AI Has Crossed the Line

AI has crossed the line and business leaders cannot get enough.

AI has crossed the line and business leaders cannot get enough.  

Key Milestones 

For years people have wondered when artificial intelligence (“AI”) would advance to the stage that computers become smarter than humans. We watched in amazement as AI technology passed key milestones:  

  • May 11, 1997 - the IBM computer called Deep Blue beat the world chess champion after a six-game match.  
  • February 2011- IBM’s Watson computer beat Ken Jennings on Jeopardy. Ken Jennings, you may recall, had won 74 games of Jeopardy in a row.   
  • June 2018 - IBM’s Debater AI argumentation tool debated the Israeli debate champion, Noa Ovadia.  

These milestones mark the progress toward singularity. In technology terms, singularity is achieved when computer programs become so advanced that artificial intelligence transcends human intelligence.  

Singularity is still the stuff of future science. Business AI is the stuff of today.  

Business AI 

In 2019, the AICPA and the CGMA studied the impact that AI was having on business, particularly on the financial industry. The report stated that “as AI develops, it's going to be replacing white-collar jobs in fields such as accounting”.1 The study was looking into the future and concluded that AI was about to surpass human intelligence.  

Here we find ourselves at the end of 2023, almost four years after the AICPA / CGMA study. AI has continued to advance. Today’s present is 2019’s future. Today, there are several AI applications that outperform human intelligence, particularly when trained for a specific task.  

Artificial Intelligence and the R&D Credit  

Consider the R&D credit industry. There are no exact numbers, but each year taxpayers pay tax consultants approximately $5 billion to compute and defend R&D tax credits. This is a significant compliance cost that business leaders are extremely interested in reducing.  

SPRX has recently released the SPRX Platform, an AI application that outperforms R&D credit consultants.  

  • AI ingests, structures, and heals data faster than consultants and spreadsheets   
  • AI performs computations faster than consultants with spreadsheets   
  • AI reads, understands, and scores R&D credit documentation faster than consultants   
  • AI drafts R&D project analysis memoranda faster than consultants  
  • AI prepares reports, schedules, and tax forms faster than consultants  

Not only is AI faster, AI consistently performs at an elevated level 24/7. AI does not have bad hair days, does not get sick, and does not take vacation.   

You may ask, “how can a computer outperform or even replicate a tax consultant?" The answer lies in understanding machine learning and neural networks.  

Machine Learning 

Machine learning is a branch of artificial intelligence and computer science that focuses on the use of data and algorithms to imitate the way humans learn. 

A tax professional learns through a step-by-step process.  

  • First, the person learns the tax rules and creates an analysis framework for making a tax decision. In the case of the R&D credit, that framework might be the four-part qualification test.  
  • Next, the person processes client data and compares the data to each element of the tax decision framework to make a prediction regarding appropriate tax conclusion.  
  • Next, the person discusses the analysis with a more experienced professional reviewer to verify the accuracy of their prediction. The reviewer provides feedback regarding the accuracy of the prediction and learning occurs.  
  • Last, the person stores their experience in their memory so that the next time they encounter a similar task they apply the learning to the repeated task.  

AI models learn through a similar process. The tax analysis framework is loaded into the system through a series of complex algorithms. Next, data is loaded, and the system compares the data with the decision framework and predicts the appropriate tax conclusion. 

The prediction is reviewed by an experienced professional reviewer who provides feedback regarding the accuracy of the prediction and learning occurs.  

The Human Brain 

Review the graph at the beginning of this article. Notice the difference between the growth rates in human performance and AI performance. AI models can learn much faster than humans because computers process data at a higher rate and run continuously 24/7.  

In the tax consulting industry, it takes a person several years to reach the manager competency level. An AI model can be trained in a matter of weeks to perform at that same level.  

Neural networks are to the computer what the brain is to the tax professional. The structure of a neural network mimics the way the human brain functions. In the human brain, biological neurons signal one another to process data. With a neural network, node layers signal one another to replicate the human brain.  

In the human brain there are billions of neurons sprouting thousands of branches. A neural network stacks layers of nodes and runs simultaneous processing on each node to replicate the operation of the human brain. 

Each node connects to another node and has an associated weight and threshold. If the output of any individual node is above the specified threshold value, the node is activated, sending data to the next layer. This process is repeated billions of times.  

The human brain is truly remarkable and performs vast computations all within the space between your two ears. Neural networks are remarkable and perform vast computations throughout scores of networked computers worldwide. 

New Expectations

Ninety-five percent of business leaders believe that their organizations will benefit from embedding AI into business operations, products, and services.2   

Today, the SPRX Platform is an AI tool that automates 80% of the manual tasks performed by R&D credit consultants. Automating the manual spreadsheet processes, document analysis, and report preparation allows the R&D credit consultants to spend more time on strategy and decisions related to the R&D credit. As a result, taxpayers get better value at a lower cost.

The SPRX Platform can help everyone involved in Research and Development. Both taxpayers and R&D credit consultants benefit from adding AI to their processes.