What Genspark is
Genspark is an AI workspace combining research, file creation and task execution. AI Workspace 6.0 organizes the product around memory, a general agent, office and creative tools, and collaboration between people and agents. The name covers a platform rather than a single language model.
The intended outcome determines the route: an answer to explore a question, a spreadsheet to analyze data, a presentation to explain findings or an application to provide an interface. First identify which object the recipient needs and how they will review it. A correct conversational response may still require editing, sources and an appropriate delivery format.
Reference documentation: AI Workspace 6.0.
Origins and company development
Eric Jing and Kay Zhu founded the company in 2023. MainFunc launched Genspark as an AI agent search product in June 2024. This origin explains its initial emphasis on retrieving information and assembling topic pages before expanding into documents, applications and other deliverables.
On 3 March 2025, it announced a $100 million Series A and its iOS application. On 20 November 2025, it introduced AI Workspace and announced a $275 million Series B at a $1.25 billion post-money valuation. These are dated financing milestones, not a permanent statement of current valuation.
Reference documentation: Genspark company and products · Lanchi Ventures: Genspark · Series A and iOS app · AI Workspace and Series B.
Sparkpages, Autopilot and the agent transition
The original Sparkpages assembled web information into topic pages with an assistant for follow-up questions; users could share and fork them. Autopilot, announced on 18 September 2024, added background research, concurrent tasks and completion notifications.
In April 2025, Kay Zhu explained retiring the original search product to focus on Super Agent, which adapts its steps and tools to a task. Describing current Genspark solely as a search engine is therefore incomplete. Search remains a useful operation within processes that may end in a file or an action.
Reference documentation: Introduction to Sparkpage · Autopilot Agent announcement · Why I Killed Our AI Search Product.
Super Agent and working files
Super Agent executes work on a project-specific cloud computer. It distinguishes delivered files from scratch files: delivered files remain in the project, while scratch files can disappear when the computer is recycled. Documentation directs recurring tasks to Workflows rather than a scheduler within Super Agent.
A well-defined assignment specifies inputs, outputs and checks: for example, an evidence-based report accompanied by the underlying data. Explicitly request delivery of auxiliary files that must be retained. An execution summary alone cannot reconstruct a formula or establish which version of a table supplied a chart.
Reference documentation: Super Agent documentation.
Conversation, search and models
AI Chat provides text conversation with model selection and a Mixture-of-Agents option that synthesizes multiple responses. It can read files and enable sourced web search, depending on the model. It has no memory across conversations and does not automatically read Hub files; interface personalization is not that memory.
Use it to refine a question before delegating execution. To compare two communication methods, request a matrix of audience, required information and format. Then inspect the original references: agreement between models is not an independent measurement and does not turn a commercial claim into technical evidence.
Reference documentation: AI Chat documentation.
Presentations and Gen-1 Slides
AI Slides combines instructions and materials to build presentations. Professional or Creative, Auto or Guide, and Standard or Ultra are project choices. Exporting PDF, PPTX or Google Slides requires a compatible paid plan; Free supports creation, presenting and link sharing.
Introduced on 10 September 2026, Gen-1 Slides is a specialized Standard-mode model post-trained from MiniMax M3 with Fireworks AI. Its published evaluation concerns presentation tasks, not expertise in every subject included. For a technical report, approve the outline, sources and figures before optimizing visual design.
Reference documentation: Genspark AI Slides · Gen-1 Slides introduction.
Documents, editing and versions
AI Docs distinguishes manually editable Rich Text and Markdown from Web Doc and Word, revised through instructions. The format is fixed once content is created. Exports include Word and PDF, plus Markdown where applicable. Save Points allow selecting an earlier version; exports use the selected version.
When maintaining an internal procedure, separate editorial corrections from technical changes. A request to shorten a document can remove necessary conditions. Identify the exact section and purpose of the change, then compare the approved paragraph with its replacement before updating the distributed file.
Reference documentation: Genspark AI Docs.
