How I Use AI in My Evaluation Work
AI has become part of my daily workflow as a practical support system that helps me manage the many moving parts of federal programs and evaluation projects. I am an internal evaluator dedicated to the evaluation of federally funded programs. That means every logic model, analysis plan, report, tool, survey, literature scan, and statistical decision flows through one set of hands. it stand out
Because of that, AI has become part of my daily job; every morning after turning my computer on, I open Microsoft Outlook and ChatGPT to start my day. AI allows me to work efficiently across multiple projects at once while still maintaining the depth and rigor that evaluation requires. Over the past year, I’ve built several custom GPTs designed specifically for the work I do in Title I, migrant education, family engagement, teacher incentives, and district-level evaluations. Each one supports a different part of the process, and together they help me focus my time on the parts of evaluation that rely on reasoning, interpretation, and professional judgment.
So here is how I use AI in practice.
1. Researching evidence-based interventions
One of the most time-consuming tasks in federal programs is researching to verify whether a proposed intervention meets ESSA evidence tiers. AI helps me quickly scan peer-reviewed studies and identify:
if they are listed in the What Works Clearinghouse (WWC) or Evidence for ESSA website,
the evidence tier,
what outcomes the intervention improves,
the student populations included,
and any cautions or limits in the research.
And when interventions are not listed in WWC, AI helps me identify what other studies have been done and published on peer reviewed journals.
This gives me an initial screening of whether a program is aligned with ESSA expectations before we move forward. I still verify the findings myself, but AI speeds up the groundwork.
2. Developing tailored data-collection tools
AI supports the early drafting of:
surveys,
interview guides,
implementation checklists,
and observation templates.
These drafts are customized to the federal program context and then refined through my evaluator lens. The structure, clarity, and alignment with federal expectations still depend on human judgment. AI accelerates the setup so I can put more time into strengthening the content.
3. Statistical guidance for ongoing projects
This one especially significant: I use AI to think through analyses such as:
which statistical method fits the question,
assumptions to check,
strengths and limits of each approach,
and how the available variables align with the proposed method.
This is especially helpful when I am planning a new evaluation or comparing multiple analytical paths before selecting the one that best fits. In the past, I would spend a lot of time thinking and going back to my statistics’ books to refresh my memory about the techniques, assumptions and conditions; AI drills down the methods and I make the decision about what to use based on the evaluation project and the available data.
4. Literature reviews to understand programs and interventions
Before evaluating a new program, I need to learn what I am evaluating. This is an essential step anyone conducting evaluations do. In this case, I use AI to gather:
key definitions across studies,
outcome patterns,
measurement approaches,
theoretical foundations,
and gaps in the research.
AI gives me a map of the landscape. I then validate, expand, and refine it through my own search process. It shortens the time needed to understand the intervention well enough to begin evaluating it.
5. Translating statistical language for stakeholders
Not everyone I work with is a statistician, and they shouldn’t have to be. I cannot start statements with “statistical significance”, “confidence intervals” or “effect sizes”; I need to translate statistical conclusions into language that fits the program context. AI helps me take:
model outputs,
assumptions,
statistical caveats,
quantitative findings,
and turn them into explanations that program administrators and school leaders can understand. This supports transparency and helps ensure that decisions are grounded in understanding rather than confusion.
AI helps me work faster, but it does not think for me.
It does not:
• define evaluation questions,
• choose methods,
• interpret meaning,
• understand context,
• or make decisions.
Those responsibilities remain with the evaluator.
AI supports the work; it does not replace the reasoning behind it.
For me, AI has simply become another tool in the evaluator’s toolkit. It lightens the load, sharpens the workflow, and helps me stay organized across many programs. But the thinking, interpretation, and professional judgment stay human.
