An AI advertising platform built in 6,031 commits over 28 days, with a judge panel that only blocks.
Fixed-price packageAI Agent PilotUSD 5,0002 weeks
Status: Built end to end by Clouditive
AI marketing software has to answer three questions before anything is published: is the piece allowed in this industry and country, is it good enough, and who approved it. Clouditive built a platform that answers them inside the product: forbidden-claims rules, an independent panel of AI judges that only blocks, and a person who approves every piece.
The platform works as a single marketing brain. It diagnoses a client's brand, sets objectives, plans campaigns, generates copy, images, carousels, documents and short videos with voice, and schedules them on the client's social networks.
Every figure on this page comes from the product's release history, its test runs and its repository. None comes from a client's campaigns.
Everything in the package, in the order it lands. The price and the duration stay the same.
Step 1: The agent
One AI agent working on one real use case of yours.
Step 2: The evals
Evals that score its output on that use case.
Step 3: The cost
Cost per task, measured.
2 weeks
USD 5,000
Want to keep going after the package? Add engineers by the hour at the published rates, with a 6-month minimum term. How staff augmentation works.
Real clients, dated figures, and a plain note on what was not done.
Clouditive built our product with us goal by goal: the AI with its judge panel, the Google Cloud platform and 854 test files. We shipped 89 releases in 10 days, each checked to run the exact tested build, and a person still approves every piece.
Sebastian Tavul, CEO
Clouditive designed and built the whole platform: 6,031 commits in 28 days, then 89 releases in 10 days.
Add engineers to your own team at the published rates instead of buying a package.
USD per hour, drawn to one scale
The one in this case diagnoses a brand, plans its calendar, generates copy, images, carousels and short videos, and schedules them on social networks, with a person approving every piece.
No. Approval stays with a person. The judge panel only blocks verifiable defects and never approves a piece.
Gemini and open models on Vertex AI, plus text-to-speech and image models, with no single vendor. Cost per approved piece decides which model a station uses.
Forbidden-claims rules per industry and country, built from 438 data files from primary sources. 10 forbidden prompts produced 0 pieces.
Forbidden-claims rules by industry and country stop a piece before it exists. They are built from 438 data files from primary sources, and 10 forbidden prompts produced 0 pieces. A craft knowledge base per industry, niche and country covers 39 countries of the Americas, with a source and a quote for every principle. A person still approves what gets published.
That is a test result on forbidden prompts, not a promise about every piece a client generates.
Yes. For this client we did product design, 16 bounded contexts, the AI with its judge panel, the Google Cloud platform in OpenTofu, the release pipeline, observability and the QA suites.
6,031 commits in 28 days, and 89 releases in the 10 days after v1.0.0, every production deploy checked to run the exact tested build.
You do. All code and intellectual property are yours from day 1, under an MSA and SOW, with an NDA before the first call if you ask.
The AI Agent Pilot is USD 5,000 for 2 weeks at a fixed price. Hourly rates are USD 30–60 by seniority and are on the pricing page.
Start with the AI Agent Pilot: USD 5,000, fixed price, 2 weeks. One AI agent works on one real use case of yours, with evals that score its output and the cost per task measured. Longer work runs by project or by hour: USD 30–60 by seniority, with a 6-month minimum for staff augmentation.
One client's case is not a promise; the pilot is how we measure yours.
An independent panel of AI judges blocks verifiable defects and never approves; approval stays with a person. Each judge was tested on labelled examples, cases whose right answer is known, before it went into service, and calibration tools tune it against people's ratings.
Cost per approved piece decides which model each station uses, and a ledger records the cost of every piece and every model call.
A release tag builds 3 images, runs the migrations, deploys by image digest and reads back the digest that is actually live. That ran 89 times in 10 days, from v1.0.0 to v1.61.0.
854 test files cover the product. Portal routes were checked with axe at 0 violations in light, dark, forced colours and zoom. Checked, not certified.
We reply to every request within 1 business day. We sign an NDA before the call if you ask.
