Authors could reuse the research history instead of rebuilding it for each report.
The source material included survey outputs, spreadsheets, and published reports from different research waves. Question labels and customer-group definitions had changed over time, so matching two similar labels didn't establish that their results could be compared.
We organized the data around the question, wave, segment, and value, with a link back to the source. Keeping definitions with the relevant wave gave the authors a way to distinguish a change in the research from a change in the population or question being measured.
That record was the input to the drafting workflow. The report archive could supply examples of the company's writing and prior framing, while the structured research data supplied the figures and comparisons.
Drafts arrived with the evidence authors needed to review them.
We worked with senior authors to document the rules they used for valid trends, sample sizes, statistical significance, and recommendations. Those rules gave us a standard against which to check the assistant's output.
The assistant prepared drafts and supporting tables from the shared record. Source links let authors inspect the evidence behind a figure, and automated checks highlighted problems for review. Where the data didn't support a figure, the workflow could leave a gap for the author to resolve instead of supplying an unsupported number.
The author reviewed the draft, its supporting evidence, and the check results before signing off. Automated checks supported that review while the author kept responsibility for the research.
New research could enter the shared record instead of another disconnected file.
A historical cleanup would lose value if the next research wave introduced another unmapped label or changed definition. We therefore built an intake process that mapped new research into the shared record and flagged changes for the operating team to resolve.
We connected the publishing workflow to the existing subscriber platform. The company could keep its client delivery channel while changing the data preparation and drafting work behind it.
We trained the team to keep new research moving through the system.
We trained the operating team to maintain the inputs and trained authors to review drafts against the company's methods. The handover covered recurring tasks such as adding research, resolving a changed definition, and checking the material before publication.
That training gave the team procedures for keeping the data usable and the research under review as new waves arrived.
More paid work contributed to $1.6M more EBITDA in the year after delivery.
The company generated $2.9M more revenue and $1.6M more EBITDA with the same author team in the twelve months after the engagement ended in July 2025. A partner later confirmed the comparison: revenue increased from $16.4M to $19.3M, and EBITDA from $3.6M to $5.2M. Revenue and EBITDA describe different parts of the same business activity, so the two gains aren't additive project value.
The CFO credited the production changes and new client-specific briefs with most of the earnings increase, while also recognizing existing contractual price increases. The operating mechanism was released author capacity being converted into paid work by the client's sales team. The financial comparison does not isolate the effect of one software component or establish a net project return after all engagement costs.