Artificial intelligence (AI) has rapidly transitioned from an emerging technology to a mission-critical capability for businesses across industries. According to McKinsey, AI could deliver up to $5.8 trillion in annual value for companies worldwide. And a recent survey by IBM found that 43% of companies accelerated their AI adoption plans due to the COVID-19 pandemic.
However, building, deploying, and maintaining AI systems is complex and resource-intensive. It requires scarce data science talent, massive amounts of data, and ongoing optimization and tuning. That‘s where AI platforms come in. These software tools provide pre-built algorithms, APIs, and infrastructure to make AI more accessible and scalable for companies of all sizes.
But with dozens of AI platforms on the market—each with different strengths, specialties, and tradeoffs—it can be overwhelming for business and technology leaders to determine which ones are the best fit for their needs. As an AI consultant and practitioner who has evaluated and implemented many of these platforms firsthand, I created this comprehensive guide to help you navigate this complex landscape.
Criteria for the Top AI Platforms
To select and rank the top AI platforms, I evaluated dozens of offerings across the following key criteria:
- Functionality: The breadth and depth of the platform‘s AI capabilities, including pre-built models, AutoML features, and flexibility to customize models
- Ease of use: How intuitive the platform‘s interface and tools are, and how much technical expertise is required to utilize them effectively
- Performance: The platform‘s training and inference speeds, scalability, and ability to handle large datasets and complex models
- Integration: How well the platform integrates with popular data sources, business applications, and development tools
- Cost: The platform‘s pricing model, TCO, and ROI potential based on common use cases and customer case studies
I also factored in the platform‘s market traction, customer feedback, analyst reviews, and my personal experiences. Finally, I looked for platforms that offered differentiating features and strong positions in specific industries or use cases in addition to the leading general-purpose offerings.
The 20 Best AI Platforms
Based on this analysis, here are my picks for the 20 best AI platforms available today:
-
Google Cloud AI Platform: Google‘s comprehensive platform offers both pre-trained APIs for vision, language, and structured data, and tools for building custom models. Key differentiators include access to Google‘s state-of-the-art ML hardware (TPUs); end-to-end ML pipelines; built-in dataset search and labeling; and AutoML Tables for structured data.
-
Microsoft Azure AI: Azure offers a broad range of AI tools, from pre-built APIs to the Azure Machine Learning Studio for no-code model development. Microsoft also has strong offerings for specific verticals like healthcare. Key differentiators include the breadth of Azure ecosystem integrations and the ability to containerize models for edge deployment.
-
Amazon SageMaker: SageMaker is a fully managed machine learning platform that enables data scientists and developers to build, train, and deploy ML models at scale. It offers built-in algorithms, one-click training, automated hyperparameter tuning, and flexible deployment options. Uniquely, SageMaker integrates with Amazon‘s broad AI/ML stack.
-
IBM Watson Studio: Watson Studio provides a comprehensive environment for data scientists, developers and domain experts to collaboratively build, run and manage AI models at scale. Key differentiators include IBM‘s proprietary AutoAI technology; strong visual modeling tools; and extensive built-in libraries for NLP, computer vision, and time series forecasting.
-
H2O Driverless AI: H2O‘s Driverless AI platform automates key machine learning workflows including feature engineering, model tuning, and deployment. It uniquely offers "reason codes" that provide transparency into how models make predictions. H2O also has strong data connectors and excels in financial services use cases.
-
DataRobot: DataRobot offers an end-to-end platform for data prep, machine learning, and MLOps. The platform supports supervised and unsupervised learning and excels in its breadth of algorithms and built-in guardrails. DataRobot also enables "human-in-the-loop" decision support and is a strong choice for business analysts.
-
Dataiku DSS: Dataiku DSS is a collaborative platform for prototyping, building, and delivering AI and analytics projects. It offers visual tools for data prep, model training, and deployment. Dataiku differentiates itself with strong governance features, the ability to integrate with a variety of ML tools, and extensibility via coding and APIs.
-
Salesforce Einstein: Einstein embeds AI capabilities into Salesforce‘s market-leading CRM platform. This includes real-time insights, recommendations, and forecasting to empower sales, marketing, and service teams. Einstein uniquely enables business users to build low-code AI apps via templates and "point-and-click data science."
-
Petuum: Petuum offers an end-to-end platform for building and operating industrial AI solutions. The platform is optimized for complex, large-scale data and excels in its infrastructure innovations that enable distributed training and deployment. Petuum is a fit for advanced developers building mission-critical AI apps.
-
CognitiveScale Cortex: CognitiveScale offers a platform for rapidly designing, developing and deploying enterprise-grade AI systems. It provides a collaborative AI Notebook, Automated Machine Learning, and pre-built solution accelerators. CognitiveScale uniquely offers "AI guardrails" to help ensure models are fair and unbiased.
-
Determined AI: Determined offers a deep learning training platform that enables teams to train models faster. It supports any ML framework, algorithm, or infrastructure and offers advanced hyperparameter tuning, distributed training, and experiment tracking. Determined is a fit for organizations with large-scale AI initiatives.
-
C3 AI Ex Machina: C3‘s Ex Machina platform enables anyone to develop, scale, and apply AI insights without writing code. It offers an intuitive UI, AutoML capabilities, and a large model library for common enterprise use cases. C3 also provides strong industry-specific templates and excels in manufacturing and energy.
-
TensorFlow Enterprise: TensorFlow Enterprise brings Google‘s popular open-source ML platform into an enterprise-grade offering. It enables easy model training, collaborative development, and production deployment. Key benefits include long-term version support and security patches, cloud-specific optimizations, and prioritized customer support.
