Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

Adaptation · Adopters

Customer-support productivity uplift from a generative-AI assistant

consistent with normal75/95 confidence, good evidence, some ambiguitygrade Bcoincidentrefreshed irregularly: A one-off field experiment; refreshed only when a comparable randomised study publishes.

1What this measures

Change in issues resolved per hour when support agents get a generative-AI assistant, from the Brynjolfsson, Li and Raymond field experiment (NBER WP 31161), with the novice effect and MIT Sloan's restatement shown alongside.

Why it matters. The best-identified productivity effect on record for a deployed assistant; the normal-technology story predicts modest average gains concentrated in novices, which is exactly what it found.

Proxy types
deployment, welfare
Unit
share
Cadence
annual
Valve
adoption to adaptation

2How we track this

Normal band
≤ 20.0%
Fast band
≥ 50.0%
Falsifying
—

Normal = an average uplift under 20%, the range prior process tools (CRM, knowledge bases) delivered in the same job; fast = 50% or more, the order of magnitude the fast scenario needs from a single tool. Between is `emerging`.

Applied to nber.customer_support_agents.productivity_uplift.pt.

3Tracker interpretation

A 14% average, 34% for novices and nothing for experts. A tool that levels the floor, not one that replaces the role.

4Evidence

Latest point
14.0%as of 2023-04-30
customer_support_agents
Value the bands apply to
14.0%as of 2023-04-30

2 observations. Hollow points are disputed (see counterevidence). Every point links to its observation.

5Status and reasoning

consistent with normalsince 2026-09-10 · evaluate

Evaluator: 0.14 (NBER working papers and chapters, 2023-04-30) is inside the consistent band (hi=0.2). Auto-reason; band rationale: Normal = an average uplift under 20%, the range prior process tools (CRM, knowledge bases) delivered in the same job; fast = 50% or more, the order of magnitude the fast scenario needs from a single tool. Between is `emerging`.

6Timeline notes

  • 2023-04-30 customer_support_agents · 14.0%as of 2023-04-30

7Counterevidence

What cuts against this reading

One firm, one tool, 2020-21 vintage models; issues per hour is a partial productivity measure; later assistants may do more.

8Update history

  1. 2026-09-10unmeasured to consistent with normalconf — → 75 · evaluate

    Evaluator: 0.14 (NBER working papers and chapters, 2023-04-30) is inside the consistent band (hi=0.2). Auto-reason; band rationale: Normal = an average uplift under 20%, the range prior process tools (CRM, knowledge bases) delivered in the same job; fast = 50% or more, the order of magnitude the fast scenario needs from a single tool. Between is `emerging`.

    obs:15a42a77

9Confidence

75 / 95 — good evidence, some ambiguity

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; below 50 limited or vague.

10Related