Choosing the first useful AI uses

Industrial group

At an industrial plant, several teams were asking for AI tools. The engagement was to choose the first uses to fund.

Approach

1. 01

   Frame

   The task was to choose two uses to deploy, with an initial investment of $63,000 CAD. Time saved was not enough: avoidable costs and quality also had to be documented.

2. 02

   Investigate

   The inventory of five activities revealed external preparation costs, errors in document matching and tasks whose value remained hard to measure.

3. 03

   Explain

   A lack of tools and poorly defined processes were examined. Some teams mainly needed shared rules before receiving an additional tool.

4. 04

   Build

   We compared a general assistant for everyone, two targeted uses, and process modernization without AI, with the existing situation as the economic and quality baseline.

5. 05

   Test

   Three uses were tested on controlled historical files. The thresholds set in advance were 25% less time and no increase in significant errors. Two uses reached 30% and 35% reductions; the third exceeded the accepted error level.

6. 06

   Decide

   Management selected the help with document preparation and the matching of purchase documents. The third use was stopped, and human validations were kept.

7. 07

   Steer

   At six months, external spending and avoidable error costs had decreased less than expected. Management adjusted the forecast and required two additional months of data before any extension.