Chat with us
← All case studies

Workflows / AI platform & research

Get a broken workflow working again.

Transcript recovery, API troubleshooting and structured research outputs, alongside AI-agent development and testing.

ContextConsulting engagement
Delivery stageDelivered engagement
FocusUsable research data delivered
Illustration: Interview archive to Recover & extract to Transform to Research outputInterview archiveRecover & extractTransformResearch output

Process illustration. Recreated for this portfolio; no client interface or data shown.

  1. 01Interview archive
  2. 02Recover & extract
  3. 03Transform
  4. 04Research output

01 / The starting point

The problem to solve.

A research project needed usable interview material from a transcription service. Recurring automation and API failures stood between the original interviews and the structured outputs required by the research team.

02 / My contribution

Where I came in.

I recovered historical transcripts, built and adjusted Make workflows, investigated integration failures and delivered structured research data. A separate part of the engagement involved building and testing catalogue AI agents.

03 / Implementation

What I worked on.

  1. 01

    Extraction and transformation workflows connecting the transcription service to the research output.

  2. 02

    Troubleshooting across API and automation modules, including coordination with provider support and another consultant.

  3. 03

    Agent prompt and behaviour refinement, scenario testing and QA feedback on team-built agents.

A design decision that mattered

Separate a plausible AI answer from a usable deliverable. Workflows needed consistent structure and agents needed testing across different inputs.

04 / Delivery

Usable research data delivered.

Recovered interview material and structured outputs delivered for the research team, alongside integration fixes and tested AI-agent behaviour.

Delivered as a consulting engagement, in collaboration with the client team and integration providers.

Could this apply to your team?

A similar problem,
your own context.

Teams with a broken integration, inaccessible historical data or an AI workflow that still needs too much manual correction.

Discuss something similar ↗

Your next project

What needs to
work better?

Tell me about the process, the people using it and the outcome you need. We can work out a useful first step.

Discuss your project