Spreadsheets and data analysis
AI Sheets combines extraction, analysis and visualization. It works with files or connected sources and can use Excel formulas, SQL and Jupyter. Users can edit the sheet directly or through instructions and export XLSX. Review data provenance and calculation logic alongside the result.
A proposed example is summarizing preventive actions by month without personal data: date, category and status. Before calculating percentages, decide whether the denominator is open, planned or overdue actions. That choice changes the indicator’s meaning. Recording it in a column or note makes the analysis reproducible at the next review.
Reference documentation: AI Sheets documentation.
Memory, notes and storage
SecondBrain assembles context from connected sources for reuse; AI Drive organizes files for use across tools. SecondBrain Note is a recording device: it can capture audio without a phone, but requires an app connection for transfer and processing. Its documentation does not provide real-time transcription from the device.
Memory, an archive and approved minutes serve different purposes. Saving a conversation makes retrieval easier; approving minutes establishes recognized decisions and responsibility. For a safety meeting, review task names, owners and dates, and retain the approved document reference rather than treating every automatic summary as a formal decision.
Reference documentation: AI Workspace 6.0 · SecondBrain Note documentation · AI Drive documentation.
Media production and delivery formats
Introduced on 27 October 2026, Media Studio coordinates scripts, storyboards, generation and editing through multiple models. It includes an asset canvas and manual editing. Documentation distinguishes flattened PNG images from layered PSD files and provides editable Premiere or CapCut projects alongside finished video.
For a training clip, retain the script, voice, captions and visual assets as reviewable elements. A person should be able to correct an equipment name without rebuilding the entire piece. Testing one clip in the destination editor before producing a series establishes whether layers, text and timing retain the required flexibility.
Reference documentation: Media Studio announcement · Media Studio documentation.
Applications and assisted development
Genspark Code, formerly AI Developer, combines agent-assisted planning, building, testing and deployment. It distinguishes web, mobile, game, e-commerce and existing-code projects. Its project list belongs to Code rather than representing every document or presentation in the account.
A proposed OSH prototype is a public resource library with activity filters and links to current documents. Acceptance should cover searches with no results, withdrawn links and mobile reading. If accounts or internal data are connected later, authentication, permissions and maintenance become project requirements. The prototype’s appearance does not resolve those decisions.
Reference documentation: Genspark Code documentation.
GenTeam and agent controls
GenTeam places people and persistent agents in shared conversations. Cloud agents consume their creator’s credits; agents connected to a personal computer use its model access, with GenCode billed to Genspark as an exception. Access controls determine who can request work and use each connected application.
Keep responsibilities visible: one person supplies documentation, an agent drafts and another person approves. A written instruction to ask permission does not replace an access restriction. A safety team should restrict sources to the assignment and separate reading tasks from those capable of changing documents or communicating with others.
Reference documentation: GenTeam documentation.
Automation and tests with real effects
Workflows builds automations with a schedule or email trigger. Test Run executes steps for real: it can send messages and modify data. Building or updating a workflow can also test actions. Configured approvals and execution history matter before enabling recurrence.
A reasonable first trial for an internal bulletin would use public documents, a test spreadsheet and the owner’s email address. Review which sources were included, what was missing and what was written. Expand only after checking destinations, content and permissions. Test status should be reflected in the resources used, not merely in the process name.
Reference documentation: Workflows documentation.
Phone calls and permitted scope
Call For Me places real calls and returns transcripts and summaries. It requires paid access and phone verification; availability depends on the destination. Documentation excludes promotional or sales calls and emergency numbers, and prohibits concealing that the caller is AI or whom it represents.
A simple logistical inquiry, such as confirming opening hours, differs from communicating a critical safety instruction. For any authorized communication, specify the contact, permitted information and intended result. A call summary can help record the reply but should not be treated as formal acceptance of conditions the recipient did not confirm.
Reference documentation: Call For Me documentation.
