Finding the knowledge projects need

Technical services firm

Decision studied

Make validated knowledge accessible without depending on a few experts.

The firm, which has 70 employees, had fifteen years of project files. Its 45 professionals were looking for approved clauses, lessons learned and deliverable templates, and answers often depended on two experienced people.

Approach

  1. 01

    Frame

    The director wanted to reduce search time and speed up the preparation of deliverables. The project was limited to approved documents; each professional remains responsible for what they reuse.

  2. 02

    Investigate

    I measured the searches of twelve team members and checked versions, access rights and duplicates, distinguishing a document that was hard to find from information that had never been recorded.

  3. 03

    Explain

    Two causes were compared: poor filing, or knowledge held only by experts. An assistant could address the first problem; the second required interviews and validated documentation.

  4. 04

    Build

    We compared the existing search engine, a reorganized library and an assistant that cites the passages it uses. The status quo included the time spent with experts.

  5. 05

    Test

    Eight users tested 180 questions over four weeks. The thresholds were 95% of answers accurate, sourced and up to date, and 30% time saved. A first pass failed because of outdated versions; after cleanup, 174 of 180 answers met the criteria and search time fell by 35%.

  6. 06

    Decide

    The director released $30,000, including $8,000 for strategy. Assisted search started on a limited document collection. A document manager keeps the updates current and checks access rights.

  7. 07

    Steer

    The firm tracks search time, source validity and the additional hours actually billed to clients. Added availability with no client demand is recorded as capacity, with no associated revenue.

Test results

Compliant answers: 96.7%174 answers out of 180 · threshold: 95%Search time: −35%Threshold: −30%
Two indicators from the four-week test with eight users: compliance of answers (outer ring) and reduction in search time (inner ring). Lines mark the thresholds set.

What AI contributed

AI helped me spot duplicates and prepare the evaluation questions. The internal assistant finds passages, cites their version and flags questions with no documented answer.

Business model (projection)

The projection assumes that 600 freed-up hours become additional services sold, at $95 of contribution per hour after variable costs, or $57,000. This conversion depends on available work. The prudent scenario assumes 300 hours sold.

Annual projection for this case
Initial investment$30,000 CAD
Annual gains$57,000 CAD
Costs to subtract$12,000 CAD
Net annual gain$45,000 CAD
Simple payback8 months (at steady state)
Prudent scenario$16,500 CAD net, payback in 21.8 months

Annual projection built from the results of the engagement. It is not a guarantee of performance.

A comparable challenge?

Describe the decision to make and the data available. I will suggest an approach to prepare it and the evidence to gather before deciding.

Let's talk strategy