Service 02 · AI Agents · set by the scope

An agent is the new
company website.

Agents built into your operation, wired to your data and tools: research, ops, support, sales. Real work, done by software, with a person on every call that matters. If you’ve been sold an agent before and watched it stall, you weren’t wrong to doubt it. The doubts get answered below.

Let’s talk

Already built something with AI that works in the demo? Getting it to hold in production is a different job: Prototype to Production.

INBOX CRM DOCS AGENT ACTS ASKS LOGGED A PERSON sign-off here safe alone placeholder · the wiring

A chatbot answers.
An agent acts.

An agent is software with hands: it reads your systems, uses your tools, and carries a task from asked to done.

The chat window was AI’s first form, and it still sits there waiting for you to type. An agent doesn’t wait. It watches the inbox, checks the calendar, pulls the record from your CRM, drafts the reply, files the result where your team already works. You give it the job once. It runs the loop every day, and it writes down what it did.

That’s also the part that should give you pause. A chatbot that’s wrong wastes a minute. An agent that’s wrong takes an action in your business.

We use a lot of agents here, to do work: a fleet of specialized agents on an agentic OS runs this practice’s research and delivery every day. What we’d wire into your operation is the system we already live on.

The work they take
off your team’s desk.

Research, ops, support, sales. Every agent shop sells the same four nouns, and the nouns tell you nothing. The wiring is the product: an agent designed around your operation, not the other way around. Here’s what that looks like, lane by lane.

Researchthe field, read overnight

An agent that reads what you don’t have time to: competitors, filings, forums, the papers that matter in your field. You get the brief in the morning, sources attached, in a form your team reads. Nobody on your team opened a tab.

Opsthe swivel-chair work

Every business runs on work between systems: copy this there, reconcile that, file it, schedule it. Nobody was hired to do it, and someone does it all day anyway. An agent wired to those systems does it every day, on time, and logs each move.

Supportanswers with receipts

An agent that drafts answers from your real docs and your real history, not the open internet. It knows which replies it may send alone and which ones wait for a person’s sign-off.

Salesthe follow-up that happens

Most pipelines leak where the follow-up doesn’t happen. An agent watches the deals, enriches the leads, drafts the note while the conversation is still warm. A person still makes the call.

The part that makes it safe

What may it do
alone?

If handing software the keys makes you uncomfortable, good. That instinct is correct, and it is the spec.

Before any agent of ours touches your operation, the answer to that question exists on paper: what it may do by itself, what it has to bring to a person, what gets logged, and who watches it work. Access is least-privilege, in your accounts, under your control. The agent gets the keys it needs and none it doesn’t.

You’ve seen agent demos that ran the whole workflow, in theory. Most stall right here: the demo could act, but nobody had decided what it was allowed to do, so it never got to. Deciding that is most of the build.

Every agent ships with an owner: a named person, a runbook your team actually runs on, an incident plan, and an answer for the day the model or a vendor changes underneath it. An agent without an owner is an orphan with your keys.

What “set by the scope” means

Even AI with hands
needs scoping.

Two agents are never the same size. One reads a feed and writes a brief; another touches your books. A price printed on this page would be a bluff, so the price is set in the scope, in writing, before any build starts. The scope is the product here. It stands whether we build it, someone else builds it, or nobody should.

01
Free Assessment$0 · two days
We read what you have and where the hours go. At this stage, we read the context and leave the code alone. Two days later you have a straight answer: is there an agent in this or not. If not, we say so and you keep the answer.
02
Production Audit$5,000 · within 5 business days
We put the agent in writing: which workflow, which systems it touches, what it may do alone, what done means, and what the build costs. Signed. It stands alone, take it anywhere, including to someone else’s build team. And if the right answer is don’t build this, the scope says that, too.
03
The build, if it’s a goset by the scope
This rung is optional, and it isn’t ours by default. The scope prices and schedules the build: fixed scope, fixed window, acceptance criteria agreed before the iteration starts. Build with us and the loop ends when the work crosses the line, not when everyone is exhausted. Build elsewhere and the acceptance criteria go with you. Either way, the agent runs owned and watched.
Set by the scope means set before the work: the price exists before the build does, and you hold it in writing either way. No meter. The number you signed is the number.

In theory, an agent could run it: read the field overnight, work the queue, chase the follow-up, draft the reply. Let’s find out.

Let’s find out

The free assessment: ten minutes about your operation, a straight answer in two days. $0. Or start with the full AI Opportunity Audit: the Free Assessment and the Production Audit, together.