The new standard

What is Resident AI?

Almost every AI sold to a company is bolted on. It sits beside the business, reached through a window of its own, fed whatever someone remembered to paste into it, running on a machine somebody else owns.

That is an architecture, not bad luck.
The failures that follow are predictable from it.

↓

Bolted on

┌── company ──────┐   ┌── vendor ──┐
│ tools    data   │   │            │
│ people machines │──▶│     AI     │
│                 │◀──│            │
└─────────────────┘   └────────────┘

Resident

┌── company ───────────────────────┐
│  tools    data     ┌──────────┐  │
│  people machines ──│    AI    │  │
│                    └──────────┘  │
└──────────────────────────────────┘

Resident AI is the other architecture. It lives inside the company that owns it. Same tools, same data, same machines, same people doing the same jobs.

That is the standard.
Not an assistant people visit. Something present.

The test

The four conditions

An AI is resident when all four are true. Fail one and it is bolted on.

Tick the conditions the AI in your building meets.

01 It runs where the work already happens. No second window, no new login, no habit anyone has to be trained into. Bolted-on AI asks people to go somewhere to use it, and adoption dies quietly a few weeks after the pilot team stops trying. 02 It sees the whole operation. Not the slice someone thought to paste in. An AI with no data will ask a question. An AI with partial data will answer, and be confidently wrong. The second is worse, because people trust it once, get burned, and then stop trusting the system at all. 03 It runs on the company’s own infrastructure. The company’s hardware, network and security perimeter. No vendor-hosted service in the middle and no copy of the operating record on the vendor’s side. This is a procurement decision: a full local setup runs on commodity hardware, well under the price of a year of licenses for the same headcount. 04 What it learns belongs to the company. That understanding lives in a store the company owns and can export in a format it chooses. Bolted-on AI leaves the company renting access to a picture of itself.

Four out of four is resident. Three out of four is bolted on with better marketing.

Why an existing company can meet this standard

AI-native

The usual phrase for this is AI-native, and it locks out almost every company worth talking to. Native means born that way. A 25-year-old firm doing $200M was not born that way and never will be, so the phrase tells them they are the wrong kind of company before the conversation starts.

Resident

Resident is not a birth condition. It is a state a company moves into, in place, without becoming a different company first.

Nothing in the four conditions requires reorganization. Two of them are statements about where software runs and what it can reach. Two are procurement decisions. None asks anyone to work differently, and a standard that required people to work differently would fail the first condition anyway.

What the conditions rule out

They rule out most of what is currently sold, including software that is good at its job.

A chat product with an excellent interface fails the first condition.
A platform that has to ingest your systems into its own environment to be useful fails the third.
A well-integrated agent whose accumulated context cannot leave fails the fourth.

They also rule out work we do ourselves. A deployment that ends up beside the business has failed the first condition, and the fact that we built it does not change the result. A definition that only ever disqualifies other people is a marketing claim wearing a lab coat.

So the conditions are written to be checked, including against us.

What to do with this

Run the four conditions against whatever AI is in the building now. The one that fails will name the problem the company already has:

01 the pilot nobody uses
02 the system nobody trusts
03 the bill nobody forecast
04 the picture of the business that cannot leave the vendor

We build resident AI for companies that already exist.

Change nothing. Transform everything.

How that works in practice →