-
Cloudera Data Science Workbench: Cloudera‘s data science platform enables teams to collaboratively build and scale ML and analytics workflows. It offers Jupyter-style notebooks, built-in frameworks, and tools for model deployment and monitoring. Cloudera is known for its strong capabilities around massive, complex datasets.
-
Domino Data Lab: Domino provides an end-to-end platform for code-first data science. It offers environment management, reproducibility, collaboration, and one-click deployment and integrates with popular IDEs. Domino is well-suited for expert data science teams who want flexibility and control.
-
Databricks Lakehouse Platform: Databricks‘ unified data platform supports the full ML lifecycle from data preparation to model training and deployment. It offers collaborative notebooks, automated cluster management, and a large library of pre-built models. Databricks is known for its strong performance and ability to handle massive data volumes.
-
Noodle.ai: Noodle.ai offers a suite of pre-built AI applications for the manufacturing and supply chain industries. The applications leverage Noodle‘s Enterprise AI platform which uniquely combines data preparation, AutoML, and business workflows. Noodle is differentiated by its proprietary algorithms and strong industry expertise.
-
Veritone aiWARE: Veritone‘s aiWARE platform enables organizations to rapidly deploy and integrate AI-powered solutions. It offers a library of hundreds of best-of-breed AI models for tasks like transcription, object recognition, and content classification. aiWARE uniquely supports both structured and unstructured data.
-
Aible: Aible offers an AutoML platform that optimizes AI models for specific business goals and constraints. The platform uniquely enables business users to iteratively guide model creation to align with real-world tradeoffs. Aible also provides strong model monitoring and continuous improvement capabilities.
-
Ada: Ada offers an AI-powered platform for automating customer interactions. It can understand and respond to customer questions, personalize content, and seamlessly handoff to live agents. Ada‘s key differentiators are its multilingual support, ability to rapidly train industry-specific models, and strong integration and security.
Getting Started with AI Platforms
With this overview of the top AI platforms in mind, here are a few recommendations as you evaluate options for your organization:
- Start with the business problem: Before diving into any platform, get crystal clear on the business outcomes you‘re trying to drive with AI. Engage stakeholders to prioritize use cases based on feasibility and impact.
- Assess your data and talent: Evaluate the quality, quantity, and accessibility of the data you‘ll need for your AI applications. Also, consider whether you have the right talent in place to implement and scale the platform, and where you may need to supplement via hiring or partnering.
- Take advantage of free trials: Most leading AI platforms offer free trials or proof-of-concept projects. Take advantage of these to get hands-on experience with the tools and gauge their fit for your needs and environment.
- Don‘t neglect change management: Driving adoption of AI platforms requires significant change management, communication, and training. Be sure to put a change strategy in place to drive awareness and uptake of the platform and address any concerns around job impact.
- Plan for governance from the start: AI projects can introduce significant risks around data privacy, security, ethics, and transparency. Define clear governance processes upfront to mitigate these risks and put ongoing model monitoring and oversight in place.
The Future of AI Platforms
The AI platform market is rapidly evolving. On the near-term horizon, I predict we‘ll see the following developments:
- Continued consolidation: The market is currently fragmented with both a bevy of startups and the tech giants building broad AI platforms. I expect the well-capitalized players to continue to acquire and consolidate the market to offer more comprehensive, end-to-end offerings.
- Deeper verticalization: While the general-purpose AI platforms will continue to expand, I expect many of them to build deeper vertical offerings to compete with the industry-specific AI vendors. Vertical solutions will get more turnkey and accessible to business users.
- AI-centric data prep: Data preparation and feature engineering are consistently cited as the most time-consuming parts of AI projects. Expect more platforms to bake in intelligent automation and recommendations for data prep, leveraging techniques like weak supervision and augmentation.
- Rise of intelligent applications: Currently, most AI platform adopters build models to enhance existing applications and processes. Increasingly, I expect organizations to develop applications that are AI-first from the ground up—with AI/ML as the core.
- Continued abstraction and democratization: While many AI platform vendors tout drag-and-drop interfaces and AutoML, most still require significant technical chops to utilize. Expect more platforms to offer truly no-code development that enables business users to build and deploy models.
- Transformer-based models: Transformer-based language models like BERT and GPT-3 are rapidly expanding the possibilities of natural language AI. I expect leading platforms to incorporate these models into more turnkey offerings for translation, text generation, Q&A, and more.
- AI marketplaces: As AI becomes more ubiquitous, I predict the growth of marketplaces where practitioners can buy, sell, and share customized models and datasets. This will accelerate AI development and allow for more niche solutions.
These are just a few of the innovations I expect to shape the future of AI platforms. As Gartner predicts, AI augmentation alone will create $2.9 trillion of business value in 2021. And by 2025, IDC forecasts that 75% of organizations will shift from piloting to operationalizing AI, driving a 5X increase in streaming data and analytics infrastructures.
In other words, AI capabilities are rapidly becoming table stakes. Businesses that fail to harness these platforms will get left behind. But in the race to adopt AI, it‘s critical to be intentional in selecting platforms that align with your business needs, technology environment, and talent model. By starting small, continuously measuring value, and planning for scale, you can harness the transformative power of AI to drive efficiency, growth, and competitive advantage.
What AI platforms are you using or evaluating? Do you agree with this assessment of the top platforms? What challenges and opportunities do you see in operationalizing AI? Leave a comment with your thoughts!
