I simply replaced the existing blocks with questions asking that information. As you can see, it still all fits on one screen! Make sure to save each input as a variable in Landbot, so we can pass these variables to the Akkio model.
Next, we add a custom webhook, and under “send params?”, we enter the data from Akkio’s API deployment.
The last “key” field, named “data,” is where we input the data to feed the Akkio model. The webhook should then look something like this:
Now, we can take the output of our webhook, and give it to the user.
By hitting “preview” in the top-right of Landbot, we can test it out, and voila, it works! Users answer a series of questions, which gets sent to the AI model we made and deployed in Akkio, and a prediction is returned to the user.
In a matter of minutes, you can build an AI-powered, automated credit approval chatbot. This is revolutionary, because it’s an example of something much bigger: You can build and deploy AI models to predict practically anything in a number of minutes.
In late 2020, an AI startup called Zest raised $15 million for its AI-based credit underwriting software, for a total of over $230 million. In 2019, a similar AI lending startup called Upstart raised $50 million.
More recently, UK startup Fintern raised £8 million in February 2022 to bypass credit scores in makinng lending decisions. Instead, the firm deploys AI models that read through alternative data to better assess an individual's spending habits and ability to repay a loan.
What do these startups have in common? They're all using AI to re-evaluate how we assess creditworthiness.
For decades, credit scores have been the primary way that lenders determine whether to approve a loan and at what interest rate. But there are many problems with this system. For one, credit scores don't take into account an applicant's entire financial history. They also don't consider an applicant's current circumstances, such as a recent job loss or medical emergency.
This is where AI comes in. By reading through an applicant's entire financial history, AI can get a more holistic view of their creditworthiness. AI can also be used to constantly monitor an individual's financial situation and make lending decisions in real-time, rather than relying on outdated credit scores.
In short, AI-based credit lending is incredibly valuable, and it used to be incredibly difficult. Today, you can use no-code AI to accomplish wildly difficult tasks in a fraction of the time.
In this guide, we’ve explored using Akkio, a no-code AI tool, and Landbot, a no-code chatbot builder, to build an automated credit approval bot for financial services.
However, the same principles and steps apply to a wide-range of use-cases. If you visit Kaggle Datasets, you’ll find thousands of ways to apply no-code AI. For instance, you could use AI to predict corporate bankruptcy, credit card fraud, Kickstarter project success, customer churn, or any of a million other metrics. As long as you have sufficient, high-quality historical data, you can make a predictive model.
A number of these use-cases could be integrated into existing chatbot platforms like Landbot to provide valuable services without writing a line of code. Let’s explore how this might work for two different applications: customer support and product recommendations.
In the customer support context, a chatbot could be designed to take the customer’s reported issue and pass it through an AI-based issue classification system. This system would use historical data to classify the issue and route it to the appropriate support team. The bot could also provide the customer with an estimated time to resolution based on past similar issues.
In the product recommendations context, a chatbot could be used to extract user preferences from natural language input (e.g., “I’m looking for a gift for my wife”) and then use those preferences to make personalized product recommendations. The bot could also keep track of products that have been recommended in the past, so that if the user asks for a “recommendation for a gift for my wife” again in the future, the bot can provide an updated list of products based on her recent activity.
Both of these applications are just a few examples of how AI-based bots can be used to provide valuable services without any coding required. If you have access to high-quality historical data, there are endless possibilities for what you can build. So get creative and see what you can come up with!
This guide can be used by any company that issues credit to make their processes more efficient, by improving customer service while cutting costs.
That said, this guide can be so much more. Consultants and agencies can use this guide to increase their repertoire of services, and offer AI-powered chatbots to their clients.
Entrepreneurs can use this guide to build an AI startup faster and cheaper than ever before, and intrapreneurs can use this guide to implement AI in their organizations, even without having any technical expertise.