How to Manage AI Video Production: Plan, Track, Approve

To manage AI video production, treat every video as a job with an owner, a plan, a price and a status. Decide who approves and what one video may cost before work starts, approve each storyboard and its quote before anything renders, track every job through its stages, clear what waits on a person, review each finished cut before release, and read the month as a whole. This guide shows how in Alpha, Nagent's video studio, where that pipeline is built in.
Why does AI video production need managing?
AI made video fast to start and easy to lose track of. A team asks for a few cuts, renders run for an hour or two each, some stall for a missing still or a consent form, and nobody can say what was spent or what is waiting on whom. Video is also the most expensive thing an agent in your workspace can make, so surprises cost real money.
A managed pipeline answers four questions at any moment:
What was asked for, and by whom?
What did we agree to pay?
Where is each job, and why is it stuck?
Who signed off the cut before it shipped?
Alpha is the video teammate on Nagent's Content team, a team lead under CREA, currently in Beta. Alpha quotes the credits before anything renders and waits for a person to approve the storyboard. Every request becomes a job, and a person reviews the finished cut before it is delivered. The steps below set that pipeline up and run it.
Step 1: Who should approve video work?
Set the roles before the first request. Alpha separates three permissions:
Making a video needs the content write permission. Give it to the people who brief video.
Following jobs needs only content read, so anyone who can read content can watch progress.
Approving a storyboard and its quote needs the settings write permission, which workspace admins hold.
Requests that wait under Awaiting approval in Approvals are decided by an Organisation admin, or by someone with a deciding role where the work belongs. Personas have a stricter rule: a new persona is a draft until another admin of the workspace, or Nagent, approves it, and the person who asked cannot approve it. When you invite people, give at least one other person a role that can decide, so requests never wait on you alone.
Step 2: How do you set what a video may cost?
Cost is controlled in layers, and every one is checked before anything is spent.
The quote. Alpha prices each video in credits before anything renders, and nothing paid starts until a person approves it.
Alpha's own budget. Alpha carries a daily and a monthly budget and a cap on any one job. A quote above a cap parks the job until a person lets it go on.
The brief's cap. A brief can carry its own cap. A quote above it waits for a named approver, and no agent may lift that cap on its own.
The workspace allowance. Your workspace's monthly credit allowance covers every agent, Alpha included.
Spending past a cap always waits for a person, however far Alpha is trusted. The credits and budgets guide explains how the limits fit together.
Step 3: How do you plan each request?
Open the Content room, choose Team surface, then the Alpha tab, marked Beta. Alpha's studio opens inside the room, under the heading Alpha. Video ads and UGC.

Across the top sit Alpha's six views. Studio, where requests are made, works right there in the room. Jobs (every request tracked), Assets (finished videos), Personas (who can be on screen), Mission control (the month so far) and Routes (the ways a shot can be made) open on Alpha's own page.
In Studio, the requester filters the presets by desk (ADS, FEED, AVATAR, TWIN, UGC, LENS, FASHION or DEMO), picks a preset or none, sets the details and describes the video. The bar above the composer shows the credits quote before anything is sent. Alpha replies with a storyboard: each shot with its timing, purpose and route, the layers on top, and the quote, broken into planning, shots, finishing, composition and scale. Every voice-over line reads the same way: from the brief or supplied copy. Alpha does not write the words.
Step 4: How do you approve a storyboard and its quote?
Nothing paid starts until this is passed.
In the studio, the storyboard arrives with Approve storyboard and Request changes. Someone who may not approve sees Send for approval instead. Request changes puts the plan back in the composer.
In Approvals, a sent request waits under Awaiting approval for someone who can decide it.
On the job's own page, a job waiting for its storyboard or quote carries the approval card. If you may approve, you decide there; if not, the card names who can.
From a team room or chat, Alpha first shows a card naming the settings it will send; nothing is sent until someone approves it, and the job then waits for its storyboard approval like any other.
Typing "approve" in a thread doesn't approve anything; only the buttons do. The quote is fixed at approval, and a delivered job bills at most the quote, less anything that wasn't made. If a render costs Nagent more than it was quoted at, Nagent pays the difference.
Step 5: How do you track every job?
Every request becomes a job with a reference such as al-job-0004. Jobs lists them all, each with the request, who asked, the crew progress with its clock, and the credits quoted and spent. Filter by All, Running, Needs a human and Delivered. Make Needs a human your daily check: it's the list of jobs that can't move until someone acts.
Paid video often starts in Marketing, where HOOK (Beta), the ad creative strategy lead, writes the briefs and hands video ones to Alpha.

