DAM & Content Intelligence Glossary

A three-dimensional model with geometry, materials and textures, in formats such as glTF/GLB, USD/USDZ, FBX or OBJ. 3D assets can render any angle or configuration of a product without a photo shoot.

ExampleA furniture brand renders every fabric option from one 3D model instead of photographing 40 variants.

A durable, distinct identity for each agent, with credentials scoped more narrowly than a person's, instead of a shared service account. It makes agent actions attributable and revocable.

ExampleDisabling the 'tagging-agent' identity stops it everywhere without touching other automations.

Coordinating several agents and deterministic steps: what runs in which order, what passes between them, how failures are handled and when a person is involved.

ExampleAn orchestrator runs enrichment, then compliance; if compliance fails, it routes to a reviewer instead of publishing.

Agents are usually scoped by job. Enrichment agents describe content. Compliance agents check rights, claims and brand rules. Transformation agents create renditions, crops and variants. Orchestration agents coordinate others across a process.

ExampleSeparating jobs keeps permissions narrow: the enrichment agent can write tags but cannot publish.

Systems in which models plan and act over several steps with some autonomy, using tools, memory and feedback. In practice, what matters is which actions it may take on its own and how those actions are controlled.

ExampleAn assistant that drafts a recommendation is helpful; an agent that unpublishes assets needs the same controls as a person with that permission.

A workflow where one or more steps are executed by agents, with defined triggers, allowed actions, review points and records. It extends rule-based automation to steps that require judgment.

ExampleOn upload: an agent tags, a compliance agent checks claims, a person approves exceptions, the asset publishes.

Software that uses a model to pursue a goal by deciding on steps and calling tools to carry them out, rather than producing a single response. Its reach is defined by the tools and permissions it is given.

ExampleAn enrichment agent receives 'tag new uploads for the spring launch', reads each asset, writes fields and reports what it changed.

Applying machine learning models to tasks inside a DAM: describing content, finding it by meaning, detecting risk, generating variants and predicting performance. Value depends on how outputs are reviewed, stored and governed.

ExampleEvery upload receives AI-generated tags and a description, marked as machine-generated until confirmed.

A property on the asset record saying whether and how AI generated or altered it. It must persist through derivatives, crops, resizes and republication, or disclosure obligations break downstream.

ExampleA crop of an AI-extended image inherits the flag and the C2PA manifest, so the social post is disclosed correctly.

A defined way for software to request data or actions from another system. A DAM API lets other tools search, read, create and update assets without a person using the interface.

ExampleThe e-commerce platform calls the DAM API to fetch all approved images for a new SKU.

Assets retained for reference, legal or historical reasons but removed from active use. Archived assets stay searchable for authorized users and cannot be distributed.

ExampleA 1990s campaign is archived: the brand team can study it, retailers cannot download it.

The status values an asset passes through, typically draft, in review, approved, published, expired and archived. Each state decides who can see the asset and where it may go.

ExampleAn asset moves from 'published' to 'expired' automatically when its rights window closes.

Recorded links between assets: master and derivative, native file and export, video and its captions, layout and the images placed in it. Relationships let changes and restrictions propagate.

ExampleWhen an image's license ends, the DAM finds every layout and crop derived from it.

The share of assets that are actually used, and how often, in a period. Low utilization signals overproduction, poor findability or assets that do not fit needs.

ExampleA library where only 22% of approved product shots appeared in any channel last quarter is producing more than it uses.

Recording both the agent that acted and the person or rule it acted for, as separate facts.

ExampleThe audit entry reads: compliance-agent, on behalf of A. Rossi, rule 'claims-check v4', changed status to 'blocked'.

A tamper-resistant record of who or what did what to an asset, when, and under which rule or permission. It answers compliance questions and makes errors traceable.

ExampleAn auditor asks who approved a claim in March; the trail shows the reviewer, version and timestamp.

Generating keywords and descriptions with AI when assets are added. Useful tags are mapped to the organization's controlled vocabulary and carry a confidence score; raw model labels alone produce noise.

ExampleThe model's 'footwear' maps to the taxonomy node 'Shoes > Running'; low-confidence tags are queued for review.

A graded description of how much an AI system may do without a person: suggest only, act with approval, act and notify, act silently within limits. Autonomy should be set per action type, not per agent.

ExampleThe same agent auto-applies tags, requests approval for rights changes and is never allowed to delete.

Keeping data consistent in two systems that both read and write it, with rules for which system wins each field. Without field ownership, sync loops and overwrites follow.

ExampleProduct attributes flow PIM to DAM; image URLs flow DAM to PIM; neither overwrites the other's fields.