Plans, platforms and administration
The general offering distinguishes Free, Plus, Pro, Team and Enterprise. Free has a daily allocation also subject to a lifetime allowance; paid personal plans combine credits and conditional quotas. Commercial use is listed among Plus and Pro benefits. Personal and team subscriptions do not automatically merge.
Team and Enterprise add centralized administration, connectors and SSO. Administrator status does not grant access to members’ private projects. Procurement should compare a task’s full route: generation, review, storage and export. Identify the application to be used, since web and mobile availability does not establish parity across every editor.
Reference documentation: Membership Plans · Team & Enterprise Plans.
Data, training and sharing permissions
The privacy policy, updated on 24 July 2026, provides for external AI processors and US processing. Team and Enterprise document training opt-out by default; certain Media Studio modes depend on enabling AI Data Retention. Check the applicable configuration and contract rather than inferring them from the plan name.
When sharing a result, distinguish a single file from the whole project, which may include conversations and working materials. Initial OSH trials should use public documents or aggregated information. Personal health records require a separate handling process and are outside this profile’s proposed editorial case. Also establish what each connection retains after it is disconnected.
Reference documentation: Media Studio documentation · Team & Enterprise Plans · Privacy Policy.
Applications in occupational safety and health
It can support occupational safety campaigns and awareness materials by turning reviewed technical content into different formats. Define the central message and its sources before generating assets, so that an attractive presentation preserves preventive meaning and does not introduce unsupported facts. Consistency between the different versions should be part of the final review.
For a safety campaign, Genspark can help turn approved content into a visual sequence and supporting document. For a review meeting, it can prepare a narrative from supplied indicators while retaining their definitions. For training, it can produce an initial script subsequently adapted to the actual job and materials. Sabentis proposes these uses as assisted editorial production, without attributing the ability to certify instructions or assess exposures to the system.
Worked example and checks
For example, a team prepares a hearing protection campaign from approved instructions. It requests a presentation script and supervisor summary, checks terminology and recommendations, and adapts images to the protective equipment actually used at its sites.
- Prepare a defined source set: current instructions, authorized photographs and aggregated data, with dates.
- Request an outline about hearing protection first, identifying the objective, audience and message of each slide.
- Approve the outline and generate the presentation and summary without adding unrequested external statistics or recommendations.
- Compare each claim with its source and each image with actual equipment; explicitly identify conceptual illustrations.
- Export and open files in the destination program, checking text, charts, cropping, links and notes.
If a chart shows percentages, verify denominators and labels against the original data. If a slide changes, review the summary too, avoiding two versions of the same message. The exercise is successful when both materials retain their content, scope and sources, and the exported files support the intended use. Record the approved versions so a later design revision does not silently change an instruction.
This is a hypothetical workflow, not a report of a campaign tested with Genspark.
Limitations and conditions of use
Visual polish does not establish accuracy. Review figures, references and depictions of protective equipment. Some functions may require additional services or permissions, and generated materials need approval before entering the company’s communication process.
Visual mode and format determine which elements can be edited afterward. Check the delivered file and compare its claims with their sources. Sharing permissions also matter: a link may show only the result or include the process and conversation. Review that scope before distributing internal materials.
Reference documentation: Media Studio documentation.
Choosing and complementing the tool
Choose Genspark when coordinating research and production of several deliverables in one workspace. PowerPoint can support final review of an exported presentation; Canva can complement specific graphics. Base the choice on the required final format, editing needs and approval process, rather than the appearance of a preview alone.
A brief trial containing a table, image and link checks portability before producing an entire campaign. Retaining an approved copy helps distinguish content changes from later visual adjustments. It also provides a stable reference when different people review the document and slides separately.
Frequently asked questions
Does viewing a presentation mean I have a downloadable PPTX?
No. Export and available formats depend on the module and plan.
Can I manually edit every generated document?
Do not assume so. AI Docs distinguishes formats with manual editing from those changed through instructions.
Does automated review establish accuracy?
No. It can assist, but numbers, sources and technical representations require specific checks.
Reference documentation: Genspark AI Slides · Genspark AI Docs.
