Details sanitized and shared with appropriate discretion. Specific metrics and artifacts available upon request.
Operationalizing Microsoft Copilot Across the Research Workflow
Embedded Microsoft Copilot into three high-friction ResearchOps workstreams — template creation, SharePoint repository organization, and transcript analysis — cutting time spent on repetitive work and giving the team more hours for actual analysis.
Detailed case study available on request
This project contains confidential process details and internal metrics. The full case study — including methodology decisions, process documentation, stakeholder artifacts, and specific outcome data — is available to hiring managers and recruiters.
Summary
A large share of my team’s time was going to repetitive operational work: building documents from scratch, hunting through a sprawling SharePoint for past research, and manually combing interview transcripts. Because United runs on Microsoft 365, Copilot was available natively and inside our enterprise data-governance boundary — which meant it could safely touch internal documents and transcripts in ways consumer AI tools couldn’t. I designed and rolled out three Copilot-powered workflows — templates, repository organization, and transcript analysis — that reduced operational drag and let researchers spend more time on insight.
Role: Lead UX Researcher · Timeline: 2024–Present [VERIFY] · Team: 5 researchers · Methods: ResearchOps, template systems, knowledge management, AI-assisted analysis
Problem & context
Three recurring time sinks were slowing the team down:
- Document creation — research plans, screeners, discussion guides, and readouts were rebuilt from scratch each time, producing inconsistent formats and wasted hours.
- Findability — the research SharePoint had grown organically into a hard-to-navigate sprawl, so past studies were frequently re-run or simply lost, and dissemination was slow.
- Transcript analysis — first-pass review of interview and session transcripts was almost entirely manual, consuming the time researchers should have spent on synthesis.
The key constraint: any solution had to respect participant privacy and enterprise data policy. That ruled out pasting internal or participant data into consumer AI tools — and made the M365-native, tenant-governed Copilot the right fit.
My role
As Lead UX Researcher I identified the three opportunities, designed each workflow, set the standards and guardrails, and trained the team to use them consistently as part of our standard operating procedures.
Approach & decisions
The deciding factor for Copilot specifically was governance, not novelty: because it operates inside United’s Microsoft 365 tenant, it could work with internal documents, SharePoint content, and Teams/meeting transcripts within our compliance boundary — something a consumer tool couldn’t do safely. I focused it on three workstreams:
- Templates (Word / PowerPoint): Used Copilot, seeded with our best existing exemplars, to generate and standardize a reusable set of research-plan, screener, discussion-guide, and readout templates — so every researcher started from a consistent, on-brand baseline instead of a blank page.
- SharePoint / insights repository: Used Copilot to help restructure the repository’s information architecture, generate consistent summaries and metadata for past findings, and apply a tagging taxonomy — making prior research discoverable and reducing duplicated effort.
- Transcript analysis: Used Copilot for first-pass summarization, theme surfacing, and quote extraction from transcripts — explicitly as a draft, with the researcher validating every theme and quote before it became a finding.
Across all three, the rule was the same: enterprise tenant only, human verification required, no final insight without researcher review.
Process
I audited where the team’s hours were actually going, then piloted each workstream, built out the artifacts (template set, repository taxonomy, transcript workflow), trained the team in working sessions, and embedded the workflows into our SOPs so they’d persist. [PLACEHOLDER: note the pilot order and rollout timeline]
Deliverables
- A standardized, reusable research template library (plans, screeners, guides, readouts)
- A SharePoint information architecture, tagging taxonomy, and metadata/summary conventions for the insights repository
- A transcript-analysis SOP with the human-in-the-loop verification step built in
- A Copilot prompt cheat-sheet for the team [PLACEHOLDER: sanitized examples]
Impact & outcomes
- Sharply reduced time spent on document setup, repository hunting, and first-pass transcript review — redirecting those hours to analysis and stakeholder work.
- Improved findability and reduced duplicated research through a cleaner, well-tagged repository.
- Greater consistency across the team’s deliverables.
- [VERIFY/PLACEHOLDER: insert the specific figures you can defend — e.g., hours saved per study, % reduction in transcript-review time, contribution to the 66% turnaround reduction or 30% efficiency gain. Do not invent a number here.]
Reflection
The repository work delivered compounding returns — every well-tagged study made the next project faster — and in hindsight I’d have invested in the taxonomy even sooner. Naming conventions and template versioning needed governance early to avoid drift. And for transcripts specifically, the human-in-the-loop gate proved essential: AI is excellent at a fast first pass but will confidently misattribute a quote or overstate a theme, so researcher validation was the difference between a time-saver and a liability.