The scope of harm a single automated error could cause: how many assets, channels or partners one wrong action reaches. Limiting it is a design choice, through scoping, batch sizes, staging and approvals.

ExampleA new rule first runs on 50 assets in a test collection before it is allowed on the full library.

A DAM scoped to brand materials: logos, guidelines, templates and approved imagery, usually with a self-service portal for internal teams and partners. It prioritizes consistency of use over production depth.

ExampleFranchisees download the current logo lock-ups and point-of-sale templates instead of copying an old version from email.

Automated review of an asset against brand rules: logo use, colors, typography, mandatory legal lines, prohibited imagery. AI can flag likely violations and explain them; people decide edge cases.

ExampleA partner-made banner is flagged for an outdated logo and a missing legal line before it goes live.

A curated, self-service view of a DAM for a specific audience, such as agencies, retailers, press or franchisees, showing only what they are allowed to use with simple download options.

ExamplePress receives a portal with approved launch images and logos; the internal team keeps its full workspace.

Exposure from off-brand, outdated, unlicensed or non-compliant content reaching audiences. Detecting it means looking at what is live in channels, not only at what is approved in the DAM.

ExampleA scan finds an old logo still live on three retailer sites after a rebrand.

A coalition, and the open technical specification it publishes, for attaching cryptographically signed provenance information to media: who created it, with which tools, and what edits followed, including AI generation. It was founded in 2021 by companies including Adobe, Microsoft, BBC, Intel, Arm and Truepic.

ExampleAn image exported with a C2PA manifest lets anyone verify it came from the brand and see that generative fill was used.

Unprocessed sensor data from a digital camera, in formats such as CR3, NEF, ARW or DNG. It retains far more tonal range than a JPEG and needs processing before use.

ExampleRetouchers start from RAW files; marketers only ever see the processed, approved derivatives.

Any destination where content reaches an audience: website, app, marketplace, social platform, email, retail media, store, print. Omnichannel means consistent content and governance across all of them at once.

ExampleThe same product video runs on the brand site, Amazon and TikTok, each in its own format and rights scope.

Verifying policy at the moment content leaves the governed platform for a channel: rights, embargo, AI disclosure, market rules. It closes the gap between correct records and incorrect publishing.

ExampleAn asset with an expired rights window is blocked at the hand-off to the social scheduler, even though someone queued it last week.

How content performs by destination: which formats, styles and messages work on each channel. The same asset can win on one channel and fail on another.

ExampleLifestyle imagery outperforms packshots on social; the reverse holds on marketplaces.

Tracking every product or marketing claim (for example 'clinically tested', '30% less sugar'), its evidence and where it is used. When a claim is withdrawn, every asset carrying it can be found.

ExampleA reformulation removes a nutrition claim; the DAM lists 212 assets and 9 retailers to update.

A DAM delivered as a hosted service rather than installed on the customer's servers. The vendor runs the infrastructure, scaling, backups and updates; the customer configures metadata, workflows and integrations.

ExampleA team moving off an on-premises archive gains browser access for agencies and removes server maintenance from IT's backlog.

A codec compresses and decompresses video or audio (H.264, H.265/HEVC, ProRes, AV1). A container wraps encoded streams and metadata into one file (MP4, MOV, MXF). An MP4 file can hold different codecs, which is why 'MP4' alone does not guarantee playback.

ExampleMasters are archived as ProRes in MOV; the DAM transcodes H.264 MP4 renditions for web and social.

A saved group of assets assembled for a purpose, such as a campaign, a pitch or a hand-off, without moving the assets. One asset can sit in many collections.

ExampleAn agency receives a link to the 'Summer 2027 hero' collection, which updates as assets are approved.

Data embedded in a file that tells software how to interpret its color values, for example sRGB for screens or a CMYK profile for a specific printing press. Without it, the same file can look different on every device.

ExampleConverting packaging artwork to the printer's CMYK profile before release avoids a shifted brand color on shelf.

Assembling a technology stack from specialized, replaceable services connected by APIs instead of one suite. Each component can be swapped without rebuilding the rest.

ExampleA brand replaces its search provider without migrating its DAM or CMS.

The field of AI that extracts information from images and video: objects, scenes, text, faces, logos, colors, actions. It turns pixels into data a DAM can search and govern.

ExampleVision models detect that an image contains a bottle, a beach and the brand's logo.

A score a model reports alongside an output, typically between 0 and 1. It is often not a calibrated probability, so thresholds are set by testing against labeled examples. Thresholds decide whether an output is applied automatically, sent for review or discarded.

ExampleTags above 0.9 are applied; 0.6 to 0.9 go to a librarian; below 0.6 are dropped.

A prebuilt integration between two systems that maps their data and actions, so they can exchange information without custom code. Connectors package authentication, field mapping and error handling.

