​ 

​ 

Seenit AI Co-Producer Overview

Overview of Seenit's AI Capabilities, Technology and Controls

Version history 

1. Executive summary 

Seenit’s Co-Producer is an AI-assisted set of tools built into the platform designed to simplify workflows and boost productivity, with robust measures in place to meet enterprise security, compliance, and privacy needs. It supports the workflow for employee-generated video, from shaping a project brief and suggesting filming prompts to organising clips, assisting with edits and creating supporting content from an approved edit. 

The platform is designed around a human-led workflow. Users provide the brief, source material or instruction; AI proposes an output; and the user decides whether to accept, regenerate, edit, remove, approve or publish it. The specific control depends on the feature. 

2. What Seenit AI does 

Seenit’s AI capabilities fall into two broad groups. This distinction helps explain when a user actively requests an AI action and when a platform event starts processing. 

User-initiated assistance 

A user asks Seenit to help with a task, such as developing a project idea, suggesting scenes, generating filming prompts or creating a script framework. The user can review the suggestion, regenerate it and adapt it to the project. 

Content-initiated assistance 

An action in the platform, such as uploading a video or creating an edit, can trigger AI processing against the relevant asset. Examples include transcript-based tagging, clip shortlisting and the creation of supporting content from an edit. 

3. AI-assisted capabilities 

The table below sets out the main AI-assisted capabilities. Outputs are suggestions or generated assets that remain subject to the relevant user controls. 

These generative capabilities are delivered using Google Gemini through Google Vertex enterprise language models. 

AI-assisted editing and distribution 

Seenit provides AI-assisted editing and supplementary content features. Depending on the customer’s plan and feature configuration, these include suggesting impactful clips, trimming and storyboarding content, applying sound, adding b-roll overlays, generating subtitles, translating subtitles, adding lower thirds and creating supporting text or image assets from an edit. 

Some AI-powered editing features may be provided in beta. Feature availability and status should be confirmed for the relevant customer Studio before they are treated as part of a committed service. 

4. How AI works in the Seenit workflow 

At a high level, Seenit combines user instructions, project context and content already held in the Studio to produce a suggested output. The output is returned to the platform for user review and does not bypass the customer’s normal approval process. 

Examples 

1. A user enters an idea for a project. Co-Producer suggests a brief and scenes. The user refines and accepts the content before using it to collect footage. 

2. A contributor’s video is uploaded. The platform can transcribe and tag the content, using only categories configured by the customer. The user can remove tags that are not helpful. 

3. An edit is created. Co-Producer can suggest clips or create supporting copy and images. The user reviews the assets before using them outside the platform. 

General AI Flow 

 We have 2 main flows for AI usage, User Initiated, where a user can request something directly from the AI, such as ideas for a project or filming suggestions, and Content Initiated, where the creation of some content, such as a video being uploaded or an Edit being created, creates a request to the AI engine to run some processing against the assets. These are things like AI tagging, or the creation of further content from the initial asset. 

The AI processing is managed under a single switch in the studio that can turn off the Generative AI across the platform for that studio, and we try to fall back to more traditional smart features rather AI based features when it is disabled. An example of this is the auto-trimming of videos, that can either use the transcription data to trim the video to the first and last spoken word, if AI is off, or when AI is switched on, it can trim the video to remove any unnecessary preamble and trim it just for the most impactful statements. 

User Initiated -  Examples – Brief generation, script Generation 

Content Initiated -  Examples AI Tagging, Content generation 

5. Technology used 

This section describes the technology behind Seenit’s AI features and the platform that supports them. It is intended as a technical overview rather than a substitute for a customer-specific security review or contractual schedule. 

AI service layer 

• Seenit uses Google Vertex enterprise language models, including Google Gemini, for its generative AI features. 

• The AI service supports tasks such as ideation, brief generation, script frameworks, tagging, shortlisting and supporting content generation. 

• The exact model, version and regional routing may vary by feature and may change as the service evolves. These details can be confirmed where a client requires model-level due diligence. 

• The enterprise model provider operates under data protection agreements, with no retention or provider training use of customer data for these interactions. 

