AI Video Ads for Ecommerce (Drafted by Your Co-Founder)
AI video ads for ecommerce: how an AI co-founder drafts HeyGen avatar video and static ads, a human approves, and live metrics feed the next round.
AI video ads for ecommerce are short promotional videos that software generates instead of a film crew: the AI writes a script, produces an avatar presenter with a synthetic voice, and assembles a feed-ready clip in minutes. The fastest path is to let the AI co-founder that built your store draft them, then approve each before it publishes.
This is a SoGood post, so the disclosure up front: SoGood.ai is an AI co-founder for ecommerce businesses that builds and runs a physical-product store. It is not a dedicated ad platform, and it is not a creative studio. What follows is how its ad drafting actually works, where it loses to specialist tools, and the one place it genuinely wins.
SoGood is priced in tiers: Basic is free, Pro is $29 a month, and Expert is $99 a month, and you can add credit packs on any plan. Ad campaigns are part of the Expert tier, while Pro covers brand, website, SEO, social, and payments.
How an AI co-founder drafts a video ad
The drafting flow starts from context the agent already has. Because the same co-founder generated your brand identity, seeded your products, and set your prices when it built the ecommerce store, it does not ask you to paste any of that in. It writes the ad script against the brand voice it defined.
For the video itself, the agent uses HeyGen to generate an avatar presenter and a synthetic voice reading the script. The output is a short clip sized for a social feed, not a thirty-second TV spot. Alongside the video, it drafts static image variants in the same brand palette so you can test motion against a still.
The agent does not film anything or hire a model. That is the trade you are making: speed and near-zero creative cost in exchange for the synthetic look of an AI avatar. For most early ecommerce tests, that trade is worth it, because you are trying to find a message that converts before you spend on production.
The drafting also reuses the brand photography the co-founder already generated for the store. Because it produces images in multiple aspect ratios, the same product shot can anchor a square feed ad, a vertical story, and a still frame inside the video. You are not starting the creative from a blank canvas each time.
Video or static: which to draft first
Video is not automatically the better ad. It earns attention in a busy feed and suits products that benefit from demonstration, like apparel on a body or a gadget in use. It also costs more in production effort and can read as cheap when the avatar looks obviously synthetic.
Static ads are faster to draft, easier to read at a glance, and often cheaper to test at volume. A clean product image with a sharp headline frequently beats a mediocre video on cost per click. The honest move is to draft both, because the agent generates them together anyway, and let the metrics decide rather than your gut.
For a new store with no creative history, start with two or three static variants and one video, run them on a small budget, and read the click-through rate. Promote whatever wins. The point of AI drafting is that running this test costs you minutes, not a production budget, so there is no reason to bet everything on one format.
The draft, approve, publish, learn loop
The whole cycle is built around a human gate. The agent drafts, a person approves, the ad publishes, and the metrics feed the next batch. Here is the flow in one picture.
Step one is the draft, which the agent produces on its own. Step two is the part people get wrong about AI ads: a human reviews every draft in the Ads dashboard and either edits, rejects, or approves it. The agent cannot publish on its own.
Step three is publish. Approved ads go live on Facebook, Instagram, or X, paid from credit packs you load for those platforms. Step four is metrics: the dashboard reports live impressions, clicks, click-through rate, spend, and ROAS for each ad, and what worked informs the next round of drafts.
The honest limitation: a human owns publish and pause
SoGood does not auto-publish, auto-scale, or auto-pause ads. The agent will draft ten variants and surface their numbers, but it will not move budget around without you. This is a deliberate guardrail, and it is the right one for a tool spending real money on a solo founder's behalf.
If your goal is fully hands-off media buying, that is a real limitation, and you should know it before you sign up. A platform autopilot like Meta Advantage+ will optimize delivery inside one network with far less human input. SoGood trades that automation for a human checkpoint and cross-platform drafting.
The upside of the guardrail is trust. You never wake up to a drained ad budget because an agent decided to scale a losing creative overnight. For founders who are nervous about handing spending authority to software, the approve-first model is a feature, not a constraint.
