The Latest

19 May 2021: Google has launched Vertex AI, a platform that strives to accelerate the development of machine learning models (aka, algorithms). According to Google and IBRS discussions with early adopters, the platform does indeed dramatically reduce the amount of manual coding needed to develop (aka, train) machine learning models. 

Why it’s Important

The use of machine learning (ML) will have a dramatic impact on decision making support systems and automation over the next decade. For the majority of organisations, ML capabilities will be acquired as part of regular upgrades of enterprise SaaS solutions. Software leaders such as Microsoft, Salesforce, Adobe and even smaller ERP vendors such as Zoho and TechnologyOne, are all embedding ML powered services into their products today, and this will only accelerate.

However, developing proprietary ML models to meet specific needs may very well prove critically important for a few organisations. Recent examples of this include: customise direct customer outreach with specific language tailored to lessen overdue payment, and creating decision support solutions to reduce the occurrence of heatstroke.

IBRS has written extensively on ML development operations (MLOps). However, the future of this disciplin e will likely be AI-powered recommendation engines that aid data teams in the development of ML models. In a recent example, IBRS monitored a data scientist as they first developed an ML model to predict customer behaviour using traditional techniques, and then used a publicly available tool that leveraged ML itself to build, test and recommend the same model. Excluding data preparation, the hand-coded approach took 3 days to complete, while the assisted approach took several hours. But more importantly, the assisted approach tested more models that the data scientist could test manually, and delivered a model that was 3% more accurate than the hand-coded solution.

It should be noted that leveraging ‘low-code’ AI does not negate the need for data scientists or the pressing need to improve data literacy within most organisations. However, it has the potential to dramatically reduce the cost of developing and testing ML models, which lowers the financial risk for organisations experimenting with AI.

Who’s impacted

  • CIO
  • COO
  • CFO
  • Marketing leads
  • Development team leads

What’s Next?

Prepare for low-code AI to become increasingly common and the hype surrounding it to grow significant in the coming two years. However, the excitement for low-code ML should be tempered with the realisation that many of the use cases for ML will be embedded ‘out of the box’ in ERP, CRM, HCM, workforce management, and asset management SaaS solutions in the near future. Organisations should balance the ‘build it’ versus ‘wait for it’ decision when it comes to ML-power services. 

Related IBRS Advisory

  1. Six Critical Success Factors for Machine Learning Projects
  2. Options for Machine Learning-as-a-Service: The Big Four AIs Battle it Out
  3. How can AI reimagine your business processes?
  4. Low-Code Platform Feature Checklist
  5. VENDORiQ: BMC Adds AI to IT Operations
  6. Artificial intelligence Part 3: Preparing IT organisations for artificial intelligence deployment