I Don’t Want Your Bank’s UI. I Want A ChatGPT-Like Natural Language Interface.

There’s a tedious, time consuming task that we all dread doing, but yet are compelled to do several times a month: interact with our bank’s UI.

Regardless of the convenience of mobile over web, it still ends up being a multiple step process that requires several presses/clicks, screen loading icons, scrolling up and down the page, sifting through a bombardment of calls to action, promotions, etc.

I love Chase bank, but this repetitious flow is a time suck and unenjoyable.

As an example, here are the steps I currently go through to pay my credit card bill:

First, I go to the home page.

For whatever reason, I have to wait for the welcome back widget to load asynchronously. Then I have to press I sign in button to get to the sign in page.

Then I’m brought to a page where I enter my username and password.

Finally, I’m at the home screen, where I’m bombarded with plethora calls to actions, promotions, navigation options, rewards, etc. I have to scroll and strain to find the credit card I want to pay and the Pay card button to press.

After clicking Press card, I and am brought to a screen with a bunch of options I have to sift through.

Now I’m brought to another screen, with all the same information I put in the last screen.

After reviewing that, I click Pay it.

Now I get the reward of another screen showing me all the things I just went through.

Finally, I can press Done and finish this 6 screen journey, just to pay my credit card.

ChatGPT-4 and plugins will soon be able to integrate with your bank’s APIs, and when that happens, the flow is going to be much more succinct and less tedious.

Rather than being bombarded by promotions, calls to actions for things you don’t want to do, etc, you will simply tell ChatGPT what you want to do in a natural feeling textual exchange (ie Natural Language Interface), with very little hassle.

Instead of having to log into your bank’s app every time you want to do financial business, an AI chatbot will retain a long lived OAuth access token which it uses to interact with your bank’s API.

Consider this simulation, which I had ChatGPT-4 create for me, using the following prompt:

Please simulate an interaction between a chatbot which has access to my banking api.
The user will ask to show Chase bank accounts.
The chatbot will respond with the following account data:
The accounts should be listed as a table with two colums: one for the Account Type, and one for the balance or amount owed
Checking 9999 with 4,003,567.32 as the balance
Savings 8888 with 333,353.24 as the balance
Credit Card 7777 with an amount owed of 4000
The user will then ask to pay 3000 towards their credit card, using money from their Checking 9999 account.
The chatbot will confirm with the user that they want to make the transfer.
The user will respond: yes
The chatbot will then interact with the banking api, then display the same table as before, but this time with balances that are updated to reflect the recent credit card payment

We can go even further and explore current capabilities of financial assessments.

Using GPT-4, after 4 revisions, I ended up with this zero shot prompt, where I provided a list of transactions, in csv format, and asked ChatGPT to categorize each transaction, and give me a percentage breakdown of each category.

I am going to give you a list of transactions from my credit card.
For the month of May 2023, please break down the expenses into a table of categories with 2 columns: one for the category and one for the percent spent in that category.
In a separate table with two columns, please provide me the total amount spent in May 2023, and the amount paid in May 2023.
The list is provided in csv format.
The first row of the list is the description of the columns.
The list is as follows:
Post Date,Description,Type,Amount,Memo
5/12/23,INFINITY CABLE COMM,Sale,-55,
5/11/23,GOOGLE *Colab,Sale,-9.99,
5/10/23,APPLE.COM/BILL,Sale,-0.99,
... (rest of transactions)

If I want fancier visuals, I can send that table to another plugin, like Wolfram Alpha and get a pie chart visualization. There will likely be orchestrations between many plugins in the future.

e.g. A financial data retrieval plugin that hands off data to an assessment plugin, which then hands off to a visualization plugin, etc.

As we gaze into the future of digital banking, it’s clear that the next logical step is to incorporate functionalities such as trading stocks and cryptocurrencies, applying for credit cards, opening accounts, and perhaps even applying for loans, all within the same natural language interface.

While it’s conceivable that each banking institution could develop its own proprietary bot, a more plausible scenario is the emergence of a handful of powerful, versatile chatbots — think ChatGPT, Bard, and the like. These AI powerhouses will likely integrate with a multitude of companies and their respective backends, essentially becoming a universal hub for users.

I see a future where a significant portion of our daily tasks, currently performed through cumbersome and disjointed user interfaces, will be accomplished through a single, intelligent chatbot, using a Natural Language Interface. This is likely going to be a revolution in user experience, offering a streamlined, efficient, and more engaging alternative to traditional banking user interfaces.

Banking is just one of the many interfaces to consider as candidates for having their traditional UIs supplanted by NLIs. Consider also:

• Google searches where you have to open several links, and read through comments in order to glean answers.
• Scheduling appointments with doctors, dentists, reservations, etc.
• Grocery delivery (Instacart is an already popular ChatGPT plugin)
• Pharmacy refills
• Finding maintenance workers (e.g. a natural language interface for Thumbtack)

NLIs aren’t just limited to text. ChatGPT’s recent iOS app includes voice to text. A natural next step will be an always on, GPT powered AI assistant.