Seenit platform layer 

• Seenit is a cloud-based, web application for creating, curating, collaborating on and sharing employee-generated video content. 

• The platform is hosted through industry-leading ISO 27001 certified providers, including Google Cloud Platform and AWS, with availability regions used to support service availability. 

• A distributed document database is used, with logical separation of customer data. 

• Data is encrypted at rest and in transit using industry-standard encryption. 

• The software delivery lifecycle includes static code testing, penetration testing and automated vulnerability scanning. 

• The user interface is a web-based single-page application, accessible from desktop and mobile devices. 

• Co-Producer is optimised for high-performance operations at an enterprise scale, so your teams gain the productivity benefits of AI without excessive costs or compute requirements. Scalability options allow you to extend AI capabilities as your needs grow. 

Application and access controls 

• Authentication uses JSON Web Tokens, commonly called JWTs, and authorisation uses platform roles authenticated using Auth0. 

• API requests must pass identity authentication and authorisation checks before access is granted. 

• Within AI features, customer controls include Studio-level enablement, user acceptance, regeneration, tag removal, content safety filtering and review before distribution. 

6. Security, privacy and compliance principles 

Data security and privacy 

Co-Producer uses an enterprise language model provider under strict data protection terms. Customer data is not retained or used by the provider for its own purposes or model training. Interactions are used only to deliver the requested action. 

Data loss management tools are used to remove personal information from requests before they are sent for processing. This works alongside the customer’s own information governance requirements and the applicable data processing agreement. 

Data residency and compliance 

Processing follows regulatory requirements and can support designated regional requirements, including the EU where necessary. Customers with specific residency, transfer or sector requirements should validate the applicable region, sub-processors and contractual terms during due diligence. 

Intellectual property and brand customisation 

Customer data and content created using Co-Producer remain the customer’s intellectual property, subject to the applicable agreement. Brand guidelines and other Studio-specific customisation are isolated to that customer’s Studio and are not shared with or used to train the model provider. Co-Producer is optimised to enhance existing content and does not independently generate new material. This focus on refinement, rather than creation, minimises liability and keeps you in full control of final outputs. 

Explainability and Control over AI Processes  

Co-Producer provides clear, interpretable results rather than opaque or unpredictable outputs. This transparency helps users understand how AI enhances content and enables informed decision-making. Co-Producer requires user input for each task, giving you full control over AI-driven enhancements and making sure that the AI functions only as an intelligent assistant in your workflows. 

Safety and responsible use 

Content safety filtering is applied to the generative features to help block harmful or dangerous content and the models we use are selected and monitored with the aim of reducing bias and supporting fair representation. We continually monitor and improve these for fairness in outputs. 

Seenit does not use AI to make decisions that can affect users. Its AI supports video collection, organisation, editing and content generation and its script suggestions are frameworks with placeholders, which keeps contributors responsible for adding their own experiences and views. 

7. Human oversight and operational controls 

Seenit’s AI controls are designed to keep people responsible for the content and decisions that matter. The main controls are set out below. 

8. What Seenit AI does not do 

• It does not make decisions that affect an employee, contributor or other user. 

• It does not change what an employee or contributor has said in their video. 

• It does not replace the contributor’s own experience with a fully AI-written personal story. 

• It does not publish generated content externally without user input. 

• It does not give users unrestricted tagging choices when tagging is enabled. Tags are limited to the customer-defined list. 

• It should not be treated as a substitute for human review of accuracy, appropriateness, copyright, permissions or brand suitability. 

9. Client considerations 

Clients adopting Seenit’s AI features should agree how the controls will be used within their own governance framework. A practical approach is to: 

1. Define which Studios and user groups can use generative AI. 

2. Confirm the content types and data categories that may be submitted for AI processing. 

3. Set clear reviewers and approval points for scripts, tags, edits and supporting assets. 

4. Provide brand guidance and examples where Co-Producer is expected to support a specific tone or visual identity. 

5. Review generated outputs for factual accuracy, appropriate language, permissions, accessibility and brand fit before distribution. 

6. Confirm the current model, region, sub-processors, retention terms and feature status where these are material to the customer’s security or procurement requirements.