How we actually work
These are the working rules we apply on client systems today, in the order an engagement runs. If something is a goal rather than a habit, it isn't on this page.
The rule behind all of it
AI reads and talks. Deterministic code verifies and calculates. Humans handle exceptions.
Claude reads and drafts
It reads the source documents and proposes the next step in plain language.
Code verifies and calculates
Every amount, date and rule is checked or computed by code. Unsupported values are refused.
People approve the exceptions
A named person approves anything that reaches the system of record.
1. Before we build
Find out what is really there.
Read-only first
We start with read-only access and log every action we take in your account. In NetSuite that means a dedicated role for Claude with only the permissions the work needs, never an administrator role.
Evidence before verdicts
Every keep, simplify or drop call carries its evidence. If NetSuite can't prove a saved search is unused, we write "no recorded UI execution," not "unused."
Sandbox before production
New connections and workflows run in sandbox first, then in production for a small group, expanding role by role.
Mind the shared limits
NetSuite shares one web-services concurrency limit across all your integrations, so heavy AI queries and extracts are scheduled so they don't starve your existing feeds.
2. While we build
Make the system refuse rather than guess.
Refuse instead of guess
If the source document doesn't support a value, the system refuses and routes it to a person. It never fills the gap with a plausible answer.
Code does the math of record
Rents, dates, escalations, totals and accruals are computed and checked by deterministic code against bounded rules. The model drafts; it doesn't do the arithmetic.
Tests and evaluation gates
AI steps ship with a regression test suite and an evaluation set. On the George Urban grader, a release that drifts on accuracy is held until we understand why.
Security review on every change
Pull requests get an automated security review, and every change ships with a backup and a before-and-after check.
3. Before it goes live
Someone other than the builder checks it.
Independent review
A separate review pass, by a reviewer that did not write the work, checks the results and the claims before a client sees them.
Parallel runs on your history
When we replace a script, report or calculation, the new version runs alongside the old one on your own data. It goes live only when there are no unexplained differences.
A client-side release reviewer
On George Urban's platform, a George Urban operator reviews each release before it ships.
Re-tested at every NetSuite release
Roles and connector behavior are re-tested in sandbox before each NetSuite release reaches production.
4. After launch
Keep a person in charge and the record clean.
A named approver
A named person on the client's team approves before anything reaches the system of record. At George Urban, that is their leasing lead.
Exceptions become rules
Every exception a reviewer catches is written up as a proposed rule, so the same mistake is stopped by code next time.
One source of truth
We fold working files and side spreadsheets into one database and one application, so there is a single record everyone trusts.
Drafted, not sent
Messages to customers, vendors or tenants are drafted for a person to send. Nothing goes out automatically.
What we don't do
The short list of things that sound efficient and cause the expensive problems.
- Give an AI an administrator role in your ERP.
- Let a model calculate a number that goes into a ledger, a lease or a filing.
- Send anything on a client's behalf without a person pressing send.
- Put protected health information through the NetSuite AI connector, which Oracle states hasn't been assessed for HIPAA.
- Describe designed features as built. On this site, "in development" means not live yet.
- Call work done before the evidence says it is.
Have a process that runs on copy and paste?
Tell us where your team re-keys data today. We'll tell you plainly whether Claude should do that work, and how we would check it.