Enterprises need more artificial intelligence and machine-learning (ML) solutions to drive value, transform their businesses, and outperform the competition. But firms find it challenging to navigate the lifecycle of developing, deploying, and maintaining their ML models and AI solutions. A key problem? They don’t have the right PAML (predictive analytics and machine learning) solutions that make it possible to scale AI — in a rapid, reliable, repeatable, and governable fashion — across the organization.

Pursue The Right Power PAML Tools For Your People And Purposes

Thankfully, there is a growing landscape of vendors offering PAML solutions designed to help enterprises rapidly develop custom AI and ML solutions and push them beyond proof-of-concept (PoC) purgatory to full-scale production. In our recently published report, “Now Tech: Predictive Analytics And Machine Learning, Q2 2020,” we’ve identified and researched the 37 major PAML vendors and categorized them into three segments based on their capabilities:

  • Multimodal PAML. These solutions appeal to the broadest audience by offering a diverse set of methods to develop and operationalize models. They have rich visual tools to build data and machine-learning pipelines, and they often have notebook and/or automated machine-learning (AutoML) capabilities, as well. Multimodal PAML is a great option for enterprises requiring a wide range of tools and looking to further collaboration between teams of data scientists and non-data scientists.
  • Notebook-based PAML. Vendors in this segment specialize in tools designed for coders who use programming languages such as Python and/or R to build, deploy, and manage models. These solutions are commonly built around one or more open source notebooks. Notebook-based PAML is a logical choice for enterprises with an existing pool of, or easy access to, coding talent, as well as firms interested in piggybacking off open source innovations.
  • Automation-focused PAML. Automation-focused solutions already help expedite major stages of model development like feature engineering and model training. Their coverage of certain “forgotten” aspects of ML model deployment, including data prep and model management, also continues to grow. While some multimodal and notebook-based PAML vendors tout AutoML functionality, solutions in this segmentation are differentiating by their singular focus on AutoML. Companies interested in accelerating the pilot-to-production journey should engage with vendors in this segmentation.

Presume Precipitous PAML Progress

Selecting the right PAML solution is a challenge for every large organization. Most PAML vendors have been rapidly developing extensive new capabilities, spanning new parts of the AI and ML lifecycle. Despite a convergence of capabilities in this market, PAML purchasers will still need to explore multiple solutions to meet their firm’s unique requirements. Furthermore, none of these vendors offer a universal solution to all your AI needs. Keep reminding the business stakeholders and decision makers (and yourself) that the goal of PAML is to reduce the time and cost of developing and operationalizing your AI and ML models by boosting productivity and efficiency — and that can mean regularly evaluating and replacing your PAML solutions.

Mike Gualtieri and I are evaluating vendors from the multimodal PAML and notebook-based PAML segmentations for two Forrester Wave™ evaluations that will be published in mid-Q3 of 2020. Forrester’s New Wave™ of nine automation-focused PAML vendors, published in Q2 2019, can be found here. And you can learn about ways to scale AI with automated machine learning in this report here. As always, we’re happy to discuss this research on an inquiry or learn about new PAML solutions via an analyst briefing.

(Chandler Hennig, senior research associate, coauthored this post.)

Note: A Forrester subscription is required to access the research featured in this post.