# Decide when to automate a recurring video Source: https://www.ngram.com/docs/chat-or-automate-video Published: 2026-11-03T00:00:00Z Updated: 2026-11-03T00:00:00Z Choose between chat and an event-driven video workflow by looking at the source, the decisions that change between requests and who reviews the result. Developer Test which requests need a person in Studio and which can run from a repeatable brief. [Open ngram](https://www.ngram.com/app) [All docs](https://www.ngram.com/docs) ## Read before this Read these direct prerequisites in order. Each guide lists any earlier context it assumes. 1. [1 Integration access Understand current connection surfaces, authentication paths, and coming-soon states.](https://www.ngram.com/docs/integrations-and-automation) ## When to use this guide Use this if you run a content or operations queue and are deciding which recurring videos to automate. Your team has a launch brief whose audience is still changing and an onboarding summary with the same approved fields every week. Bring a few recent requests so you can compare which decisions actually repeat. ## Step 1. List the decisions each brief still needs In the first request, someone must decide which presenter fits the brand, how long the story should run and what the final frame should say. Use [ngram’s MCP integration](https://www.ngram.com/integrations/mcp) to review those choices while the brief stays in the conversation. On Monday the launch team is still deciding whether a presenter or a product screen should open its announcement. A conversation is valuable because the next instruction depends on what the team sees in the first preview. Keep the brief, revision and chosen project together. A fixed automation at this stage would keep asking for fields nobody knows how to fill yet. The onboarding team has another request: every approved handoff supplies the viewer role, first task, source link, owner and due date. Those fields can become a generation brief with little interpretation. Make a sample record and see whether two people produce materially different prompts from it. If they agree, use an explicit ready signal and save the returned job ID beside the record. ## Step 2. Choose the route for each request Define a way out when the fields are incomplete. A missing reviewer or an unapproved change summary should route back to a human. Filling the gap with a generic prompt can produce a polished video for the wrong person. Automation works best where the team agrees what constitutes a complete input. For the onboarding request, define the trigger as “handoff approved.” Require the customer-facing task, a source page and the account owner. Send those fields to a create action once. Keep the returned video ID with the record. Route a completed result to the owner and a failed result to the person responsible for the workflow. ## Step 3. Define the trigger and the required fields For a Zapier workflow, open the Zapier card in Create integrations and select Install Zapier App. In your automation, use a test record with the approved brief, its source URL and the reviewer. Check the fields before enabling the trigger. ![In Create integrations, the Zapier card has an Install Zapier App button.](https://www.ngram.com/media/docs/studio-integrations-automation.png) [Zapier](https://www.ngram.com/integrations/zapier) suits a team that already keeps the approval trigger and follow-up there. [n8n](https://www.ngram.com/integrations/n8n) is useful when the team needs conditional routing or operates its own workflow. Neither connector removes the need to review what gets sent to the customer. ## Step 4. Identify any point where a person must take over A stable trigger can still produce a bad draft if the input is incomplete. Send a missing source back to the request owner before creation. If a customer-specific detail changes how the story should be told, pause at the brief and use chat for that decision. After creation, check the output before distribution. Consider a monthly update made from a public release note. A scheduled job can take that input directly. Now suppose the new release changes an admin setting and your video needs fresh footage. Start this one in ngram Studio. [Browser Handoff](https://www.ngram.com/docs/browser-handoff) lets you sign in when the site asks and hand control back. An unattended MCP or API job cannot wait for that action. For the next request, ask what would have to change in a working automation. If the answer is “only these three fields,” reuse it. If the answer is “the audience, story and evidence,” a conversation is likely the shorter path. ## Step 5. Try both routes on real briefs Estimate how often the same brief occurs and how much judgment goes into each result. A weekly task with stable inputs, predictable output and a named reviewer may be a better automation candidate than fifty one-off creative requests. The question is whether the next decision can be made from the fields available when the workflow starts. Keep a small sample of recent briefs. Mark where a person changed direction after seeing the preview. If those changes repeatedly concern pacing, audience or product accuracy, keep that review in the workflow. If they repeatedly concern a missing link or reviewer, improve the source form before investing in additional automation. Before activating an automated route, run two requests in the opposite order, a duplicate trigger and one failed job. Check that each result lands beside its own brief and nobody receives two reviews. Compare the review burden with the old manual process. If the team spends its time fixing missing inputs, improve the intake before automating more of the work. ## Step 6. Decide what can run again without intervention Choose a simple stopping rule for the trial. If completed results consistently reach the right reviewer and the remaining edits are creative, the automation has taken ownership of the repeatable part. If reviewers must reconstruct the source for every video, the automation has moved the work rather than resolved it. Look at several recent requests. Which decisions repeated? Which changed because the reviewer saw something unexpected? Copy the stable decisions into a candidate form and leave creative judgment with a person. A recurring launch format may automate project creation and file delivery while keeping scene approval in chat. ## References and next steps - [Connect ngram over MCP](https://www.ngram.com/docs/mcp) — troubleshoot the connection to your external AI client. - [Connect Zapier](https://www.ngram.com/integrations/zapier) — set up a repeatable trigger once the inputs are stable. - [Browser Handoff](https://www.ngram.com/docs/browser-handoff) — collect product footage when a website asks you to sign in. ## Continue in ngram Open the workspace and put this guide into practice with your own project. [Open ngram](https://www.ngram.com/app)