Adobe Sensei

Adobe Sensei was the umbrella name that Adobe used for a shared framework, services, and product features based on artificial intelligence and machine learning. Adobe introduced the name at Adobe MAX on November 2, 2016 and applied it across Creative Cloud, Document Cloud, and what was then called Marketing Cloud, later Experience Cloud.^[1][2] Sensei was not one model or a standalone application. It covered capabilities as different as image selection, document-structure recognition, marketing attribution, anomaly detection, and personalization.

Adobe's current business site says that Sensei's capabilities have transitioned into its modern Adobe AI offerings. The company now foregrounds product names such as Adobe Firefly, Acrobat AI Assistant, and Adobe Experience Platform Agent Orchestrator.^[20][21] The transition was not a clean removal of the older name: current Adobe developer documentation still includes a Sensei Machine Learning API and describes PDF Extract as using Sensei AI technology.^[22][23] Adobe has not published a dated notice establishing that every Sensei service was discontinued.

Definition and scope

At launch, Adobe described Sensei as both a unified AI and machine-learning framework and a set of intelligent services. The framework was meant to provide shared infrastructure and access to Adobe's content, document, and customer-experience data, while the services addressed particular tasks in individual products.^[1][2] Adobe's fiscal 2016 annual report called Sensei a core part of Creative Cloud and used the same framework-and-services definition.^[3]

The name therefore operated at several levels:

Scroll sideways for more →

Adobe's descriptions changed with context. It called Sensei a platform in a 2017 retrospective, an AI engine in 2022 product announcements, and cross-platform technology in its fiscal 2024 Form 10-K.^[4][13][19] These labels did not turn Sensei into a single consumer product. A Photoshop user interacted with a particular selection or editing tool, while an Experience Cloud customer configured a scoring or personalization service. Developers could work with specific Adobe APIs, but Adobe did not offer a general Sensei chat interface or downloadable model under that name.

The 2016 announcement also mixed shipping features, existing services, and planned capabilities. Adobe said that dozens of Sensei services were already running, announced features such as Adobe Stock Visual Search, and listed document summarization among areas expected to receive further work.^[1] The list should not be read as a record that every described capability became generally available on launch day.

Development and chronology

Adobe's work in computational imaging and data analysis preceded the Sensei name. In 2009, researchers including Adobe's Dan Goldman published PatchMatch, a randomized algorithm for finding correspondences between image patches. Adobe later used that research in Photoshop's original Content-Aware Fill, released in 2010.^[5][6] Adobe also had predictive and anomaly-detection functions in its marketing products before 2016. Sensei gathered this older work and new AI investment under a common technical and commercial identity rather than starting Adobe's use of intelligent algorithms from zero.^[1][4]

Scroll sideways for more →

Creative Cloud integrations

Creative Cloud made Sensei visible through features embedded in established applications. The 2016 launch examples covered computer vision, audio analysis, and search. Adobe Stock Visual Search accepted an image as a search query. Face-Aware Liquify detected facial landmarks to expose relevant editing controls. Auto Lip Sync matched recorded speech to mouth shapes in Character Animator.^[1]

By 2017, Adobe also grouped Lightroom's automatic photo tagging and natural-language search, Typekit Visual Search, and Audition Remix under Sensei. These tools solved separate problems and did not share one user interface. Adobe's statement that more than 100 capabilities were in products or its pipeline combined released functions with work still under development.^[4]

Later releases expanded the label into video and more advanced image editing. Content-Aware Fill for video, released in After Effects in April 2019, analyzed motion and surrounding frames to replace a masked object. Premiere Pro's Auto Reframe, released that November, analyzed the action in footage when converting a sequence to square, vertical, widescreen, or custom formats. Adobe kept the generated keyframes editable rather than presenting the output as final.^[8][9]

Photoshop's October 2020 release placed several systems in the same AI context. Neural Filters provided machine-learning filters in a dedicated workspace. Sky Replacement separated foreground from sky, the Discover panel recommended tools and actions, and selection updates handled difficult boundaries such as hair.^[10] At MAX in 2021, Photoshop added automatic masks for detected objects and a Harmonization neural filter for matching the appearance of composited layers.^[12] In 2022, Adobe applied the Sensei label to Photo Restoration, Lightroom adaptive presets, masking, background removal, and other automation across Creative Cloud and Adobe Express.^[14]

These examples show why Sensei should not be treated as a model version series. Adobe could revise the underlying method of a feature without changing the Sensei label, and one product release could contain several unrelated models or algorithms.^[1][10]

Document Cloud integrations

The initial Document Cloud plan included natural language processing, document similarity, summarization, optical character recognition, semantic structure analysis, and reading-order extraction. Adobe marked some of these as capability areas for future services rather than a list of products available in November 2016.^[1]

