Report:

Magic Quadrant for Cloud AI Developer Services

How does Gartner define the Cloud AI Developer Services market in 2023?

Gartner defines cloud AI developer services (CAIDS) as cloud-hosted or containerized services that enable software developers who are not data science experts to use artificial intelligence (AI) models via APIs, software development kits (SDKs) or applications. Core capabilities include automated machine learning (autoML), automated data preparation, feature engineering, automated model building and model management. Optional and important complementary capabilities include language and vision services such as sentiment analysis and image generation.

Key Facts for Magic Quadrant for Cloud AI Developer Services in 2023

• Publication Date: 22-May-2023
• Document ID: G00773391
• Coverage: As of March 2023
• Authors: Jim Scheibmeir, Svetlana Sicular
• Core Purpose: This Magic Quadrant evaluates cloud AI developer services (CAIDS) vendors that enable software developers who are not data science experts to use artificial intelligence models via APIs, SDKs or applications. It assesses vendors' ability to deliver autoML, language, and vision services that help software engineering teams build predictive and intelligent application capabilities.

Strategic Planning Assumptions

• By 2025, the AI software market is expected to reach $134.8 billion in revenue at a compound annual growth rate of 29.2%
• Software engineering leaders face a primary barrier of lacking AI/ML skills within their teams
• Organizations with AI privacy, security and risk taskforces convert a higher proportion of AI proofs of concept into production
• CAIDS vendors are expanding libraries of prebuilt ML models for vertical solutions
• Generative AI capabilities are becoming a key differentiator among CAIDS vendors
• Model portfolio management and responsible AI are increasingly important vendor capabilities

How was the Cloud AI Developer Services market evolved in 2023?

• Gartner inquiries about generative AI have increased by over 1,300% year over year
• AWS, Microsoft and Google are the most popular CAIDS vendors discussed in social media
• AI was the leading technology that survey participants had not yet deployed but expected to be especially important in the next five years
• Only 13% of organizations have already deployed responsible AI capabilities
• Lack of skills is the largest identified barrier to AI adoption, surpassing funding, legal, and ROI concerns
• CAIDS vendors are heavily investing in responsible AI components in response to customer expectations
• Vendors are expanding from ML model performance to focus on explainability, transparency and bias mitigation
• Major cloud vendors are using immense computing infrastructure to develop proprietary language models
• China-based vendors have shown immense improvements in computer vision capabilities
• Most CAIDS vendors have improved vision capabilities over the past year

What product features are required to be included in this year's evaluation?

• Automated Data Preparation — These services can prepare datasets to be used for training the models. These services can cleanse and augment datasets from the raw data that is provided by the organization and can include data visualization.
• Feature Engineering — These services automatically evaluate data and add metadata to a dataset that is submitted to the provider. This augmentation can be done via crowdsourcing capabilities or ML models. The feature may also allow for automated detection and classification of features and generation of new features from existing ones.
• Automated Model Building — These models require developers to provide datasets that can be used to train the models, as well as metadata tags that have the attributes they want the models to be trained to identify. They must also identify the variable that they want to model to predict. These services analyze the data and then recommend or select the potential best algorithms that can be used to build and optimize the model. The service automatically optimizes the performance and accuracy of the model by tuning the hyperparameters of the model. The service may also blend algorithms to optimize model performance.
• Model Management/Operationalization — These services may offer model factory/orchestration functionality to automate the building of ML pipelines, including model training, deployment, monitoring and management of models in production. The service may also provide for packaging and provisioning of infrastructure for the models created, such as Kubernetes containers. The services should also automate creation of APIs needed to access the models. This capability allows developers to monitor models in test and production environments to assess model performance and model drift. This service may provide automated retraining of the model and provision for replacing an outdated model with a better one. Additional features may include business key performance indicators (KPIs) for model value and the ability to assess the quality of data impacting the model performance.
• Responsible AI — These services analyze datasets for potential bias, explainability and interpretability

What are the common features of top products in the Cloud AI Developer Services space?

• Natural Language Processing/Understanding — This service includes functions such as tokenization, part-of-speech (POS) tagging, stemming, term frequency-inverse document frequency (TF-IDF) and other text processing functions. Natural language understanding is a subset of natural language processing that deals with machine comprehension. It takes textual input and extracts metadata from the text. Extracting metadata is straightforward, but understanding the intent of the person entering the text is challenging and often requires supplemental models.
• Speech to Text — This service is a subset of computational linguistics that converts analog input to text output. This text output can be the final product, or it can be entered into a natural language understanding model so that metadata can be extracted. Many computing devices, such as personal computers and smartphones, have a built-in automated speech recognition (ASR) capability.
• Natural Language Generation — This service creates natural language from a machine representation, such as concepts, datasets or minimal descriptions in a knowledge base; or from a logical form, such as a return form that generates a letter to the customer. The service essentially translates data to natural language text.
• Text to Speech — This service converts textual input into analog output/speech.
• Translation — This service takes text input from the source language and converts it to a target language. This is a challenging task because different languages have distinct structures. It is not as simple as translating a word from one language to the corresponding word in another language.
• Sentiment Analysis — This service analyzes language for positive, negative or neutral sentiment based on the words that the developer enters into a conversation or social model. These services may be supplemented by other services that analyze the tone of analog inputs.
• Text Analytics — This service analyzes unstructured text to extract elements such as concepts, topics and keyword attributes, and it adds these as metadata.
• Image Recognition/Labeling — This service normally identifies what objects or people are contained in an image. Some implementations can also identify other attributes in the image, such as colors or patterns. This service enables developers to identify whether people or other items of interest are in an image, to add metadata for classifying or tagging the images, and to submit image datasets for labeling.
• Video AI — This service normally combines image recognition and ASR to identify people and other objects in a video and to create a transcript for the audio. Some services also track the movement of people across multiple frames of the video.
• ML-Enabled Optical Character Recognition (OCR) — This service converts electronic images of typed, handwritten, printed text, or text in images or video into machine-encoded text and adds metadata to the content. The service also uses ML to classify the information in a given field based on its content.
• Image Generation — This service constructs images similar to the given image set or based on a word prompt. Synthetic data is a popular use of image generation.

Scope Exclusions

• Vendors that offer only language services or vision services
• Vendors that specialize only in specific use cases
• Vendors that did not include autoML services in their CAIDS offering
• Vendors that offered CAIDS only as part of a professional services contract, where the services are used exclusively by the vendor's consultants
• Vendors that offered services that were not native services created and delivered by the vendor

Inclusion Criteria

Vendors must, among other requirements:

• Demonstrate a go-to-market strategy that focuses on the persona of a professional software developer
• Include all the core capabilities of a CAIDS platform as per the Market Definition
• Have generated at least $15 million in revenue from CAIDS offerings (excluding professional services revenue) in 2022
• Have at least 75 current paying enterprise customers for CAIDS offerings

Ability to Execute — Relative Weighting

• Product or Service - High
• Overall Viability - Medium
• Sales Execution/Pricing - Medium
• Market Responsiveness/Record - Medium
• Marketing Execution - Medium
• Customer Experience - Medium
• Operations - Medium

Completeness of Vision — Relative Weighting

• Market Understanding - Medium
• Marketing Strategy - Medium
• Sales Strategy - Medium
• Offering (Product) Strategy - High
• Business Model - Medium
• Vertical/Industry Strategy - High
• Innovation - High
• Geographic Strategy - Medium

Reference

• Gartner, Magic Quadrant for Cloud AI Developer Services, 22-May-2023, ID G00773391

View Leaders
View Vendor Movements