Ausra Putvinskiene

What AI Actually Does Inside Our Ad Operations

Hand placing a wooden block marked with a rocket on top of a pyramid of blocks showing target, team, screen, gear and growth chart icons, on a desk beside a laptop

Every monetisation platform says it uses AI now. I wanted to write down what it actually does at Neuwo, day to day, because I think it is further along than most people expect.

At Neuwo, AI is not a dashboard feature or a monthly report. It is connected directly to our ad platform and to the demand partners we work with. It chooses which demand fits a publisher, sets the publisher up across all of it from one request, checks that every piece is wired correctly, and traces revenue questions back to the setup behind them. A person approves every change.

The work behind one tag

For a publisher, joining a monetisation platform looks simple: add one tag to the site. Behind that tag sits a lot of manual work. Every demand partner has its own system, its own ID format and its own naming rules, and each placement has to be created and connected by hand.

The risk is that mistakes stay silent. A wrong ID does not throw an error. The ad slot simply earns nothing, and it can take weeks before anyone notices. That is revenue a publisher never gets back.

One sentence, a finished setup

It starts with choosing the right demand. Not every demand partner performs equally everywhere: some are strongest in certain content categories, others in certain countries. Our AI looks at historical performance across our network and recommends the partners that have performed best for the publisher's content and GEOs.

Then our team gives one instruction, for example: set up this new site with our in image format on desktop and mobile, with the recommended demand.

From that sentence, the AI creates the site in our platform, defines where the tag is allowed to run, creates every placement directly in each partner's system, builds the demand setup for every device, and checks that everything is connected. Nobody logs into a single external dashboard. Nobody copies an ID by hand.

Mistakes that never go live

Speed is nice. What matters more for revenue is what no longer goes wrong:

  • IDs are never typed by hand. The AI builds each one from the partner's own response, in the exact format that partner needs.
  • Naming is always consistent. Revenue reports are matched back to placements by name, so consistent naming means every euro is attributed correctly.
  • No device is left without demand. Tablet traffic is automatically covered, so it never sits there with nothing to bid on.
  • Steps happen in the right order or not at all. Demand is never connected to a site before its domains are defined.
  • Nothing gets duplicated. If a setup is interrupted, we simply run it again.

The result is inventory that earns from day one, instead of inventory that leaks for weeks before anyone finds out why.

From a revenue number to its cause

Any dashboard can show that revenue on a site moved. It cannot tell you why, or open the setup and fix it.

Our AI works with three things at once: the full setup of our platform and demand, our performance data, and a shared team memory built from our own documents. Because setup and performance use the same IDs, it can follow a question all the way through. It sees which demand partner behaves differently on a site, opens that site's setup, checks how that partner is configured, looks up what was agreed with the publisher, and proposes a fix for a person to approve.

The team memory also means nothing depends on one person's inbox. Publishers never have to repeat themselves, even when someone on our side is on holiday.

Why this is different

When most monetisation platforms talk about AI, they mean an algorithm tuning floor prices, or automated reports about what already happened. Useful, but it sits on top of the operations, not inside them. A dashboard answers the questions someone planned for. Ours can answer the ones nobody planned for. We built it differently:

  • Our AI does the operational work itself. It does not just advise a person who then clicks through ten systems. It sets up, connects and checks across all our demand, from one request.
  • Our AI understands content. Contextual AI is where Neuwo started. At its core is Neuwo's proprietary ontology-based AI, developed over years to model content and its relationships quickly and efficiently. This specialised engine provides the speed and computational efficiency needed for contextual modelling at advertising scale, something a general purpose AI model is not built for. It helps us match demand to what a publisher's content is actually about, rather than relying only on yesterday's revenue.
  • Our AI is open, not a black box. It is built on the Model Context Protocol, an open standard for connecting AI to real systems, on top of the same tools and rules our team uses every day. We built three connectors:
    • Setup: creates and updates sites, placements and demand, directly in our demand partners' systems. It cannot delete anything.
    • Reporting: answers performance questions straight from our data warehouse. Read only.
    • Team memory: our shared knowledge, built from our own documents.
  • Every request follows the same path. Someone asks in plain language, the AI picks the right connectors, shows what it wants to change, and a person approves it before anything is written.
  • Small team, full platform. The connectors are lightweight and reuse the logins and systems we already had. The first version went from idea to working in days. AI is how a lean team runs a full ad stack.

Your data stays yours, and people stay in control

AI touching revenue only works if publishers can trust it. So the rules are built into the system itself:

  • A person approves every change before anything is written, and every change is logged under that person's name.
  • The AI can create and update, but it cannot delete anything. Performance data is read only.
  • It works with site configuration and aggregated figures only. No data about individual visitors, no cookies, no IP addresses.
  • Our credentials stay on our servers. The AI never sees them.

What this means for your revenue

For publishers, it means demand chosen for your content and audience, set up correctly from day one, with clean reporting and fast answers when something changes. You add one tag. We handle everything behind it, with AI doing the heavy lifting and our team making the decisions.