Adobe Scan, released as a mobile application in June 2017, was an early shipping example. It used image processing to detect and clean up photographed pages and OCR to turn printed text into reusable text in a PDF.^[4] Liquid Mode followed in Acrobat Reader mobile in September 2020. It identified headings, paragraphs, lists, images, and tables, then reflowed supported documents for smaller screens while leaving the source PDF unchanged.^[11]

Document intelligence also became available to developers. Adobe's PDF Extract API returns document text, tables, figures, reading order, positions, and style information in structured output. Current documentation still calls it a cloud service that uses Sensei AI technology.^[23] The separate Sensei Machine Learning API within Adobe Experience Platform manages engines, model instances, training, scoring, and service deployment.^[22] These offerings support the interpretation of Sensei as a service layer, but they are distinct APIs with their own access rules and data formats.

Experience Cloud integrations

In the marketing and customer-experience business, Sensei referred less to visible editing tools and more to prediction, scoring, and decision services. The launch catalog included attribution, personalization, intelligent audience segmentation, anomaly detection, sentiment analysis, lookalike modeling, automated advertising insights, and alerts for unusual changes in customer behavior.^[1][2] Adobe described these functions as ways to analyze customer events and recommend an action. Those descriptions were product claims, not independent findings about the quality of the recommendations.

Adobe tied the next phase of this work to Adobe Experience Platform in 2019. Customer AI was designed to score behaviors such as engagement, retention, lifetime value, and likelihood of an upsell. Journey AI was designed to choose the timing and channel for messages across a customer's interactions.^[7] In 2022, Adobe announced further Sensei functions for product recommendations, live search, sales-opportunity prediction, budget forecasting, and marketing-mix modeling.^[13]

This part of Sensei depended on an organization's Experience Cloud data and configuration. It was not a public model trained on a fixed universal dataset. At launch, Adobe said that anonymized content and customer data contributed to its machine-learning systems. The company framed the scale and domain specificity of those data as competitive advantages.^[1][2] That account came from Adobe; the launch material did not provide an independent audit of datasets or model performance.

Technical scope and research lineage

Adobe never published one architecture that applied to every Sensei-branded feature. The umbrella included conventional image-processing methods, statistical models, deep learning, recommendation systems, language processing, and later generative models. The common thread was product integration and shared Adobe infrastructure, not a single neural network.^[1][19]

Content-Aware Fill illustrates the distinction. Adobe Research wrote in 2018 that the original feature did not use deep learning. It copied suitable patches from nearby image regions. The research prototype Deep Fill instead used a generative adversarial network trained on millions of photographs to synthesize missing content.^[6] Both belonged to Adobe's broader work on intelligent editing, but they had different training requirements and failure modes. Marketing material that calls a feature Sensei-powered does not, by itself, identify its exact algorithm.

Research projects also did not automatically become products. Adobe showed many experimental systems at its MAX and Summit conferences, while its 2017 count explicitly combined capabilities already in products with those still in the pipeline.^[4] A reliable product chronology therefore depends on release notes or documentation, not a conference demonstration alone.

The developer-facing framework provides a more concrete view of one Sensei implementation. The current Experience Platform API organizes machine-learning logic into engines, model instances, experiments, trained models, and deployable services. It supports packaged artifacts and batch or service scoring inside an organization's platform sandbox.^[22] This is enterprise infrastructure rather than the hidden implementation of every desktop feature Adobe ever labeled Sensei.

Relationship to Firefly and Sensei GenAI

Generative AI entered the Sensei naming structure in 2023, but Adobe used two related labels for different offerings. Firefly was introduced as a family of creative generative models, first for images and text effects. Adobe said Firefly would be integrated across its clouds and would be part of a series of new Sensei generative AI services.^[16] That statement placed Firefly within the broader AI program; it did not establish that every Firefly model ran on one pre-existing Sensei model.

Sensei GenAI was the Experience Cloud counterpart. Adobe announced it for Experience Manager, Journey Optimizer, Customer Journey Analytics, Marketo Engage, and Real-Time Customer Data Platform. Its early text functions included copy generation, rephrasing, summarization, natural-language queries, and conversational assistance.^[17][18] Adobe said these services could use multiple large language models, including Microsoft Azure OpenAI Service and FLAN-T5, selected according to the business use case.^[17] Sensei GenAI was therefore not simply another name for Firefly.

The distinction became clearer as Adobe built separate Firefly applications, model generations, APIs, and consumption plans. Firefly became the company's principal brand for creative generation, while Experience Cloud adopted names such as Experience Platform AI Assistant and Agent Orchestrator. Adobe's current materials describe the older Sensei work as a foundation and origin for these systems rather than maintaining it as the main customer-facing AI brand.^[20][24]

Governance and limitations

Adobe announced formal AI ethics principles in February 2021 after, it said, two years of internal development. The principles were responsibility, accountability, and transparency. Adobe also described an AI Ethics Committee and Review Board intended to examine new AI features, possible harmful bias, user controls, and mechanisms for reporting problems.^[15]

Later Adobe material says product teams submit AI impact assessments and that higher-risk features receive additional review, testing, filtering, or retraining before release.^[24] These statements document Adobe's governance process, not the outcome of an external evaluation. They also postdate Sensei's 2016 introduction, so the principles should not be projected backward as launch requirements.

