Reported operating use · Education research and evidence synthesis

Reported manual evidence review and table preparation

This example concerns the researcher's preparation workflow, not the educational effects discussed in the studies. The AI evidence-table workspace below is a separate unexecuted design.

Reported manual study review and evidence-table preparation

David Deming, describing his own research preparation

First-person account of manual research and podcast preparation, published 16 November 2026; no AI review deployment or measured workflow gain established.

A published third-party example, not a Nuvepro customer case or evidence of a partnership.

The workflow at a glance

Reported study synthesis and reference-table preparation

Illustrative summary of the published account

Reported work components. Their arrangement does not imply a fixed execution order or simultaneous work.

  • David Deming

    Read and synthesize education studies

    Deming describes reading studies to prepare a podcast series and grouping his synthesis into themes.

  • David Deming

    Check registry status and retain uncertainty

    The account distinguishes confirmed preregistration from cases whose registration timing could not be verified.

  • David Deming

    Assemble a themed linked table

    Deming reports a table with study descriptions and paper links. Its individual entries were not independently inspected here.

  • David Deming

    Invite additions and plan updates

    The author asks for missing research and intends periodic updates; a completed update cycle is not established.

First-person manual research preparation; AI use to perform the review is not established. Search and inclusion rules, extraction protocol and independent review are unspecified. Neither the table nor underlying study outcomes were independently checked; no fixed order is inferred. See the source account and limitations below.

What the source reports

David Deming says he read education studies to prepare a podcast series, grouped findings into themes, examined preregistration status and assembled a linked study table. He notes uncertainty about some registry timing, invites missing research and intends periodic updates. This account describes his synthesis process; it does not establish a systematic-review protocol or AI use in carrying out the review.

Reported technology: Researcher-led reading and synthesis; AI assistance for this review not established. Scope and outcomes are source-reported.

What the enterprise environment would need

The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.

Proposed practice environment

Prepare a cited evidence table with unresolved registry questions

Illustrative project design to scope with your enterprise.

  1. 1 · Set up the environment

    Fictional studies and frozen registry extracts

    A research lead supplies a permitted synthetic study set, dated registry snapshots and scope rules. An independent reviewer authors expected passages, study identities and registry-date relationships before agent extraction.

  2. 2 · Preparation and agent contributions

    Extract a cited table draft

    An agent drafts study descriptions, population, intervention, comparison and outcome fields, passage references and uncertain registry flags. It preserves missing or conflicting evidence and leaves research conclusions to the research lead.

  3. 3 · Independent checks

    Review rows against independently authored references

    An independent reviewer checks source versions, duplicates, population, intervention, comparison, assisted versus later unaided outcomes and preregistration versus data-collection dates against frozen passage and registry references, separately from the generating agent. Missing evidence stays unresolved.

  4. 4 · People decide acceptance

    Research lead accepts the exact table version

    The research lead reviews the checked evidence-table draft and unresolved registry flags, accepts only that exact version and agreed scope, or returns it for repair, independent rechecks and renewed owner review.

Only within the accepted practice scope
Practice output

A reviewable evidence package

A cited evidence-table draft, independent review record and unresolved registry questions. Checks designed, not executed.

If a check fails → withhold acceptance.

An unsupported outcome, duplicate study, changed source version or unconfirmed registration date withholds acceptance. Repair the affected rows or retain explicit uncertainty, rerun independent checks, then return the revised exact version to the research lead.

Failure or owner rejection returns to review

  1. Hold
  2. → 2 · Revise work
  3. → 3 · Rerun independent checks
  4. → 4 · Owner review

Operating boundary. Permitted synthetic materials in a disconnected rehearsal only. No publication, study-outcome validation, automated research conclusions or claim of a comprehensive systematic review. No production operation or human readiness is established.

Proposed, not an executed Nuvepro project or a description of the named company’s implementation. Owners agree the data, rules, responsibilities and acceptance criteria for their enterprise.
Fictional study and registry reference set
Proposed: a research lead agrees an extraction schema and inclusion rules, then freezes synthetic study excerpts, paper versions, registry records and independently authored expected rows. Retain unavailable passages and unresolved registration dates explicitly.
AI drafts a cited evidence table
Proposed: an agent drafts rows linking each extracted population, intervention, comparison and outcome to a supplied passage and version. It flags uncertain preregistration, duplicate reports and missing evidence without deciding that AI improves education.
Independent methods and citation review
Proposed: a separate reviewer checks each field against the fixed study and registry references, distinguishes assisted performance from later unaided learning, compares registration and data-collection dates, and verifies duplicate-study handling. The generating agent cannot validate its own conclusions.
Research lead accepts the exact revision
Proposed: the research lead accepts only the independently checked table or returns it for repair, independent rechecks and renewed review. The researcher retains inclusion judgments and conclusions; the disconnected exercise cannot publish a table or podcast.

