Profile Machine is Slalom’s internal platform where consultants maintain profiles used for staffing and client pitches. It centralises work experience, skills, and roles across 200+ consultants and five practice areas.
Consultants typically write project descriptions quickly between billable work. These entries are later pulled by senior leaders during live client pitches to showcase relevant expertise.
The problem was not access. It was quality.
Senior leaders repeatedly encountered vague, task-level descriptions such as “Worked on data migration project” or “Improved processes.” During pitches, this forced an immediate rewrite loop. Leader flags issue. Consultant rewrites between client work. Leader reviews again. Pitch momentum is lost.
This happened constantly. Business development estimated the “profile quality tax” at ~40 hours per month in rewrites and reviews.
Quality varied wildly across practice areas. There was no shared definition of what “client-ready” meant, no quality gate before profiles were used, and feedback always arrived too late.
A senior leader approached me with an initial idea. Use AI to validate and rewrite project descriptions.
I led the initiative end to end as the sole designer, working with three engineers and stakeholders across five capabilities. I owned experience design, workflow definition, evaluation logic, prompt strategy, and LLM evaluation. My role extended beyond UI into decision-making around uncertainty, cost, and reliability.
Enable our in-house consultants to create client-ready profiles without senior-leader intervention, while keeping cost, accuracy, and control predictable.
The initial concept proposed a separate validation screen where ChatGPT would score content in one panel and generate suggestions in another.
This failed for four reasons: