A Financial AI Chatbot SaaS Company

We used our AI Expert System to deliver an AI chatbot SaaS company to a client of us. The system has access to 1.5 million Swedish companies, and revenue for 5 years. This includes all important figures such as net sales, tax payments, payroll costs, etc.

The system uses our AI Agents feature to connect ChatGPT to the database. This allows users to use natural language to ask questions and have the LLM provide advice about the company's health, KPI figures, and assess if the company is in good shape or not. You can watch me demonstrate the system in the following video.

Just like in our previous blog, the system will allow the user to ask some handful of questions for free each day. Once the free questions have been exhausted, the system will automatically ask the user to pay for a monthly subscription.

Subscription payments

This allows our client to create a SaaS company based upon our technology and combine it with ChatGPT.

ChatGPT can't do this

If I had a dollar for every time somebody asked ChatGPT for financial advice, I'd be a very rich man. The problem is that in order to assess a company you'll need a lot of data. The system we're delivering has historical financial data for the last 5 years for companies in its database. Each year has 150+ different values, and the system works with information such as for instance.

kr_ret_on_tot_cap_ebitda_perc - Return on total capital (EBITDA %)
kr_ret_on_working_capital_perc - Return on working capital (%)
kr_degree_of_debt_intr_bearing - Interest-bearing debt ratio
kr_opp_result_num_of_employees - Operating result per employee
kr_change_in_num_of_employees - Change in number of employees
kr_accounts_rec_turnover_perc - Accounts receivable turnover (%)
kr_current_liab_turnover_perc - Current liabilities turnover (%)
kr_risk_buffer_on_op_cap_perc - Risk buffer on operating capital (%)
n_perf_bonus_to_oth_employees - Bonus to other employees
kr_return_on_capital_percent - Return on capital (%)
kr_return_on_cap_emp_percent - Return on total capital (%)
kr_avg_debt_eq_ratio_percent - Average debt/equity ratio (%)
kr_risk_buffer - Risk buffer total capital
kr_ebitda - Earnings before Interest, Taxes, Depreciation and Amortization, EBITDA
kr_ebitida_margin_percent - EBITDA margin (%)

I'm not going to bore you with the complete lists of columns, but basically the system has access to "everything" that's relevant to perform an accurate evaluation of a company's health. The idea is to sell access to the AI chatbot to others needing advice related to investments into Swedish companies, providing an AI based financial assistant, capable of extracting accurate information about individual companies.

Due to having access to real time live information, this allows ChatGPT to use this information on demand, and answer complex questions the user might have about individual companies. This completely eliminates AI hallucinations, and allows the chatbot to always give accurate answers, based upon factual information, avoiding guesswork.

The system is scheduled to going live in its MVP form in a couple of weeks, and my client has already started onboarding testers and contacting journalists to gather interest in the system and start testing it. When the system is done, I've spent about 6 weeks on it in total, and the client has a complete SaaS company selling access to a highly specialised AI chatbot.

I'm just delivering the infrastructure, allowing the client to white label our solutions - So any revenue the company is generating becomes our client's revenue. We simply sell the infrastructure to create and host the system, in addition to custom software development to implement any customisations.

Implementation details

The system contains about 10 database tables, hosted in PostgreSQL, and managed by us. In addition it builds upon Magic Cloud, and the database is created by importing financial information as reported by the companies to the government each year. Then when the user is asking the AI questions, we use our AI functions feature to retrieve information from the database.

The system allows for searching for companies, retrieving directors, basic information such as address and other details - In addition to accounting details and core financial numbers.

This allows us to "seed" ChatGPT with factual information as questions are being asked, and such have it function as a financial assistant, helping its users to assess companies they're potentially interested in investing in.

Got AI idea, need SaaS?

My objective with writing this article is that we've got all the infrastructure required to create, implement, and host an AI SaaS company. If you've got custom data, and/or an API, and you're interested in creating your own SaaS company, you can contact me below.