---
title: "AI product development"
description: "Assistants, search, generative features and agents built into your existing platform, tested on your data and run in production with the rest of your system."
canonical_url: "https://netmedia.agency/services/ai-product-development"
lang: "en"
content_type: "page"
collection: "pages"
slug: "ai-product-development"
alternates:
  -
    lang: "en"
    url: "https://netmedia.agency/services/ai-product-development"
---

# AI product development

## AI features

Your users tried the AI demo. Now it has to work on your data, inside your permissions, at a cost you can plan. Assistants and search inside your product, generative features in the platform you already run, and agents that act through your own APIs, built on your data and run in production with the rest of your system.

AI product development

your users keep using

after the pilot

Book an automation review

## Why AI features stall after the demo

A working AI demo inside a product takes a week. The trouble starts when real users, real data and real costs arrive.

### The demo ran on clean sample data

Your production data has gaps, old formats and the edge cases your domain collected over years. A feature that looked sharp on test data gives wrong answers on yours.

### It sits beside the product, not inside it

A chat window added on the side cannot see permissions, account state or your business rules. Users try it twice and go back to the screens they know.

### Nobody owns it after launch

Model versions change and vendors change prices. Without someone checking answer quality and cost every month, the feature gets worse and users notice first.

### Usage costs nobody modelled

Model costs grow with every request. A feature that cost little in the pilot becomes a real line item once every customer uses it every day.

Why pilots stall

## Three kinds of AI work inside a product

Each one starts from a job your users already do in your platform. The AI part is one component of the system, built and released like the rest of it.

### Assistants and search inside your product

Answers from your platform's own records and documents, with sources. You get retrieval over your data, user permissions enforced on every query, and a test set built from real user questions.

### Generative features in an existing platform

Drafts, summaries and structured output where users already work, such as a proposal from a form. Users review before anything is saved, and running cost is measured before launch.

### Agents that act through your APIs

An agent with one defined job that works only through the APIs your product already has. Every action is validated and logged, and a person approves changes to money, bookings or customer data.

What you get

## From one user workflow to a feature in production

### 1. Pick one workflow

It starts with the automation review: one workflow in your product, the users who do it, and what a correct answer looks like. You get a go or a clear no-go, with the reasons.

### 2. Test it on your real data

A working version runs against a copy of your data, in your own environment. It is measured on the questions your users actually ask.

### 3. Build it into the product

The feature goes into your codebase through the same review and tests as everything else. Permissions, logging and fallbacks are part of the build.

### 4. Release it to your users

The feature is released like any other change, after it passes the test set built from real user questions. A person approves agent actions that change money, bookings or customer data.

How it runs

## Where we run AI agents in production today: our own CMS

We do not have a client product yet where we built the AI part. The system we can show you is ours: the content platform behind this website, where AI agents draft and update content.

### Agents draft, they do not publish

AI agents create and update pages, case studies and FAQ through our command line tool. Every agent write is saved as a draft. There is no path for an agent to publish on its own.

### Plain code checks every write

Before a draft is saved, deterministic checks run: required fields, length limits, page structure and banned phrases. A failed check returns a structured error, and the agent must fix the draft. The model never checks itself.

### A named person approves each batch

Publishing, discarding and live writes pass only with an explicit approval phrase typed by a person for that specific batch. Earlier approvals do not carry over, and a kill switch stops all live writes.

### The same pattern, inside your platform

An agent in your product works the same way: it proposes, code validates, a person approves what matters, and every action is recorded.

Proof from our own operation

## What can go wrong with AI in a product, and how we handle it

These are the risks we plan for on every AI feature. None of them is solved by picking a better model.

| Risk | Likelihood | Mitigation |
| --- | --- | --- |
| Customer data reaches a model vendor | Depends | We build on Azure AI Foundry. Microsoft's terms for Azure AI Foundry state that prompts and documents are not used to train the models. Each request is designed to send only the data it needs. |
| Users see data they should not | Medium | Search and assistants check the user's permissions in your system before retrieval, not after. A user's answer can only use records that user could already open. |
| An agent takes a wrong action | Medium | Agents get a short list of allowed API calls, each validated by code. Anything that changes money, bookings or customer records waits for a person to approve it. |
| EU AI Act obligations | Depends | Each feature is mapped to its risk category early, AI output is labelled where your users must know, and you keep the documentation and logs you need. Legal sign-off stays with your counsel. |
| Quality drops after a model update | Medium | Before a feature uses a new model version, it runs against a test set built from real questions in your domain. |
| Cost grows faster than usage | Medium | Running cost per request is measured on your real data during the prototype and projected at your expected usage. Where a smaller model scores the same on your test set, it is the cheaper choice. |

## Questions about AI inside your product

What founders and CTOs ask before they add an assistant, a generative feature or an agent to a platform that already has users.

### Have you built AI features for client products before?

Not yet. We have no client product where we built the AI part. What you can see is our own CMS, where AI agents work in production behind validation and human approval. Our team has built and run production systems for 20+ years, and the AI features plug into systems like those.

### Which models and platforms do you use?

We build on Azure AI Foundry. We pick the model per feature, based on how it scores on your own test questions and what it costs per request. The model sits behind one layer in your code, so changing it is a contained piece of work.

### Will our data be used to train someone else's model?

See the data row in the risk table above for the Azure AI Foundry training terms. Each request is designed to send only the data it needs, and if a feature cannot work that way, you hear it before anything is built.

### Do I need to hire an AI team to keep it running?

No. You do not need to hire an AI team to keep it running. The feature lives in your codebase and runs with the rest of your platform. We can run it with you as part of ongoing operations, or hand it over to your developers with documentation and the test set.

### What if the prototype shows it does not work?

Then you get a clear no-go with the reasons, before you spend on a production build. That is the point of testing on your real data first. Some workflows are better served by plain software, and we will say so.

## Other Applied AI

AI product development is one of four Applied AI lines. All four start with the same automation review on one workflow.

service lines

### AI strategy & discovery

- Feasibility check on one real workflow
- Prototype on your own data
- Roadmap with cost, risk and EU AI Act exposure

[AI strategy & discovery](https://netmedia.agency/services/ai-strategy-discovery)

### AI search & knowledge

- Search across documents, tickets and manuals
- Assistants that answer from your knowledge, with sources
- Questions to your data in plain language

[AI search & knowledge](https://netmedia.agency/services/ai-search-knowledge)

### AI automation

- Document capture and processing
- Handoffs between inbox, CRM and ERP
- Content and catalogue work with human approval

[AI automation](https://netmedia.agency/services/ai-automation)

## Start with one workflow in your product

Tell us which job your users do that an assistant, a generated draft or an agent could take over. The automation review tests it on your data and ends with a production plan or a clear no-go.

Book an automation review