ExampleA connector syncs SKUs and attributes from the PIM into DAM metadata every hour.

Linking a business outcome to the content that contributed to it. It requires stable asset identifiers that survive renditions and distribution, so channel data can be joined to the right asset.

ExampleA product video's identifier is carried into ad platform reports, so its spend and sales return to the DAM record.

The user-facing name for C2PA provenance data, shown as a small 'cr' mark that reveals the asset's origin and edit history when inspected.

ExampleA retailer checks the credential on a supplied image before using it in an ad.

A distributed network of servers that caches files close to users, cutting load times and origin traffic. Assets published from a DAM to websites are typically served through one.

ExampleA homepage hero loads from a server in Frankfurt for German visitors and from Singapore for Asian ones.

The policies, roles and controls that decide what content may exist, who may change it, how it must be described and when it must be retired. Effective governance is enforced in the system through required fields, permissions and rules.

ExampleUploads missing rights fields are rejected, so the policy holds even on a busy Friday.

Using AI and data to understand what content an organization has, how it is used and how it performs, across systems and channels. It connects the asset record to outcomes so decisions about creating, reusing or retiring content rest on evidence.

ExampleFor example, a team might see that 30% of last year's campaign imagery was never published and that one style drove most conversions.

A complete, current view of content across all systems where it lives, including DAMs, drives, CMSs and channels, not only one repository. It is the starting point for intelligence, migration and governance.

ExampleConnecting three DAMs and two drives might reveal that a fifth of assets exist in more than one system.

Software for building, editing and publishing web pages and other digital experiences. A CMS manages the page; it relies on a DAM for the governed media placed inside it.

ExampleWeb editors pick images from the DAM inside the CMS editor, so a recalled asset disappears from every page that used it, because pages reference the DAM asset rather than a copy.

The people, processes and tools that plan, produce, govern and measure content as an ongoing operation rather than a series of projects. It covers capacity planning, standards, tooling and performance review.

ExampleA content ops lead sets the metadata standard, owns the DAM roadmap and reports reuse rates to marketing leadership.

Measuring how specific assets perform in the channels where they run, such as views, engagement, conversion and sales, and attributing results back to the master asset rather than to each copy.

ExamplePerformance of a hero image across 14 retailer pages rolls up to one asset record.

Using an existing asset, or a variant of it, for a new purpose instead of producing a new one. Reuse depends on findability, clear rights and modular design.

ExampleA seasonal campaign could reuse 40% of last year's product photography with new backgrounds.

The end-to-end path content follows from request and brief, through creation, review, storage and localization, to distribution and measurement. Thinking of it as a supply chain makes bottlenecks, waste and handoffs measurable.

ExampleIf mapping the chain shows assets waiting six days for legal review, that delay matters more than faster design tools.

A composite measure of an asset's worth across dimensions such as usage, reach, performance and cost, used to compare assets objectively. Its formula should be transparent enough to argue with.

ExampleTwo equally liked images differ in score because one was reused in nine markets and the other in one.

The context window is how much information a model can consider in one step. Agent memory stores facts across steps or sessions. Both decide what an agent knows when it acts, so both need the same data permissions as the agent.

ExampleAn agent remembers that a market uses a different legal line, but only for users allowed to see that market.

A fixed list of approved terms for a field, so people choose a value instead of typing one. It removes spelling variants and synonyms that split search results.

Example'Hoodie', 'hooded sweatshirt' and 'hoody' map to one approved term, so all three searches return the same assets.

A DAM extension inside creative tools such as Photoshop, InDesign, Illustrator or Figma, so designers search, place and check in assets without leaving their application.

ExampleA designer places a DAM image into InDesign; when the image is updated, the layout shows it as modified.

The discipline of running a creative team as a service: intake, briefing, resourcing, proofing and delivery. It is the production slice of content operations.

ExampleCreative ops replaces email requests with a brief form that pre-fills campaign metadata the DAM will later need.

The degree to which people actually use the DAM as their source for assets, measured by active users, searches, downloads and the share of assets published from it.

ExampleIf teams still swap files on chat, the DAM's metadata and rights checks are bypassed.

The documented argument for investing in a DAM or content intelligence platform: current costs and risks, expected benefits, measurement plan and payback period.

ExampleThe case quantifies content waste from the last three campaigns to justify the investment.

The project of configuring a DAM for an organization: metadata model, permissions, workflows, integrations, migration and training. User adoption usually takes longer than configuration.

ExampleLaunching with three teams and their real workflows produces faster adoption than a big-bang rollout.

Moving assets, metadata, versions and relationships from one system to another. Most effort goes into mapping and cleaning metadata, not moving files.

