3 ways SMBs use machine learning to power digital transformation Posted on : May 01 - 2020

Companies leverage AI for marketing, document management, and contract lifecycle management.

Businesses should start small and fail fast with machine learning (ML) projects to get the best ROI. Some of the most common use cases for small- to medium-sized businesses (SMBs) include fraud detection, sales optimization, marketing, and document analysis.

But the benefits of implementing ML goes even further.

Steve Tycast, director of data and analytics at AIM Consulting, said ML efforts focused on operational analytics can reduce costs, drive efficiencies, and increase speed to market. And Craig Kelly, vice president of analytics at Syntax, said SMBs can use ML to anticipate the short-term and long-term impact on sales and adjust the strategy accordingly.

Further, the global impact of the recent coronavirus is something that is driving trends in ML usage.

"A specific example of this would be companies that manufacture PPE can better anticipate fluctuations in demand, and understand specifically where limited materials will be needed most," Kelly said.

Where else are SMBs using machine learning and artificial intelligence (AI)? Here are three more examples of current trends.

Marketing

Kristina Conely, director of data and analytics at AIM Consulting, said AIM worked with a hotel company to improve its marketing programs using machine learning. At the start of the project the company had no centralized data repository, and all the reporting was manual. The first goal was to analyze products and services sold in a particular region and look for new marketing pitches. The company owns hotels and wanted to automate the process of recommending upgrades and add-on experiences for guests. To accomplish this, AIM helped the company create multiple machine learning models.

Conely said she and the client's marketing team refined the algorithm over the course of three weeks to make sure the team trusted the results.

"At first the response was: 'These two things will never sell together.' But as they saw results they came on board," she said. "They also realized that now they had time to do the analyst-type role they were brought on for."

Tycast said part of the process of implementing ML is helping clients understand how to create and modify the algorithms.

"You can underfit and overfit models. Sometimes if you remove lots of variables you can gain very high accuracy scores," he said. View More