AI engineering

AI projects for one named problem.

Most clients start with growth engineering. Some then ask us to fix something else with AI: a website that people and AI assistants can both read, visibility in search and in AI answers, or agents that take a slow operational job off someone's desk. We take those on as projects, built in tools you own and quoted after the first call.

We began taking these projects because growth clients asked for them. No result is published here, because none has been cleared by a client.

No published priceQuoted per projectHow a project is scoped →

Three kinds of project.

Each one starts from a problem you can name, and ends with something that is yours.

  1. A website people and AI assistants can read

    Your site was built to look right in a browser. A buyer now asks an AI assistant about firms like yours, and the assistant cannot tell from your pages what you sell, to whom, or what it costs.

    What gets built

    • The site rebuilt, with what you sell and who it is for said plainly on the page
    • A plain-text copy of every page, and an index of the pages, for AI assistants
    • Structured data, so search engines read the same facts a person does

    What you keep

    The site, its code and its hosting, in your name.

  2. Visibility in search and in AI answers

    You do not know whether an AI assistant names you when a buyer asks about your category, or what it says when it does.

    What gets built

    • A check of what the AI assistants and the search results say about you today
    • Pages written to answer the questions buyers ask, each one checkable
    • The same check run again on a set rhythm, so you can see what moved

    What you keep

    The pages, and every check, with its date.

  3. Agents for an operational problem

    A job inside the business is done by hand, by someone with better things to do: reading, sorting, re-keying, chasing. It is slow, and it stops when that person is out.

    What gets built

    • The job mapped first: what comes in, what is decided, what goes out
    • Agents that do the reading, the sorting and the first draft, in your tools
    • A person's approval wherever a wrong answer would cost you, and a record of every run

    What you keep

    The agents, their instructions and the record, under your logins.

An AI project that stays a demo, and what replaces it.

AI projects go wrong in three places, and each one is settled before anything is built.

Scope

The project starts from a tool someone saw, so nobody can say when it is finished or whether it worked.

The project starts from one named problem and a written scope: what gets built, what done means, and what it costs.

Trust

An agent is given wide access and left to act, and the first wrong answer ends the project.

What the agents may touch, what waits for a person's yes and where the record is kept are written down before they run.

Ownership

What was built lives in the vendor's software, and stops working when the contract does.

It is built in tools you own, under your logins, and documented as it is built, so it stays with you.

Who does what.

The same rules as our growth work. Agents do the volume, which is why a project like this is affordable. A person answers for what they produce.

Agents do

  • Do the reading, the sorting and the checking
  • Write the first draft: of a page, a record or a reply
  • Run the same check again on a rhythm, and log what they found

A person

  • Scopes the project with you and writes down what done means
  • Decides what the agents may touch
  • Reviews what the agents produce before it reaches you

Waits for your yes

  • The written scope and the quote, before anything is built
  • Anything a customer or a buyer will read
  • Any change to your records, before it is made

How a project is scoped and run.

One call to name the problem, a written scope and quote, then the build, in your tools.

  1. Scope

    The first call

    What it takes from youOne call, and the person who lives with the problem in the room.

    Thirty minutes on the problem: what is slow or missing today, who it costs, and what done would look like. If it is not a fit for us, we say so on the call.

  2. Quote

    After the call

    What it takes from youRead the scope, and say yes or no.

    A written scope: what gets built, what done means, what we need from you, and one price for the project. Nothing is built until you agree it.

  3. Build

    The project

    What it takes from youAccess to the tools involved, and your yes at each approval the scope names.

    We build in your tools, under logins you grant and can revoke, and show you the work as it goes. What it does and how it is run are documented as it is built.

When the project ends, what was built is yours. If you want it run and improved afterwards, that is quoted separately, and it is your choice.

Who it is for, and who it is not.

Three situations these projects are built for, and three we would turn down on the first call.

Built for

One named problem
You can point at the job that is slow, or the page that does not say what you do. A project needs an edge, and a named problem gives it one.
Nobody to build it
Your team knows the problem well and has nobody free to build the fix. We build it with the person who lives with it, so it fits how the work is done.
You want to own it
You want the result in your own tools, documented, and able to run whether or not we are there.

Not for

AI for its own sake
If the goal is to be using AI and the problem is still to be found, a project will not have an end. Find the problem first, then call us.
Training your whole team
We build and run systems. We do not run company-wide AI training or coaching, and a firm that does will serve you better for that.
Growth work
If the problem is pipeline, this is the wrong page. Our five growth services have published prices, and that is where to start.

What buyers ask about AI engineering.

What is AI engineering at WarmUp?

Projects where we use AI to fix a problem outside our five growth services. Today that is three kinds: a website that people and AI assistants can both read, visibility in search and in AI answers, and agents built to take over a slow operational job. Each is scoped as a project and built in tools you own.

How is it priced?

Per project, quoted after the first call. We do not publish a price for this work, because the scope decides it and no two have been alike. The quote is one written figure for a written scope, agreed before anything is built. Our five growth services are different: every one of those prices is published.

Do we have to be a growth client first?

No. Most clients start with growth engineering and ask for this work later, once they have seen what the agents do in their own tools. A project can also stand alone. The first call is the same either way: thirty minutes on the problem.

Is this a sixth service?

No. Growth engineering is what we do first and most: outbound, paid media, LinkedIn content, RevOps and BDR, each with a published price. AI engineering is project work we take on beside it, for a problem that sits outside those five.

What does a website built for AI assistants mean?

A site that says what you sell, to whom and for how much in plain sentences on the page, and gives AI assistants a clean way to read it: a plain-text copy of each page, an index of the pages, and structured data. This site is built that way, so you can look at how it works before you call.

What is AI-answer visibility?

Whether an AI assistant names your company when a buyer asks about your category, and whether what it says is right. We check what the assistants and the search results say today, write the pages that answer the questions buyers ask, and run the check again on a rhythm so you can see what moved.

Can you promise we will show up in AI answers or rank in search?

No. Nobody controls what an AI assistant or a search engine says, and a firm that promises a position is guessing. What we commit to is the work in the scope, a dated check before and after, and a plain report of what changed and what did not.

What kind of operational problem can agents fix?

A job that is mostly reading, sorting, checking or drafting, done by hand today and done the same way each time. We map the job first. If part of it needs a person's judgment, that part stays with a person, and the agents do the work around it.

What can the agents touch, and who approves what they do?

Only what you grant. The agents work in your tools, under logins you can revoke. The scope names what they may change on their own and what waits for a person's yes, and every run leaves a record you can read. Anything a customer or a buyer will read is approved first.

Who owns what gets built?

You do. The site, the pages, the agents, their instructions and the records are built in tools you own, under your logins, and documented as they are built. When the project ends they stay where they are, and they do not depend on us to keep working.

What happens after the project?

What was built is yours and keeps running. Some clients ask us to keep running and improving it, and that is quoted separately. If you would rather your own team take it on, we walk them through the documentation. Neither is a condition of the project.

Do you have results to show for this work?

Not published. No client has cleared a result from this work for the site, and we do not print a number or a name without that. On the first call we can walk you through how this site is built, and through the agents we run for our growth clients.

Tell us the job that is slow.

WarmUp is a growth-engineering firm based in the Midwest. In thirty minutes, name the problem and who it costs, and you leave knowing whether it is a project for us and what we would build first.

Or skip the form