ExampleMapping 60 legacy fields to 18 in the new model, and dropping duplicates, can cut migration volume substantially.

The return from operating a DAM: time saved searching and recreating, avoided rights penalties, faster launches, higher reuse and lower tool costs, set against license and operating costs.

ExampleA business case might count hours saved per marketer each month and rights claims avoided each year.

The component in which a model chooses which governed action to take next, as opposed to executing rules a person wrote. It must operate inside the same permissions, logs and guardrails as rule-based automation.

ExampleInstead of a fixed rule, the model decides an asset needs a new crop for a channel, then calls the governed rendition action.

A digital file that an organization has decided is worth governing: it is described, owned, findable and cleared for some use.

ExampleA product photo on a designer's laptop is a file. The same photo in a DAM, tagged with its SKU, photographer, usage rights and approval status, is an asset.

Software and practice for storing, describing, finding, governing and distributing an organization's digital assets from one place. A DAM sits between the teams and tools that create content and the channels that publish it, and records what happens to each asset along the way.

ExampleCreative uploads a campaign once; e-commerce, regional marketing, agencies and the website all pull approved versions from the same record.

Getting approved assets to where they are used: websites, commerce platforms, retailers, social tools, print vendors and partners. Good distribution sends references or governed copies, so updates and withdrawals propagate.

ExampleRetailer portals receive updated packshots automatically when a pack changes, without anyone emailing files.

Technology that encrypts content and controls playback or copying, as used by streaming services. It differs from rights management in a DAM, which records and enforces usage terms but does not lock files.

ExampleA film studio uses DRM for consumer streaming and DAM rights management for trailers sent to partners.

Identifying identical or near-identical assets. Exact duplicates are found with file checksums; near duplicates (resized, recompressed, lightly cropped) need perceptual hashes or embeddings.

ExampleBefore a migration, near-duplicate detection shows how many of 1.2 million files are really unique assets.

Generating the needed rendition at request time from parameters in the URL (size, crop, format, quality), usually served through a CDN. Pages request exactly what they need instead of storing every variant.

ExampleThe product page asks for a 640px WebP on mobile and a 1600px AVIF on desktop from the same asset.

A restriction that keeps an asset hidden or undistributable until a set date and time, typically for launches and announcements.

ExampleLaunch imagery is visible to the press team now but cannot be downloaded until 09:00 CET on launch day.

Metadata written inside the file itself, so it travels when the file is downloaded. Common standards are EXIF (camera data), IPTC Photo Metadata (captions, credits, rights) and XMP, Adobe's extensible framework standardized as ISO 16684-1.

ExampleWriting the copyright notice and usage terms into XMP keeps them attached to the image after it leaves the DAM.

A list of numbers that represents the meaning of a piece of content, produced by a model. Items with similar meaning get nearby vectors; a multimodal model can place images, text and video frames in one shared space.

ExampleA photo of a red sneaker and the phrase 'red running shoe' produce vectors close to each other.

Adding information to an asset after ingest: tags, descriptions, transcripts, detected products, brand checks. When enrichment runs inside the governed pipeline, its outputs carry version, rights state and an audit trail like any other change.

ExampleEnrichment writes detected SKUs to the asset and records which model and version produced them.

A DAM built for many brands, regions, business units and external partners at once. What separates it from a team-level library is scale of governance: granular permissions, configurable metadata models, audit trails, integrations and service levels.

ExampleA CPG group with 40 brands runs one enterprise DAM with separate brand spaces, shared product data and a single rights model.

The transparency provision of the EU AI Act. It requires providers of generative AI systems to mark synthetic output in a machine-readable way, deployers to disclose deepfakes and certain AI-generated text, and anyone deploying AI that interacts with people to tell them so. Most obligations apply from 2 August 2026; under the AI Omnibus (Regulation (EU) 2026/1744), generative systems already on the market before that date have until 2 December 2026 to add machine-readable marking.

ExampleAI-generated campaign images carry a machine-readable marker and a disclosure where required, and the marker is designed to survive common edits such as crops and resizes.

Systematic tests that measure an AI system's accuracy and behavior on representative cases before and after changes. For agents, evals check both outputs and actions taken.

ExampleA tagging update is released only if it matches or beats the previous model on 2,000 labeled brand assets.

A design where systems react to events (asset created, rendition finished, rights window closing, upstream record changed) as they happen. For automation and agents, the event stream (Tenovos calls it the event spine) is what tells them when to act.

ExampleA 'rights window closing' event triggers notification, unpublish and channel checks without a nightly batch job.

Detection finds that a face is present; recognition matches it to a known identity. Recognition processes biometric data, a special category under GDPR Article 9 that usually requires explicit consent, and is restricted by law in many other regions.

ExampleA brand uses recognition only for contracted talent, to link assets to release agreements.

Narrowing results by selecting values from metadata fields, such as type, brand, market, status or date, with counts shown for each value. Facets turn a long result list into a few clicks.

Example'Video' + 'Approved' + 'Germany' + 'Last 90 days' reduces 40,000 results to 23.

The specification that defines how data is encoded in a file. The format decides what a file can hold (layers, transparency, color depth, audio tracks) and which tools can open it.

ExampleA PNG keeps transparency for a logo on colored backgrounds; a JPEG of the same logo would show a white box.

A file is the bytes: pixels, frames, vectors, text. An asset is the file plus its context: metadata, versions, relationships, permissions and rights. Storage tools manage files; a DAM manages assets.

ExampleTwo identical JPEGs can be different assets if one is licensed for social only and the other for all channels.

Further training a pre-trained model on an organization's own examples so it adopts specific styles, vocabularies or tasks. It changes the model; prompting and retrieval do not.

ExampleAn image model fine-tuned on approved product photography generates backgrounds in the brand's lighting style.

Models that create new content: text, images, video, audio or 3D. In content operations it is used for variants, backgrounds, localization and copy, with rights, provenance and brand rules applied to the output.

ExampleBackground replacement creates seasonal versions of a packshot from one approved photo.

Tenovos uses this term for a named action with a known, typed signature, drawn from a finite and documented set, such as 'set property', 'set relationship', 'set policy'. Because it is defined, it can be permissioned, logged and reversed. Open-ended updates to any field cannot be governed the same way.

ExampleAn agent may call 'set policy: expire on date' but cannot rewrite an asset record freely.

A query language for APIs where the client specifies exactly which fields it needs in one request, including related objects. It reduces over-fetching for front ends.

ExampleOne query returns an asset, its 400px rendition URL and its product name.

Constraining a model's output to verifiable sources supplied at run time, and linking claims back to them. Grounded output can be checked against its sources; ungrounded output cannot be verified.

ExampleEvery product claim in a generated description links to the PIM field it came from.

Constraints that bound what an AI system may do or produce: allowed actions, data scopes, thresholds, blocked content, required approvals. They are enforced by the platform outside the model.

ExampleAgents may suggest unpublishing but only a person can confirm it for assets live on more than three channels.

Model output that is fluent but false or unsupported, such as an invented product feature or citation. Grounding, constrained output formats and review reduce it; they do not remove it.

ExampleA generated caption claims a product is waterproof; the check against the PIM blocks it.

A CMS that stores structured content and serves it through APIs, without a built-in presentation layer. Front ends such as websites, apps and kiosks render the content themselves.

ExampleThe same product story is delivered to the website, the mobile app and in-store screens from one content entry.

Using a DAM as a back-end service through APIs, with other applications providing the user interface. Assets are governed centrally and appear inside the tools where people already work.

ExampleA commerce site and a mobile app render DAM assets through the API in their own interfaces.

Designing automated processes so people review, approve or correct specific outputs or actions, usually by risk or confidence. Review is concentrated on high-risk actions and low-confidence outputs.

ExampleAI tags above threshold apply directly; claim changes always wait for a reviewer.

Combining keyword matching, metadata filters and semantic similarity in one query. Filters enforce hard constraints such as rights or brand; semantics ranks by meaning.

Example'Approved, licensed for France' as filters plus 'cozy autumn living room' as meaning.

A cloud platform for building, running and monitoring integrations between many applications through connectors, triggers and visual workflows, with little or no code.

ExampleA no-code recipe moves approved assets to the social scheduler and writes post IDs back to the DAM.

A record of each automated execution: what ran, on which assets, under which rule or instruction, what changed and how to reverse it.

ExampleA bulk agent run is reviewed as one job with 4,312 changes and a single undo.

Search that matches the words in a query against words stored in metadata or extracted text. It is precise when vocabulary is consistent and misses results described with different words.

ExampleSearching 'beach' finds assets tagged 'beach' but not ones tagged 'coast' unless a synonym is defined.

Data stored as entities and relationships, following an ontology, so software can answer questions that span several hops. In content work, it links assets to products, people, campaigns, rights and performance.

Example'Which images of discontinued products are still live on retailer sites?' is a graph query across products, assets and channels.

A neural network trained on very large text collections to predict and generate language. LLMs summarize, classify, extract fields, write copy and, with tools, take actions.

ExampleAn LLM reads a creative brief and extracts audience, products, markets and mandatory claims as structured fields.

Adapting an asset for a market: language, imagery, legal copy, units, cultural fit. Localized assets are variants linked to their source, so changes upstream can be traced to every market.

ExampleA claim change in the master triggers review tasks on all 14 localized variants.

An architecture principle set: Microservices, API-first, Cloud-native SaaS and Headless. The MACH Alliance, founded in 2020, certifies vendors that follow it.

ExampleA retailer requires MACH-certified vendors in its RFP to keep its stack composable.

Methods that learn patterns from data instead of following hand-written rules. In content work it powers tagging, similarity, prediction and generation.

ExampleA model trained on past campaign data predicts which image style performs better on a channel.

Systems specialized for time-based media in production and broadcast: ingest from cameras, high-bitrate video, proxies, timecode, edit integration and playout. MAM optimizes the production pipeline; DAM optimizes reuse and distribution across the business.

ExampleA sports league edits match footage in a MAM, then publishes approved highlights and stills to a DAM for sponsors and social teams.

Structured information about an asset: what it shows, who made it, where it may be used, how it was produced and how it relates to other assets. Search, permissions, rights checks and analytics all run on metadata.

ExampleFiltering for 'approved, product shot, SKU 1042, licensed for EU social' only works if those four facts exist as fields.

The defined set of fields an organization uses, with each field's type, allowed values, whether it is required and which asset types it applies to. It is the contract that keeps search and automation predictable.

ExampleThe model requires 'brand', 'market' and 'rights expiry' for every image, so uploads without them are blocked.

The review process regulated industries such as pharmaceuticals use before content is released, with medical, legal and regulatory sign-off on claims and references. It requires version-exact approvals and records.

ExampleA claim on a patient brochure links to its source reference and the approval record for that exact version.

An open protocol, introduced by Anthropic in November 2024 and donated in December 2025 to the Agentic AI Foundation under the Linux Foundation, that standardizes how AI applications connect to external systems. An MCP server exposes tools, resources and prompts; any compatible AI client can use them without a custom integration.

ExampleA DAM's MCP server lets an AI assistant search assets and read rights in the user's own permissions.

A model that processes several types of input, such as text, images, audio and video, in one system. It can describe an image, answer questions about a video or compare a layout with brand rules.

ExampleA model checks whether a banner uses the approved logo, colors and legal line, and explains what it found.

The editable file from the creative application, such as PSD, AI, INDD, AEP or a Figma file, with layers, linked assets and fonts intact. Final exports are derived from it.

ExampleStoring the layered PSD next to the exported JPEG lets a regional team swap the headline without recreating the design.

Building integrations and automations with visual tools and configuration instead of programming. No-code needs none; low-code allows small scripts for edge cases.

ExampleA marketing ops manager builds the Slack notification for new approvals without a developer ticket.

Locating and labeling items within an image or frame, returning both the label and its position. It enables cropping around products and checking logo placement.

ExampleSmart crops keep the product centered when a 16:9 image becomes a 9:16 story.

Extracting text from images, scans and video frames. It makes text inside packaging, slides and screenshots searchable and checkable.

ExampleOCR finds every asset whose packaging still shows the old price.

A formal model of concepts and the typed relationships between them, such as 'product belongs to range', 'campaign features product', 'talent appears in asset'. Where a taxonomy classifies, an ontology connects.

ExampleBecause talent is linked to assets, ending one ambassador contract identifies every affected asset in seconds.

Checking whether an action is allowed when it executes, not only when a session starts. Agents running for hours then follow the same current policy as people.

ExampleA user's market access changes mid-run; the agent acting for them is refused on the next call.

Rules that decide who can view, download, edit, approve or share each asset. Role-based access control (RBAC) grants rights by role; attribute-based rules also consider properties such as market or rights status.

ExampleAgencies see only assets for their brand, and only in 'approved' state.

Images, video and audio that identify people are personal data under laws such as the EU GDPR. Consent, purpose, retention and the right to erasure apply to the assets as well as to databases.

ExampleA customer withdraws consent; every asset showing them is located through face and release links and removed.

A digital representation of a physical product that combines its 3D model, product data and approved imagery, kept in sync as the product changes. It becomes the source for renders, configurators and AR experiences.

ExampleWhen packaging changes, the twin is updated once and new renders are generated for every retailer.

The system of record for product data: SKUs, attributes, descriptions, prices, dimensions and translations. A PIM describes the product; a DAM holds the media that shows it. Linking the two by SKU lets product pages assemble data and imagery automatically.

ExampleWhen a pack size changes in the PIM, the DAM finds every asset tagged with that SKU and flags it for update.

The instructions and context given to a generative model. In production, prompts are templated and versioned like code, so the same task produces consistent results.

ExampleA versioned prompt template writes alt text in the brand's tone and length limit for every new image.

A lightweight, low-bitrate copy of a high-resolution video used for browsing, review and editing. Edits made against the proxy are later applied to the full-resolution master.

ExampleReviewers comment on a 720p proxy in the browser while the 6K master stays in storage.

An image stored as a grid of pixels, such as JPEG, PNG, WebP, TIFF or HEIC. Quality depends on pixel dimensions; enlarging beyond them loses sharpness.

ExampleA 1200-pixel-wide photo is fine for a web banner but too small for a printed poster.

A cap on how many API requests a client may make in a time period. Integrations and agents must respect it with batching, queues and retries.

ExampleA bulk re-tag job is paced at 50 requests per second to stay within the limit.

A copy of an asset produced for a specific use: a different size, crop, format, bitrate or color profile. The master is kept unchanged; renditions are generated from it, stored or produced on request.

ExampleOne master photo yields a 2000px web JPEG, a 1080x1350 social crop and a CMYK TIFF for print.

For screens, the pixel width and height of an image. For print, pixel density per inch (PPI/DPI) at a given physical size. The same file can be high resolution for web and low for print.

ExampleA 3000 by 2000 pixel photo prints at about 25 by 17 cm at 300 PPI.

An API style that exposes resources at URLs and uses standard HTTP methods (GET, POST, PUT, DELETE), usually returning JSON. It is the most common way DAMs expose their data.

ExampleGET /assets?sku=1042 returns the matching asset records.

Rules for how long assets and records must be kept and when they must be deleted, driven by legal, contractual and privacy requirements.

ExampleImages with identifiable customers are deleted 24 months after the campaign ends, as the consent form states.

A pattern where a system first retrieves relevant, authorized content (assets, guidelines, product data) and then gives it to a language model to ground its answer. It reduces invented answers and keeps output tied to sources.

ExampleAsked for the current packaging rules, the assistant retrieves the latest brand guideline pages and cites them.

In content operations, the value an asset generates relative to its cost of production and management. In finance, ROA means return on total assets, a different measure.

ExampleIf a product video costing $20,000 is linked to $400,000 in attributed sales across channels, its return can be compared with other assets.

The ability to restore the state before an automated action, for one asset or a whole job. Actions that cannot be undone need stricter approval than those that can.

ExampleAn agent's mistaken re-tag of 4,000 assets is rolled back from the job record in one step.

Collecting comments, markups and decisions on a specific version of an asset, with a record of who approved what and when. Proofing ties feedback to pixels or timecodes rather than to email threads.

ExampleA reviewer pins a comment at 00:14 of a video; the editor resolves it in the next version.

Recording and enforcing what each asset may be used for: channels, territories, time periods, media and any restrictions. It protects against using content that was never licensed for that use.

ExampleA photo licensed for EU web use until June is blocked from US print and from everything after June.

The fields that hold an asset's usage terms: licensor, license type, allowed channels and territories, start and end dates, credit line and restrictions.

ExampleA required 'end date' field turns every license into a date the system can act on.

The period during which an asset may be used. When it closes, the asset must be withdrawn from every channel that still shows it, not only from the DAM.

ExampleThirty days before expiry, owners are notified; on the day, the asset is unpublished and flagged on connected sites.

Configured logic of the form 'when this event happens and these conditions hold, take this action'. Rules automate repeatable decisions and are often configured without code.

ExampleWhen an asset arrives in the 'Packaging' folder with a SKU, set its brand and attach the product relationship.

A packaged set of code libraries and tools that wraps an API for a given programming language, so developers do not write low-level requests themselves.

ExampleA developer uses a JavaScript SDK to add DAM search to an internal tool in an afternoon.

Search by meaning rather than exact words. The query and the assets are compared as embeddings, so 'family breakfast in a sunny kitchen' finds matching images even if no tag contains those words.

ExampleA brief pasted into the search box returns candidate assets across all connected sources.

A URL that gives someone access to assets or a collection, with optional expiry, password and download rights. Unlike an attachment, it can be revoked.

ExampleA link sent to a photographer expires after seven days and allows upload but not download.

A separate file, often .xmp, that stores metadata for a file that cannot or should not be modified, such as a camera RAW file. The two must travel together.

ExampleA RAW shoot is delivered with matching .xmp sidecars carrying the photographer's captions.

Finding assets closest to a given asset in embedding space. It powers 'more like this', near-duplicate detection and style matching.

Example'More like this' on an approved hero image surfaces alternates from the same shoot.

Signing in to many applications with one corporate identity, usually through SAML 2.0 or OpenID Connect. Disabling the identity blocks new sign-ins to every connected application.

ExampleWhen a contractor leaves, disabling them in the identity provider blocks new DAM sign-ins immediately; automated provisioning (SCIM) removes the account itself.

The unique code a business uses to identify a sellable product variant. In a DAM, the SKU is the key that joins assets to product data from a PIM or commerce platform.

ExampleTagging packshots with SKUs lets the online store pull the right images automatically for each variant.

Converting spoken audio to text with timestamps. Transcripts make video searchable by what is said and provide captions for accessibility.

ExampleSearching 'free returns' jumps to 01:42 in a product video where it is said.

A set of terms the search engine treats as equivalent, without changing what is stored. It catches the words people actually type.

ExampleA search for 'sneakers' also returns assets tagged 'trainers'.

Content wholly or partly generated or altered by AI. Organizations need to know which assets are synthetic for disclosure, licensing and trust.

ExampleA synthetic model in a fashion image is labeled as such in metadata and in the published credit.

The one system whose copy of a piece of information wins when copies disagree. Each data type needs exactly one: typically the PIM for product data, the DAM for approved media and rights, the CMS for page content.

ExampleIf an image's usage rights differ between a shared drive and the DAM, the DAM record is authoritative and the drive copy is treated as stale.

A signed agreement in which a person shown in content grants permission for its use, often limited by media, territory and term. The release terms must be attached to every asset that shows that person.

ExampleAn athlete's contract covers social and web for two years; outdoor advertising needs a new agreement.

A hierarchical classification of terms, from broad to specific, used to organize assets consistently. It answers 'what kind of thing is this?' in a way every team reads the same.

ExampleFootwear > Running > Trail lets someone browse broadly or narrow precisely, and reporting rolls up by level.

A locked design with defined editable areas, so non-designers can produce on-brand variants by changing text, images or sizes within rules. The rules carry the brand guidelines into the output.

ExampleRegional marketers localize a promotion banner in 12 languages without touching layout or logo placement.

The mechanism by which a model requests that a defined function be run, such as search_assets or set_metadata, with structured arguments. The application executes the call and returns the result; the model never acts directly.

ExampleThe model outputs set_metadata(asset_id, 'market', 'FR'); the platform checks permission, runs it and logs it.

Converting media from one encoding to another, such as ProRes to H.264, or changing resolution and bitrate. It is how one video master becomes many playable outputs.

ExampleUploading a master triggers transcodes for 1080p web, 720p mobile and a vertical cut for stories.

Descriptive metadata says what the asset is about (subject, product, campaign). Administrative metadata covers ownership, rights and status. Technical metadata records properties such as dimensions, codec and color profile, usually extracted automatically. Structural metadata describes how parts fit together, such as pages in a document or clips in a sequence.

ExampleTechnical fields fill themselves on upload; descriptive and administrative fields need people, rules or AI.

Assets that were paid for, approved and never used, or used far less than planned. It is a direct cost, and a signal about briefs, findability and creative fit.

ExampleIf 3,000 unused lifestyle images already fit the next brief, surfacing them can replace a reshoot.

A store optimized to find the approximate nearest vectors to a query vector among millions, quickly. It is the engine behind semantic and similarity search.

ExampleEach asset's embedding is indexed; queries return the 50 closest in milliseconds.

A graphic stored as mathematical shapes, such as SVG, EPS, PDF or AI files. It scales to any size without losing sharpness, which is why logos and icons are kept as vectors.

ExampleThe master logo is an SVG; the DAM generates PNGs at fixed sizes for tools that cannot read vectors.

Keeping every saved state of an asset under one record, with the ability to see differences and restore an earlier version. Links can follow the current approved version or pin a specific one.

ExampleThe website's product image updates itself when version 3 is approved, because it references the asset, not a file.

A version replaces its predecessor over time. A variant exists alongside the original for a different context: language, market, size or channel. Treating variants as versions overwrites assets that are still needed.

ExampleThe French and German banners are variants of one campaign asset, each with their own versions.

The search feature that lets a user submit an image instead of words and get back similar or identical assets. It runs on similarity search over image embeddings.

ExampleA designer drops in a competitor's ad to find comparable shots already owned.

A mark added to content to indicate ownership or status. Visible watermarks overlay text or a logo; invisible watermarks embed a signal in the pixels or audio that software can detect after copying.

ExamplePre-release images carry a visible 'Embargoed' mark; final assets carry an invisible identifier for tracking.

A message a system sends to a URL as soon as something happens, such as 'asset approved'. Instead of other systems polling for changes, the change announces itself.

ExampleWhen an asset is approved, a webhook tells the CMS to refresh pages that use it.

A defined sequence of steps, owners and conditions an asset moves through, such as request, create, review, approve, publish. A workflow makes the path visible, enforceable and measurable.

ExampleAssets for regulated markets automatically add a legal review step; others skip it.