A proposed simulation

An illustrative practice design derived from this work, not an executed assessment.

Proposed: prepare a cited evidence-table draft from synthetic study and registry excerpts, preserving uncertain timing, missing passages and researcher-owned conclusions.

Review gate: The research lead accepts only the exact table whose fields, paper versions, duplicate handling and preregistration dates passed independent review. Acceptance does not validate educational effects or authorize publication.

Challenge the result: A paper changes after review, or an uncertain registry timestamp becomes a confirmed preregistration claim. Withdraw acceptance, repair the row, repeat independent source and date checks, and return the exact revision to the research lead.

What needs checking

  • Source boundary: the original describes reading, thematic synthesis, a linked table and uncertain preregistration timing. Search strategy, inclusion criteria, extraction schema and independent review procedures are unspecified.
  • Proposed: test an unavailable paper, a revised paper, two reports of the same study, a registry with no verifiable pre-data-collection timestamp, and outcomes measured with versus without AI assistance. Compare rows with a separately authored reference table. Designed, not executed.
  • Proposed: change a paper version after review or mark unknown registration timing as confirmed. Withhold acceptance, repair affected rows, repeat independent passage/version/date checks and return the revision to the research lead. Designed, not executed.

Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.

How an enterprise could build similar work

A proposed build order for an enterprise doing similar work, written by Nuvepro from the published account and our interpretation of the environment and checks above. It is not how David Deming, describing his own research preparation built or operates this work.

  1. Agree references and extraction rulesProposed: the research lead freezes synthetic study excerpts, registry dates, versions, inclusion rules and expected rows.
  2. Draft cited evidence rowsProposed: an agent extracts supplied fields, attaches passage/version references and flags missing or uncertain evidence.
  3. Independently review methods and citationsProposed: compare fields, duplicate studies, assisted versus unaided outcomes and registry timing with separately authored references.
  4. Repair changed or unsupported rowsProposed: reject stale versions or unsupported registration claims, repair and independently recheck affected rows.
  5. Research lead reviews the exact tableProposed: accept or return the checked draft. Conclusions and publication remain human decisions outside this disconnected exercise.

Your systems, data, rules and accountable owners decide the real project. Agree acceptance criteria, escalation owners and operating responsibility with the people accountable for the work before anything runs on live records. Treat the first run as a rehearsal on test records: compare the draft output with an independently prepared answer, and record what the people reviewing it change.

Questions about this example

Did Deming deploy AI to conduct this review?
The account does not establish that. It describes researcher-led reading, synthesis and table preparation; AI is the research topic.
Were the linked table and underlying studies verified here?
No. Neither the linked table contents nor the underlying study results were independently inspected for this example.
Does this establish a systematic-review protocol?
No. Search, inclusion, extraction and independent-review procedures are not specified, and some preregistration timing remains uncertain.
What would the proposed AI environment test?
Cited extraction, version and duplicate handling, uncertainty and registry-date checks against synthetic reference rows. These checks are designed, not executed; the research lead retains conclusions and publication decisions.

Shape this around your enterprise

Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.

Sources and limitations

Curation review: . This is the date this source example was reviewed, separate from source publication dates.

  • This is a first-person preparation account, not an independently audited review workflow. AI is the subject of the studies; AI use to perform this synthesis is not established.
  • The linked Datawrapper table was not independently inspected, and underlying studies were not reviewed here. This example does not repeat their educational outcome claims or validate the author's conclusions.
  • The account does not specify a search protocol, inclusion rules, extraction schema or independent reviewer checks. Unconfirmed preregistration and missing-study concerns remain unresolved.
  • The 16 November date-only header was treated as seven-day research overlap, not proof of a new deployment or precise rolling-48-hour publication. The proposed synthetic environment and acceptance checks are designed, not executed; no customer or partner relationship is claimed.
  1. The education of an AI skepticSource publication: 2026-11-16

More research and regulated documentation examples

All 11 research and regulated documentation examples

Explore more enterprise work