Elephant Index
목록으로
AI Nativeness5분EN

AI Nativeness: When AI Becomes an Organizational Capability

Working with AI · Part 1

연재 「AI와 일하는 회사」 2번째 글

한국어로 읽기

If a report that once took an hour can be drafted in ten minutes, how much has the company changed? The team has gained time. A deeper change begins when people examine different evidence, assign decisions differently, or use the outcome of one decision to improve the next.

AI nativeness shows up in decisions

The strategic management class in New Machine Age & IT looks at AI as an organizational capability, from supporting analysis to informing plans and reviewing the results. I pay particular attention to decision quality. Did the team compare meaningful alternatives? Could it tell when an assumption failed? Could it trace that failure back to the original decision? Marco Iansiti and Karim Lakhani describe how putting data and algorithms at the center of operations changes how firms scale and learn. Andrew Ng starts with a manageable pilot whose lessons can inform training and strategy.

What changes in a monthly business review?

AI can summarize sales and draft slides. The team can use that time to compare demand scenarios and decide under what conditions to increase hiring or marketing spending. If it allocates more budget to a market, it can record the expected conversion rate and delivery lead time. At the next meeting it can compare both assumptions with what happened. An inventory team can similarly track stockouts and overstock alongside decision time. A software team might help customers trace budget variances to the underlying transactions, then measure how long that review actually takes.

What do the experiments establish?

Noy and Zhang’s writing experiment reported 40% less time and 18% higher quality scores for participants using ChatGPT. Those participants completed writing tasks. A firm still needs to measure review, approval, and implementation. Dell’Acqua and colleagues found gains on several tasks in a BCG field experiment, alongside a 19-percentage-point drop in the likelihood of a correct answer on a separate task. In a business review, AI can propose explanations for a variance and the team can verify them against transaction records and market evidence.

Noy and Zhang’s writing experiment · Dell’Acqua and colleagues’ BCG field experiment

After drafting gets faster

If sales miss the plan, weaker demand and delayed deliveries call for different responses. Check orders and shipments separately before choosing between additional marketing and a revised delivery schedule.

Workflow and decision points
  1. Draft the report
  2. Review evidence and alternatives
  3. Record the decision and assumptions
  4. Compare outcomes at the next meeting
Report preparation time (minutes)
Manual preparation
60
AI draft only
10
AI draft and evidence review
35

Illustrative figures. The draft takes 10 minutes and evidence review adds 25, leaving a net saving of 25 minutes.

Record the response, supporting records, and a date for revisiting the other explanation. At the next review, check whether orders recovered and whether late shipments went out.

Series index

목록으로