AI vs. Machine Learning vs. Decision Engines for SME lenders

Man in black coat using mobile phone and laptop

The powerful technologies of artificial intelligence (AI), machine learning(ML), and decision engines have given SME lenders a plethora of benefits—faster loan processing, efficient lending workflows, and accurate risk assessments, to name just a few.

The three technologies are often spoken about interchangeably, and there certainly is a degree of overlap between them. However, just under the surface, there are significant differences regarding the benefits they bring about and how they are applied in alternative lending.

In this article, you’ll learn what AI, ML, and decision engines are, and how each of them is taking the alternative lending industry to a new level. We’ll also explore what we expect to see from these technologies.

Before we get into it, a note for SME lenders: adopting these technologies sooner rather than later is a wise business decision.

What Is Artificial Intelligence?

Artificial intelligence (AI for short) refers to the ability computers have to reproduce or emulate human thinking and cognitive skills. It encompasses several more specific technologies and learning methods, such as machine learning, deep learning, natural language processing, and speech recognition.

The first AI program was successfully implemented in 1952 by Christopher Strachey from the University of Oxford and was designed to play the game of checkers.  Although AI was initially applied to the computing industry, it has since spread widely to other sectors, including alternative lending. In fact, AI could deliver up to $1 trillion in additional value for the global financial services sector each year.

It is important to note that for many lenders (particularly mid-sized lenders), AI is rather cost-prohibitive. Do you build your own technology? Pay for a custom solution? Will you need to make changes to the infrastructure of your business? Hire additional employees? What about maintenance and upkeep? 

While costs tend to vary based on the type of technology, most cloud-based systems (i.e., the most common type of AI tool for medium-sized businesses) will run a pretty penny, especially when you consider training and maintenance costs. Just something to keep in mind as you work your way through this article. 

If you want to learn more about pricing, check out this great write-up by Wired.

What Is Machine Learning?

Machine learning (ML) is a subset of AI. It’s one of the many learning methods that AI technologies use to interpret and act on data.

ML encompasses the algorithms and technologies that enable machines to leverage comprehensive data sources, learn from them, implement decisions, and iterate on them. After being initially programmed manually with code, these intelligent algorithms essentially improve upon themselves without needing explicit human intervention.

For instance, companies in the automotive industry collect data from all aspects of driver performance and feed it into machine learning algorithms. These algorithms then churn data to catch seemingly minute, yet significant factors to improve safety, such as turn speed and angle and even the mental state of the driver (read more here).

Machine learning has also produced far-reaching applications in the aerospace, hospitality, communications, and logistics industries, among many others.

And, of course, the financial services industry has also benefited from the use of machine learning. A study conducted on the state of AI in the market reported that machine learning and deep learning accounted for the largest market share (38.9%) of AI. 

What Is a Decision Engine?

A decision engine is a software program that helps businesses automate decision-making. It does this by leveraging customizable preset rules created by the user.

Decision engines are also data analytics powerhouses, capable of crunching large volumes of data in minutes, even seconds. In essence, decision engines allow your business to make smarter data-driven decisions with minimal oversight by your employees.


Ever used Google flights to book your travel? That’s actually a decision engine. It considers the parameters you set and then searches the web for the best deals for you. Travel agents would struggle to find a good deal without their own air travel decision engines. 

Decision engines are being used in different industries to automate manual processes that rely on consistent requirements and countless inputs—like loan applications.

SME lenders can benefit specifically from credit decision engines to automate their lending workflows. A widely customizable decision engine can enable you to deny or accept applicants based on parameters that usually only your underwriters can assess. This most often arises in the form of business credit scorecards.

By integrating the work capacity of decision engines, SME lenders can increase the accuracy and speed of the underwriting process, leading to more significant revenue potential, with less risk.

Benefits of AI, ML, and Decision Engines for SME Lenders

Now that you know what these technologies are and how they work, let’s explore some of the specific benefits brought by ML, decision engines, and AI in SME lending.

Accelerate the Loan Application Process

Decision engines are often deployed to speed up the manual processes of the application stage of your lending operation.

To be eligible for funding, SMEs must meet the criteria that you (as the lender) determine beforehand. Decision engines can, without manual supervision, help discard unwanted applications and move along those that meet your standards. 

AI programs also help by auto-classifying applicants before they reach the manual assessment stage, so it’s easier for your employees to sort through and process the eligible applicants.

Improve the Customer Experience

While decision engines go a long way in streamlining the loan application process, they also help improve the borrower experience. 

For example, lenders are using decision engines on the back end of their websites to recommend the most appropriate financial products to borrowers based on factors like their current financial situation, previous financial products, and business goals. This approach helps lenders strike the right balance between automation and human intervention, while also providing ample opportunity to pivot as new information emerges.

Financial service providers are also using AI and ML-based applications to analyze customer behavior from every communication touchpoint on their website and mobile applications. This allows them to better understand customer preferences and improve overall interactions. A great example of this is AI-driven chatbots.

Finally, AI-automated “know your customer” procedures allow for a smoother and quicker client onboarding experience. EXL service, a company that worked with HSBC to streamline KYC procedures, claims that automating KYC can reduce onboarding costs by up to 70% and reduce turnaround time by up to 90%. 