There is a workflow cost to the gate: you have to show up. If you ignore the dashboard, drafts pile up and nothing ships, because the agent will not publish to fill the silence. Plan a short weekly review where you approve, reject, and adjust budgets, and the loop runs smoothly. Skip it, and the whole system stalls by design.
What the agent cannot do on the ad platforms
Be precise about scope. SoGood drafts ads and reports their numbers across Facebook, Instagram, and X, but it is not a replacement for the platforms' own ad managers. Advanced audience building, lookalike modeling, and granular bid strategy still live in the native tools.
It also does not do multi-touch attribution. The dashboard tells you an ad's own ROAS, but it will not stitch a customer journey across five touchpoints. For that level of analysis you stay in the platform's reporting or add a dedicated analytics layer. Treat the co-founder's ad module as a fast drafting and launch surface, not a measurement stack.
Where dedicated ad tools beat SoGood
Be clear-eyed here: on pure creative depth, dedicated studios win. They exist to do one job, and they do it deeper than a bundled module ever will. SoGood scores medium on creative depth, on purpose, because it is not a studio.
| Tool type | Creative depth | Video ads | Knows your brand already | Runs the rest of the business |
|---|---|---|---|---|
| Dedicated studio (AdCreative.ai, Pencil) | High | High | Low, you paste it in | No |
| Platform autopilot (Meta Advantage+) | Medium | Medium | Shallow | No, one network only |
| AI co-founder (SoGood.ai) | Medium | Yes, HeyGen | High | Yes |
A dedicated studio like AdCreative.ai ships a large template library, dozens of variations per concept, and creative scoring against past performance. Pencil leans hard into video ad generation and predictive performance. If creative breadth is your single biggest bottleneck, buy the specialist.
The matrix below scores the three tool types honestly, so you can see exactly where SoGood sits and where it does not.
The one place an AI co-founder wins: context
The edge is not the creative. It is that the ad tool already runs the company. A dedicated studio asks you to paste in your brand guidelines, product details, audience, and offer every time. SoGood does not, because the same co-founder that drafts the ad also built the store and knows all of that already.
That context compounds. The agent can tie an ad to a specific product's margin, draft it in the exact voice it set for your brand, and report ROAS next to the orders flowing through the store it deployed. For a solo founder, that is fewer logins and one AI co-founder coordinating marketing instead of a stack of disconnected tools.
This is the same logic behind running a whole lean stack under one roof. If you are weighing the bundle against buying every specialist separately, the honest math is in our take on building an AI marketing stack when you cannot afford an agency: the bundle only wins once you actually use several of the modules.
Who should use AI video ads inside a co-founder
The fit is a solo or very small ecommerce team that wants ads drafted fast, in-brand, without standing up a separate creative workflow. If ads are one of five or six jobs you need an AI agent for your ecommerce business to cover, the bundled drafting is genuinely useful.
The poor fit is a performance marketer who lives in creative testing all day. That person should buy a dedicated studio for depth and a platform autopilot for delivery, and treat the co-founder's ad module as a backup. Match the tool to whether ads are your core job or one of many.
Either way, treat AI video as a test, not a finished campaign. Approve a couple of variants, run a small daily budget, read the ROAS in the dashboard, and scale only what earns its spend. The AI removes the cost of making the ad; it does not remove the discipline of reading the numbers.
Budgeting AI video ads honestly
The creative is nearly free, but the media is not, and that distinction trips up new founders. Generating a video draft costs a tiny amount of compute; running it in front of real shoppers costs real money on the platform. Do not confuse cheap creative with cheap advertising.
In SoGood you fund the paid platforms with credit packs, then each approved ad spends against your platform budget once it goes live. Set a small daily cap, let an ad gather a few hundred impressions, and judge it on click-through rate and ROAS before you scale. Most variants will lose; the job is to find the one that pays and pour budget into it.
This is the same lean discipline that lets a non-technical founder launch without developers: use AI to drive the cost of trying to near zero, then spend real money only on what the data already proved works. The co-founder makes testing cheap. Reading the result still has to be your job.