Ninsei LabsAn AI readiness assessment answers one question before you commit budget: do your current operations give automation something real to work with? For a small business, the answer usually isn't binary. It's a score across four dimensions, and each dimension has a floor below which AI reliably fails regardless of how good the tooling is.
Most assessments you'll find online were built for enterprise procurement teams. They ask about "organizational AI maturity" and recommend 18-month transformation roadmaps. This one is built for operators running 5-to-25-person companies who want to know if they should spend $8,000 to $15,000 on a workflow automation project right now, or sequence some groundwork first.
Why AI Readiness Is Not About the Technology
The tooling has never been more accessible. n8n, Make, and purpose-built AI agents can handle complex multi-step workflows at costs that were unthinkable three years ago. The failure mode isn't picking the wrong tool. It's automating a process that isn't stable enough to automate.
When an automation project fails, the post-mortem almost always points to one of four things: the process changed mid-build and nobody told the dev, the source data was inconsistent so the model kept producing garbage outputs, the integrations required screen-scraping because there was no API, or the team never had bandwidth to supervise the ramp-up and the whole thing got abandoned after the demo.
These are readiness failures, not technology failures.
The Four Dimensions of an AI Readiness Assessment
Score yourself 0 to 3 on each. Honest scores only.
1. Process Stability
- 0: The process exists in someone's head. Steps vary by who does it.
- 1: The process is partially documented but has exceptions that aren't written down.
- 2: The process is documented, consistent, and most edge cases are handled explicitly.
- 3: The process runs the same way every time, inputs are defined, outputs are defined, and you can describe it to a new hire in writing.
You cannot automate tribal knowledge. A process that relies on undocumented judgment calls will produce an automation that fails on exactly the cases that matter most.
2. Data Quality
- 0: Relevant data is scattered across spreadsheets, emails, and PDF attachments. No consistent format.
- 1: Data lives in a system but fields are inconsistently filled, naming conventions vary, and deduplication hasn't been done.
- 2: Core data is in a structured system, mostly clean, with some gaps.
- 3: Data is structured, consistently populated, accessible via export or API, and someone owns data hygiene.
"We have data" is not the same as "we have usable data." An AI reading your CRM contacts will produce proportionally better output when those contacts have complete, consistent records. One client I worked with discovered their CRM had four different spellings of the same company name across 800 records. The automation worked. The outputs were a mess until we cleaned the source.
3. Integration Access
- 0: The systems you need to automate don't have APIs or native integrations. You're looking at scraping or manual export/import.
- 1: Some systems have APIs but require custom development to connect. Auth is complex or undocumented.
- 2: Most systems have native integrations or documented APIs. Some legwork required.
- 3: Your core systems (CRM, email, calendar, project management) have first-class API access or are already connected in a platform like Make or n8n.
Fragile integrations are the most common reason an automation breaks in month two. If the connection relies on a browser session or an unofficial API, budget for ongoing maintenance from day one.
4. Change-Management Capacity
- 0: Nobody has been assigned to own this project. The team is at full capacity.
- 1: There's an owner in name, but they have no dedicated time. The team is skeptical.
- 2: One person has 20-30% of their time available to oversee the ramp. Team is neutral to curious.
- 3: An owner has dedicated time, the team is bought in or at least willing to test it, and there's a clear escalation path when something goes wrong.
AI automation is not a "set and forget" deployment, at least not initially. Someone has to review edge cases, catch failure modes, and push the feedback loop back to whoever built it. Teams that skip this step end up with an automation nobody trusts and everyone routes around.
How to Interpret Your AI Readiness Score
| Total | What It Means |
|---|---|
| 10-12 | Ready. A scoped automation project will likely succeed. |
| 7-9 | Conditionally ready. Identify which dimension scored lowest and fix it before contracting work. |
| 4-6 | Not ready yet. Sequence matters: process stability first, then data, then integration. |
| 0-3 | Stop. Budget spent here goes to waste. Fix foundations first. |
The threshold that matters most: if any single dimension scores 0, stop. A zero in any category is a blocker, not just a drag on the total. You can score a 9 overall and still fail if nobody owns the implementation.
The Score That Gets Underrated Most
Change-management capacity is the dimension operators consistently underrate on the self-assessment. A 1 here has killed more technically successful automation projects than bad data or broken integrations combined.
Here's what it looks like in practice. A company invests in a client intake automation. It works. The demo goes well. Three months later, the team is still copy-pasting form responses into the CRM manually because the person supposed to supervise the automated version "never had time to learn the new flow." The automation runs fine. Nobody uses it.
The tool wasn't the problem. The adoption ramp had no owner.
If your team is already stretched and you're thinking about AI automation as a way to get out from under the workload, the math only works if you have capacity to absorb the ramp before the capacity returns. That ramp is typically four to eight weeks for a single-workflow automation, longer for anything touching multiple systems.
How to Use This Assessment When Evaluating Vendors
Run this diagnostic before you talk to anyone. A credible operator will ask you some version of these questions during discovery. If they skip straight to scoping and pricing, that's a signal worth paying attention to.
The assessment also gives you a real negotiating frame. A score of 7 with a clear weakness in data quality means you can go to a vendor and say: "We know our CRM is messy. What's the right sequence, and what does it actually cost to fix the data layer before we build on top of it?" That's a more productive conversation than "here's our budget, what can you do?"
At Ninsei, the first thing we do with a new client is work through exactly this: where are you on each dimension, what's the gap, and is an automation project the right next move or should we sequence something else first. Sometimes the honest answer is not yet.
That answer saves both sides time, and it's the one the enterprise-maturity frameworks are too expensive to give you.