The PromptLens field guide
How to assess AI adoption beyond seat counts
A practical checklist for AI adoption, operations, and enablement leaders who want to understand what is changing in everyday work.
Download the assessment checklist ↓1. Start with the decision you need to make
An AI adoption assessment examines who has access, how consistently tools are used, and what that activity means for a specific workflow. It should help you decide where to provide support, what to test, or what evidence to collect next.
Start with one team and one question. For example: “Would a shared drafting approach help our support team produce replies that need less editing?” This gives the analysis a practical purpose. A broad request to prove AI ROI usually combines too many questions for a first assessment.
Record the task, the people involved, the tool, the reporting period, and the decision owner. Confirm that the data can answer the question before choosing a metric.
2. Separate access, adoption, and outcomes
These measures describe different things. Define the denominator and reporting period so a percentage can be interpreted correctly.
| Measure | Working definition | What to watch for |
|---|---|---|
| Access coverage | Licensed people ÷ people in the chosen team | Low usage may reflect limited access rather than a training problem. |
| Active use | People with recorded activity ÷ licensed people, in a defined period | A login or one event does not show proficiency. Define which activity counts. |
| Repeat use | Active people meeting a chosen frequency threshold ÷ active people | Choose a threshold that fits the task. Infrequent tasks may still benefit from AI. |
| Workflow quality | Outputs meeting an agreed review rubric ÷ outputs reviewed | Use comparable tasks and a documented rubric, including errors and rework. |
| Editing effort | Time spent reviewing and revising a defined task | Include verification time. Differences in task difficulty can distort comparisons. |
Tool reports may define “active” differently and expose different levels of detail. Document the definition from each source. Do not add platform user counts together without checking overlap, and do not infer time saved from prompt counts.
3. Work through a small example
Hypothetical example, not customer results
A support team has 60 people. Of those, 40 have licenses, 28 recorded activity during one week, and 14 used the tool on at least three days.
- Access coverage: 40 ÷ 60 = 67%, rounded.
- Active use among licensed people: 28 ÷ 40 = 70%.
- Active use across the whole team: 28 ÷ 60 = 47%, rounded.
- Repeat use among active people: 14 ÷ 28 = 50%.
The same activity can produce different percentages depending on the denominator. None of these numbers establishes productivity or explains why people use the tool less often. The next step is to investigate access, task fit, and team feedback, then review actual work quality using approved samples.
4. Prepare your assessment
Use this checklist in your planning conversation. In a PromptLens pilot, we lead this preparation and identify what your team needs to approve.
- Decision: Name the workflow question, the decision owner, and the action the findings could inform.
- Scope: Identify the team, tools, reporting period, and any access changes during that period.
- Available data: List approved exports, activity definitions, timestamps, and gaps in coverage. Confirm that exports are available on your plan.
- People and privacy: Review employee communication, access permissions, reporting groups, retention, and deletion before collecting data.
- Baseline: Choose comparable tasks and define quality and effort measures before changing the workflow.
- Context: Record task complexity, workload, tool changes, and other factors that may explain a pattern.
- Review: Set a review date and name the person who will decide whether to adapt, expand, or stop the experiment.
Start with the minimum data needed. Do not send employee conversations, credentials, personal information, or company exports through the website inquiry form.
5. Turn a finding into a coaching experiment
A useful experiment changes one part of a workflow and sets out how the team will judge it. For support replies, you might test a shared drafting prompt with approved tone and response examples.
- Define the hypothesis. For example, a consistent structure might reduce avoidable rewriting while preserving response quality.
- Choose the comparison. Review a baseline sample and a subsequent sample of similar tasks. Record differences in difficulty and reviewer experience.
- Measure both quality and effort. Include factual accuracy, completeness, tone, editing effort, and verification time. Record failures as well as successes.
- Set a decision rule. Decide what evidence would justify continuing, adapting, or stopping. Avoid universal percentage targets that have no basis in your workflow.
A before-and-after improvement is a signal to investigate. It does not by itself show that AI caused the change. Stronger claims require a study design that addresses other explanations.
What should you leave with?
A useful assessment produces a baseline with clear definitions, findings supported by the available evidence, a record of limitations, and a prioritized coaching plan. Each recommended experiment should have an owner, a measure, and a review date.
You do not need a complex dashboard to begin. You need a clear question and a measurement approach the team can interpret and act on.
A guided next step
We can lead the assessment.
PromptLens shapes the pilot, reviews available data, interprets the findings, and builds a practical coaching plan. Your team provides context and approves the scope and data access.
Explore a PromptLens pilot ↗