AI implementation readiness assessment: 16 criteria to evaluate enterprise gaps
AI implementations often stumble not from weak models but from data that cannot be extracted, insufficient process change, or compliance constraints that eliminate the plan entirely. These obstacles remain invisible during demonstrations and emerge only at launch. This assessment uses 16 scored criteria across four dimensions (data, process, team, security) to show what your enterprise must address before deploying an AI system your staff will actually use.
By
Tenten AI FDE 團隊
導入方法論
Published
October 8, 2025
Read time
5 分鐘
An AI readiness assessment measures your enterprise before you begin, showing how far you are from a functioning AI system in production that your staff actually uses. Rather than asking whether to pursue AI, it asks whether your organization as currently structured can sustain adoption or will create another unused tool.
We run this assessment with clients early in most engagements. In practice, most obstacles come not from weak models but from data that cannot be extracted, process changes stakeholders will not adopt, or a compliance constraint that eliminates the entire plan. Demos never surface these issues. They emerge immediately when the system goes live.
Why four dimensions and sixteen criteria
A functioning AI system requires four elements: accessible data, integration into actual workflows, someone driving adoption, and security and compliance clearance. Failure in any area prevents deployment. The assessment evaluates each of these four dimensions with four criteria. This yields sixteen scored items, each rated 0 to 5, then combined by their weighted contributions into a final score.
Weights vary by dimension. Data accessibility receives 30% weight because insufficient data makes everything else irrelevant. In regulated industries like finance and healthcare, security carries veto power: low security scores prevent production deployment regardless of other dimensions.
| Dimension (Weight) | Criterion | What a 0-Score Looks Like |
|---|---|---|
| Data (30%) | Data Accessibility | Critical data locked in another system; requires manual export |
| Data Quality and Consistency | Same field recorded three different ways; duplicates; missing values | |
| Access Control and Governance | No classification, either everyone sees everything or no one sees anything | |
| Unstructured Knowledge Retrievable | Contracts, tickets, PDFs scattered in folders with no indexing | |
| Process (25%) | Scenario Clarity | Target is 'the whole company uses AI,' with no narrow starting case |
| Current State Baseline Measurable | Can't say how long one transaction takes or how often it fails | |
| Decision Points and Human Review | Undefined: what goes to AI, what requires human sign-off | |
| Exception Handling Path | No handoff mechanism when the system lacks confidence | |
| Team (25%) | Business Owner in Place | No one accountable for the adoption rate KPI |
| Frontline User Buy-In | The actual operators think it's a waste | |
| Technical Integration Point | IT or data engineering can't make room in the roadmap | |
| Change Management Capacity | No one assigned to training and rewriting procedures | |
| Security (20%) | Data Residency and Sovereignty | Unclear if data can leave the region or whether private deployment is required |
| Access Control and Audit Trail | No logs; can't trace who accessed what | |
| Regulatory Compliance Mapped | Privacy law, financial regulator rules, healthcare standards not yet itemized | |
| Vendor and Model Risk | Haven't read third-party model terms on data handling |
How to calculate and interpret
Each criterion receives a score from 0 to 5, multiplied by that dimension's weight, then converted back to a 0-5 scale. If data criteria average 3.0, process 2.0, team 3.5, and security 4.0, the composite score is 3.0×0.3 + 2.0×0.25 + 3.5×0.25 + 4.0×0.2 = 3.075.
Your next steps depend on your score:
- Below 2.5 - Stay off production. Invest one to two months in establishing data foundations and process baselines. Deployment now produces a functional demo with negligible adoption.
- 2.5 to 3.5 - Conduct a pilot on a single scenario. Verify that a complete workflow operates and users engage with it.
- 3.5 to 4.5 - You can deploy, but treat your lowest-scoring dimension as your pre-launch priority. Weakness there will limit adoption.
- 4.5 and above - You can expand across multiple scenarios.
A high composite score does not guarantee readiness. Your lowest dimension matters most. One manufacturing client posted 4.0 for data and process but 1.5 for team because shop floor staff rejected the system. The total score seemed adequate, but adoption fell to single digits within a week of launch. The score's utility is in pointing to which dimension requires work.
From scores to action
The assessment converts general concerns into specific work items: data that needs cleaning, process decisions that need clarification, missing team roles, and unmet compliance requirements. Each low score indicates a task to complete before deployment.
At Tenten, we use this sixteen-criterion scorecard to guide our first week of deployment work. Engineers prioritize strengthening the lowest-scoring dimension first, not moving directly to model development. Technical polish matters less than achieving deployment and sustained use.
One stuck workflow
is enough to begin
Tell us what the team does today, where it breaks down, and what a better working day should look like.