Sensei's breadth creates another limitation for assessment. A result about one tool does not establish the accuracy, fairness, or privacy properties of another. A photo-selection model, an OCR service, and a customer-propensity score use different inputs and evaluation criteria. Adobe's corporate adoption figures and claims that its data improved contextual results should be read as attributed statements unless supported by product-specific testing.^[1][13]

Branding transition and current status

The strongest evidence for a branding change comes from Adobe itself. Its former Sensei product page, with metadata dated November 14, 2025, is now titled around Adobe AI and answers the question "What was Adobe Sensei?" in the past tense. It says that Sensei's AI and machine-learning features have transitioned into modern Adobe AI offerings and that capabilities originating in Sensei continue in Firefly and Experience Platform Agent Orchestrator.^[20]

Adobe's regulatory filings show a similar change in emphasis. The fiscal 2024 Form 10-K still named Sensei in Document Cloud and Experience Cloud and listed Sensei as a cloud-hosted AI framework. The fiscal 2025 filing instead discusses Firefly, Acrobat AI Assistant, Experience Platform AI Assistant, GenStudio, and agentic AI without naming Sensei.^[19][21]

The technical record is less final. Adobe Developer still publishes a Sensei Machine Learning API reference, and current PDF Services documentation retains Sensei in service descriptions.^[22][23] The available evidence supports saying that Adobe transitioned its main branding from Sensei to Adobe AI and named product families. It does not support assigning a universal end-of-service date or assuming that all code, APIs, and product features associated with Sensei were retired together.

References

1. ^Adobe Sensei: Mastering Content and Data. Adobe Blog, November 2, 2016.
2. ^Adobe Sensei Lets Customers Master the Art of Digital Experiences. Adobe, November 3, 2016.
3. ^Form 10-K for the fiscal year ended December 2, 2016. Adobe Systems Incorporated, January 20, 2017.
4. ^Mastering Creative and Experience Intelligence: A Year with Adobe Sensei. Adobe Blog, November 2, 2017.
5. ^PatchMatch: A Randomized Correspondence Algorithm for Structural Image Editing. Connelly Barnes, Eli Shechtman, Adam Finkelstein, and Dan B. Goldman, ACM Transactions on Graphics 28(3), August 2009.
6. ^Leveraging Deep Learning to Fix Images. Adobe Research, February 8, 2018.
7. ^The Next Chapter of Adobe Sensei. Adobe Blog, March 26, 2019.
8. ^Everything You Need to Know About the April 2019 Releases. Adobe Blog, April 2, 2019.
9. ^Faster, Smarter, More Creative. Adobe Blog, November 4, 2019.
10. ^Photoshop: Now the world's most advanced AI application for creatives. Adobe Blog, October 2020.
11. ^Adobe unveils ambitious multi-year vision for PDF: Introduces Liquid Mode. Adobe Blog, September 23, 2020.
12. ^Photoshop ships major updates across desktop and iPad apps. Adobe Blog, October 26, 2021.
13. ^Adobe Unveils New AI Innovations for Personalizing Experiences. Adobe News, March 15, 2022.
14. ^Adobe AI Innovation Maximizes Creative Power and Precision for Creative Cloud and Adobe Express Users. Adobe News, October 2022.
15. ^Adobe unveils new AI ethics principles as part of commitment to responsible digital citizenship. Adobe Blog, February 17, 2021.
16. ^Adobe Unveils Firefly, a Family of New Creative Generative AI Models. Adobe News, March 21, 2023.
17. ^Adobe Announces New Sensei GenAI Services to Reimagine End-to-End Marketing Workflows. Adobe News, March 21, 2023.
18. ^Adobe Adds Powerful New Sensei GenAI Services Across Experience Cloud. Adobe News, October 12, 2023.
19. ^Form 10-K for the fiscal year ended November 29, 2024. Adobe Inc., filed January 13, 2025.
20. ^From Sensei to AI. Adobe for Business, page metadata dated November 14, 2025, accessed July 24, 2026.
21. ^Form 10-K for the fiscal year ended November 28, 2025. Adobe Inc., filed January 15, 2026.
22. ^Sensei Machine Learning API. Adobe Developer, accessed July 24, 2026.
23. ^PDF Extract API. Adobe Developer, accessed July 24, 2026.
24. ^Generative AI: Built for Business. Adobe, accessed July 24, 2026.