HOOK's In production view shows where each brief is and who is making it. A video brief is taken by a person in Alpha: Jobs lists HOOK briefs waiting for Alpha, each with Take the brief, so paid requests from Marketing land in the same queue as the studio's.
Step 6: What does a job's page tell you?
Select a job to open its page. The Crew list shows the eight stages, each with a note and its share of the clock:
Stage: SLATE, What happens: The request is read and quoted
Stage: BOARD, What happens: The storyboard is planned
Stage: CAST, What happens: The persona is locked and the voice renders first, setting the clock
Stage: ROUTE, What happens: A render lane is chosen for each shot, with the reason kept
Stage: RENDER, What happens: Shots render, natively in each ratio
Stage: CUT, What happens: Layers, captions and the end card are composed
Stage: PROOF, What happens: The finished cut is checked
Stage: REEL, What happens: The video is stored and returned to whoever asked
The page also shows the deliverables, the credits quoted and spent so far, and every event in order. Renders take 30 minutes to 2 hours of wall-clock, so the clock estimates when the job will be ready and pauses while it waits.
Step 7: How do you clear what's waiting?
A job that can't move says why, in one of two ways.
Needs you means a person must decide:
P0 storyboard: the storyboard and quote wait for an approver.
Missing input: the brief lacks something, such as the words or a product keyframe.
Consent: a release or rights record is needed before anything renders.
Over the cap: the quote is above the cap and waits for a named approver.
Cut review: PROOF has checked the cut and a person must read its report.
Parked means the platform isn't ready, and nothing is spent while it waits:
Queues off: rendering is switched off in this environment.
No route: no render lane can make one of the shots yet, and the page names which.
Spend cap: Alpha's spend cap is reached; it carries on when the cap allows.
PROOF offline: PROOF tries again by itself, and nothing is delivered unchecked.
A parked job keeps its stage and its approved quote and resumes where it stopped. Work the Needs you reasons yourself; raise Parked reasons with whoever runs your workspace, or with the Nagent team.
Step 8: How do you review the cut before release?
Releasing a finished cut after PROOF always waits for a person, however far Alpha is trusted. PROOF checks each rendition and writes one report: whether words or a logo sit where the platform covers the frame, whether the spoken words match the storyboard, whether a presenter's lips move with the words, whether every face in the presenter's shots is the same person, and any personal data on screen. A report that blocks fails the job.
Read the report, then watch the cut yourself, because PROOF can misread. Approving the cut review delivers the video to Assets. On a DEMO cut, each item of personal data has a tick box on the job's page, and an item left unticked sends the cut for a second, named approval instead of delivering it.
Step 9: How do you read the month at a glance?
Mission control counts this workspace's video work for the month to date, from its own jobs, renders and personas. Nothing on it can be changed.

Delivered: videos delivered this month, and how many were on time within the quote.
First-pass PROOF: PROOF reports kept this month that read PASS.
Credits per rendition: credits spent on delivered jobs, per rendition.
Waiting on a person: approvals and reviews waiting now.
Releases ending within 30 days: persona releases about to run out.
Use it in your weekly content review: first-pass PROOF tells you whether briefs are clear, and credits per rendition tells you what a finished video really costs. Two figures say "not counted yet" rather than show a number: GATE pass, and ad winners, because ad results do not reach Alpha yet.
Step 10: How do you keep personas and routes in order?
Personas lists who a talking preset can put on screen, each with whether it can be cast today and why not. Renew a digital twin's written release before its last day, because an expired release stops the twin being cast. Mission control lists releases ending within 30 days. Revoke stops new videos using a persona, stops open jobs and marks every video made with it do not reuse.
Routes is read only. It lists each way a shot can be made, its rate in credits per second, and how its work fared at PROOF. Routing prefers the routes whose work passes. Switching a route on or off is done by Nagent, not from this page.
Where does Alpha sit in the team?
Nagent's AI teammates are organised like a growth team: an AI Chief of staff heads each department, team leads own a kind of work, and leads can run a bench of sub-agents. The full chart is on the team page.
CREA, AI Content Chief of Staff
RUPA, team lead, image studio
FOLIO, team lead, long-form
Alpha (Beta), team lead, video studio
Bench on the team page: ADS, UGC and AVATAR; the studio's desk filters also show FEED, TWIN, LENS, FASHION and DEMO

How a request moves. You ask the Content room in one line. CREA decides who owns it and hands video to Alpha. Alpha plans and quotes it through the right desk, and the storyboard waits for a person in the studio. Agents do the work and report it in the room; people set the direction and approve what goes out.
How handoffs work. A sub-agent hands work back to its lead, never sideways, and every hop is checked against the receiving agent's budget. Asking another department is a written handoff to that room's lead: from another team's room, writing @ALPHA puts a brief on the Content board for CREA with Alpha named on it. HOOK on Marketing writes the briefs for paid video, and its GATE checks the videos that come back. RUPA makes Alpha's product keyframes. Alpha never publishes, traffics or touches an ad account.
What are the limits today?
Alpha is in Beta: planning, quotes and approvals work in every workspace, and render lanes are still arriving. Creator, presenter and twin shots render only where those lanes are connected for your workspace, and the plan names any shot with no lane before you approve. The studio works inside the Content room; Jobs, Assets, Personas, Mission control and Routes open on Alpha's own page for now. If a delivered video matters to a date, talk to the Nagent team first. The Alpha studio guide lists what the Beta does today.
Frequently Asked Questions
How do you manage AI video production?
Treat each video as a job with an owner, a plan, a price and a status. Set approvers and caps first, approve each storyboard and quote before rendering, track jobs through their stages, clear what waits on a person, and review every cut before release.
How do you approve AI-generated videos before they are made?
In Alpha, every storyboard and its credits quote wait for a person. An approver clicks Approve storyboard in the studio, decides on the job's page, or approves the request in Approvals. Typing "approve" in a thread does nothing; only the buttons do.
How do you control the cost of AI video?
Alpha quotes every video in credits before anything renders, and the quote is fixed at approval. Budgets and caps are checked before spend, and a quote over a brief's cap waits for a named approver.
Why is my AI video job stuck?
Open the job's page. "Needs you" means a person must act: the storyboard, missing input, consent, a cap or a cut review. "Parked" means the platform isn't ready, such as no render lane for a shot, and nothing is spent while it waits.
Run video like a pipeline, not a pile of renders
AI video stays under control when every job has a plan, a price, a status and a person who signed it off. Alpha builds that pipeline into the studio: quotes before renders, a job page that says why it waits, and a monthly view of what was delivered and what it cost.
Book a demo to see Alpha run a video job from plan to proof.