Minimize Human Biases and Errors

Basing lending decisions entirely on human judgment could result in unfair and, at times, non-inclusive decisions at the loan origination stage.

Implementing decision engines can lead to more rational and pragmatic decisions, with fewer biases. The more data you feed a decision engine with, the more it understands a borrower’s situation from a financial and business standpoint, rather than ethnic or socio-economic background. This means you can accept good applicants and reject bad applicants with more certainty. 

While AI and ML have made great strides in recent years, it’s important to recognize that the world’s best and brightest are still struggling to eliminate biases from these tools. News story, after news story, after news story has shown us that AI and ML are not where we want them to be and that we are not quite ready to fully eliminate the human element when it comes to potentially life-changing financial decisions.

Contrast this to decision engines, which still have the necessary degree of human intervention to correct the trajectory of an analysis when systems churn out biased results

Finally, beyond the bias discussion, it’s worth generally highlighting that humans are also prone to errors in repetitive tasks, like sifting through piles of loan applications or transferring and encoding data. It is here that decision engines also shine as they make decisions only when specific conditions are met, improving accuracy, saving time, and resulting in more efficient workflows for both SME lenders and borrowers.

Trends indicate that we’re yet to see the best of what they can offer to the alternative lending industry.

What Does The Future Hold?

With the rapid adoption of machine learning, decision engines, and AI in SME lending, technology from even just a decade ago seems ancient and outdated. As these tools continue to advance, their use cases in the industry will also continue to grow.

Here’s a glimpse of what the future holds for SME lenders who adopt these technologies.

Reach Underfunded SMEs

Decision engines have the potential to improve the fairness and accuracy of credit scoring by changing the credit scorecard methodology altogether.

ML models can theoretically handle infinitely more variables than standard credit models like FICO and, more importantly, don’t require credit history data. This means that previously unserved and underbanked SMEs will have new opportunities to access credit.

When lending to SMEs that seek capital for the first time, lenders understandably have doubts about the business model’s sustainability, lack of sufficient financial performance data, and the possibility of fraud. AI and ML programs will be able to harness big data to analyze the creditworthiness of all SMEs, even those with little to no financial history. 

Lenders can thus extend lines of credit to underbanked businesses that are capable of paying back. And conversely, stop lending to vulnerable companies that erroneously appear to be creditworthy. Ultimately, by leveraging machine learning and AI in SME lending, lenders could earn more revenue while taking on less risk.

Improve Credit Risk Analysis and Underwriting

The biggest hurdle to adopting these technologies is learning to interpret the results of such models. Model understanding is of prime importance in decision-making, especially if it comes to allocating capital to counteract credit risk.

This challenge could soon be solved as more methodologies such as the SHAP (Shapley Additive Explanations) for interpreting ML models become more common. In addition to interpreting results, credit risk teams will also be able to reproduce these models for similar portfolios.

Other advantages of ML credit risk models include fewer limits on the number of variables and data types, leading to more accurate risk-based pricing and fairer lending practices. A case study by Cognizant revealed that machine learning improves underwriting efficiency ten-fold and increases accuracy by 83%.

Lastly, it’s worth noting that AI and ML don’t get to have all of the fun when it comes to underwriting. 

Decision engines are already in use in most credit risk departments, and their utility is only continuing to increase as the technology advances. This will enable manual underwriters to analyze bespoke situations and emerging circumstances of clients at scale and has the potential to materially improve credit forecasting and limit losses incurred from defaults. 

Advanced Fraud Detection and Anti-Money Laundering Protection

Fraud in the small business lending industry has increased by 6.9% since 2020.

Although AI and ML are still in their infancy with respect to compliance, they’ve shown great promise for the institutions that have already adopted them. ML will likely be a potent tool used by banks and other financial institutions to detect criminal activity more acutely. 

These intelligent self-learning algorithms can work for compliance departments by interpreting existing data and then predicting and adapting to identify future malpractice instances. When trained with high-quality data sets, these programs can detect fraud with much better accuracy than even the most competent compliance teams. 

For example, the biggest bank in Denmark, Danske Bank, implemented a deep learning tool for fraud detection that was 50% more accurate at fraud detection.

The biggest near-future impact in the areas of fraud detection and anti-money laundering is likely to come from decision engines. Not only are decision engines fantastic automation tools, but they are also big data powerhouses capable of sending, storing, aggregating, and analyzing larger volumes of real-time data, sourced intelligently from a variety of sources (both traditional and non-traditional). 

These tools will help lenders design a logic engine that prevents lending fraud by performing pre and post-authorization tests. The engine will go through a checklist of “rules,” and if one of the rules you set is not met, the engine will “fail” the entire application. You can also design your system to weigh each “fail” according to predefined parameters before allowing you to engage in a final manual review. 

Onyx IQ:  A SaaS Platform for the Modern Lender

With so many benefits, it’s no surprise that SME lenders are refocusing their operational strategies to keep technology front and center. 

If you’re wondering whether it’s time for your alternative lending business to undergo a tech makeover, then be sure to check out our blog article here.

Onyx IQ leverages cutting-edge technology to bring more intelligent, faster, and fairer practices to your SME lending business. If you want to see what the lending software of the future is truly capable of, sign up to demo our platform today.