Find out whether it’s worth doing yet
Five questions we ask before recommending anything — whether or not AI is involved. You get back a straight answer: what your weakest link is, and whether the right next move is to build something, clean something up first, or wait. The rules we use are published below, so you can apply them yourself.
This is a specialty check, not the front door. If you already know something in your operation is expensive but not what, start with the Systems Assessment — it maps how the work actually moves, puts numbers on where it leaks, and ranks what is worth fixing. Automation is one possible answer it produces. So is connecting two systems you already pay for, and so is leaving something alone.
Most of these projects fail before anyone chooses a tool
The pattern is consistent across analyst research and the implementations we have seen on the ground, and it is not specific to AI. Automation gets added to workflows that were never defined, verified against examples that do not exist, and deployed without an owner who can sign off when things drift. AI is simply where it costs the most, because a deterministic script that is wrong is wrong the same way every time — and a model that is wrong is plausible.
This check is what prevents that. Before we recommend anything, we look at whether the foundation is actually there. Five questions, each scored 0–5. Fixed rules — not a judgement call — produce the answer and the recommended next step.
It is a deliberately unglamorous step. It also regularly saves people money, because the honest answer is sometimes “fix this one thing yourself and you will not need us.”
Every question has signals you can check yourself in fifteen minutes
No framework, no borrowed methodology. These are the five things that decide whether a change to how your business runs will stick, and the signals are exactly what we look at.
Is the payoff real?
Is there a defined desired outcome, a real bottleneck, and a budget/timeline signal that makes a real engagement plausible?
- ✓Outcome stated as a measurable sentence (time, dollars, or rate)
- ✓Bottleneck described in concrete operational terms
- ✓Budget and timeline are realistic, not aspirational
Is the work defined?
Can the workflow be drawn on a whiteboard in 15 minutes? Are SOPs in place? Is there a named owner?
- ✓Inputs, outputs, owners, and exception paths are documented
- ✓SOPs exist and are current
- ✓Two staff members describe the workflow the same way
Is the information there?
Single source of truth, data quality, where institutional knowledge lives, and whether known-good examples exist for validation.
- ✓One clean source of truth for the data this workflow uses
- ✓At least 20 past cases exist on record somewhere we can reach
- ✓Knowledge lives in a documented source, not one person's head
Can the systems connect safely?
Are the systems wired together (or wirable) and is data sensitivity manageable with appropriate controls?
- ✓Required systems have integrations or accessible APIs
- ✓Data sensitivity is classified and handling rules are documented
- ✓Permissions and access controls are explicit
Is someone watching it?
Is there a named approval owner? Is there any error tracking today? Is the org disciplined enough to keep validation gates in place after launch?
- ✓Named human approval owner with a weekly cadence
- ✓Errors and exceptions are tracked, not just discovered when they hurt
- ✓Validation gates and acceptance criteria are written down
Published rules, not a judgement call
The result comes from fixed rules applied to the five scores. The rules are printed below so you can apply them to your own business without us. There is nothing hidden in the scoring.
Not ready
Recommended pathAt least one foundation domain is critically weak. AI will produce fluent, plausible, and quietly wrong output that erodes customer trust. Document the foundation work first.
Cleanup first
Recommended pathReal interest, real pain, but the foundation is not yet stable enough to safely automate. A 1–2 week cleanup focused on the weakest domain comes before any AI work.
Pilot ready
Recommended pathThe foundation is good enough to build one bounded workflow. We check it against a hundred of your own past cases before it goes live — pulled from your records by us, not gathered by you — and decide on evidence after 30 days.
Build ready
Recommended pathAll five domains are strong. Multiple workflows can be sequenced through the same validation discipline. The bottleneck is implementation capacity, not readiness.
Order of evaluation: rules are evaluated top to bottom. The first match wins. If no rule matches but readiness is reasonable, the default is “cleanup first” — we do not advance a workflow into a pilot unless the criteria are explicitly met.
An answer, not a sales pitch
- →Readiness score across the five domains, plus the total out of 100.
- →Weakest domain identified — the constraint that determines what to fix first.
- →Recommended path: do-not-automate-yet, cleanup sprint, single-workflow pilot, or multi-workflow build.
- →Readiness roadmap — concrete foundation work to do before (or alongside) any AI implementation.
- →Do-not-automate-yet list — what should stay human-approved at the current readiness level.
- →Validation gates — the acceptance criteria, approval owner, and measurement plan that come before scale.
The 12-minute deep-dive on why readiness comes before tool selection, what each domain measures, and what happens after scoring.
Want your own answer?
Tell us what is breaking. Brent reads every submission personally before anything goes back to you — including the ones where the answer is ‘not yet’.