← Every exhibit
What each of these studies actually measured
Two of the five measure what people report. Three measure what changed. The last column is the one the first two cannot have.
| Study | Design | Subjects | Setting | Where a harm would show |
|---|---|---|---|---|
| Rodriguez et al. (2025) | Survey, partial least squares | Sales reps, one health-care firm | B2B selling | Not measured: the outcome is a report |
| Hautamäki and Heikinheimo (2025) | Grounded theory, interviews | 32 top-level managers | B2B sales organisations | Not measured: the outcome is capability |
| Brynjolfsson et al. (2025) | Staggered rollout, not randomised | 5,172 support agents | Post-sale support, one firm | Top of the skill distribution |
| Dell’Acqua et al. (2026) | Preregistered randomised experiment | 758 knowledge workers | 18 constructed tasks inside the frontier, 1 outside | One side of a task boundary |
| Luo et al. (2019) | Randomised field experiment | More than 6,200 customers | Structured outbound calls | The customer's side of the call |
Swipe or scroll horizontally if the table is wider than your screen.
Reference & Evidence
Source: Author's summary of the five papers cited, from the published abstract of each. Designs are described as the authors describe them. No result figures are reproduced here; the argument of the essay is that those numbers should not travel.
Each line is a claim from the register this journal publishes against, resolved from the register at build time.
- B GenAI has a positive impact on sales process effectiveness, administrative efficiency and sales performance, verbatim: "The results indicate that GenAI technology has a positive impact on the effectiveness of the sales process, administrative efficiency, and sales performance" JBIM 40(10), 2013–2027 · 10.1108/JBIM-02-2025-0097 ·
ROD25-C1 - B The instrument chain runs interviews to pilot survey to online survey, and "Data were analyzed using partial least squares (PLS)", with the manager condition stated as a result: "upper management support influences the use of GenAI technology and positively moderates the relationship between technology self-efficacy and the use of GenAI technology" JBIM 40(10), 2013–2027 · 10.1108/JBIM-02-2025-0097 ·
ROD25-C2 - B One firm, one sector, and the authors say so: "The study focuses on a single healthcare company, which may limit its generalizability to other healthcare companies" JBIM 40(10), 2013–2027 · 10.1108/JBIM-02-2025-0097 ·
ROD25-C3 - B The paper's own limitation: "relies on self-reported data from sales professionals" JBIM 40(10), 2013–2027 · 10.1108/JBIM-02-2025-0097 ·
ROD25-C4 - A Grounded theory, and the sample is small and senior: "Using a grounded theory approach, we derive empirical insights from interviews with 32" top-level managers, described again as in"terviews with 32 top-level managers of sales organizations" hautamaki-heikinheimo-2025-ai-b2b-sales ·
HH25-C1 - A "despite the wide accessibility of AI, its adoption in B2B sales remains limited" hautamaki-heikinheimo-2025-ai-b2b-sales ·
HH25-C2 - A The capabilities named: data-based human capital, social capital, "a transformative AI-positive mindset" hautamaki-heikinheimo-2025-ai-b2b-sales ·
HH25-C3 - A The published figure is 15%, not 14%: productivity rises "as measured by issues resolved per hour, by 15% on average" brynjolfsson-li-raymond-2025-generative-ai-at-work ·
BLR25-C1 - A The heterogeneity runs both ways: "the most experienced and highest-skilled workers see small gains in speed and small declines in quality" brynjolfsson-li-raymond-2025-generative-ai-at-work ·
BLR25-C2 - A The design is not a randomised trial: "the staggered introduction of a generative AI-based conversational assistant using data from 5,172 customer-support agents" brynjolfsson-li-raymond-2025-generative-ai-at-work ·
BLR25-C3 - A Not common problems and not novel ones: "the gains from AI adoption are largest for moderately rare problems" brynjolfsson-li-raymond-2025-generative-ai-at-work ·
BLR25-C4 - A Downstream, in the authors words: "customers are more polite and less likely to ask to speak to a manager" brynjolfsson-li-raymond-2025-generative-ai-at-work ·
BLR25-C5 - A The unevenness is invisible to the operator: the frontier is jagged "where AI assistance improves performance for some tasks but worsens it for others, even within the same knowledge workflow and with a seemingly similar level of difficulty" dellacqua-etal-2026-jagged-frontier ·
DELL26-C1 - A Preregistered, three arms, and it "involved 758 knowledge workers": "subjects were randomly assigned to one of three conditions: no AI access, GPT-4 AI access, or GPT-4 AI access with a prompt engineering overview" dellacqua-etal-2026-jagged-frontier ·
DELL26-C2 - A Inside the frontier, verbatim: "completing 12.2% more tasks and completing them 25.1% more quickly on average while also delivering solutions of significantly improved quality" dellacqua-etal-2026-jagged-frontier ·
DELL26-C3 - A Outside it, and on one deliberately selected task: "for a complex managerial task selected to be outside the frontier, subjects using AI were 19% less likely to produce correct solutions" dellacqua-etal-2026-jagged-frontier ·
DELL26-C4 - B The design, in the authors' words: "This study exploits field experiment data on more than 6,200 customers who are randomized to receive highly structured outbound sales calls from chatbots or human workers" luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C1 - B Undisclosed bots matched proficient workers: "undisclosed chatbots are as effective as pro"ficient workers and four times more effective "than inexperienced workers in engendering customer purchases" luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C2 - B Disclosure before the conversation is the finding: "a disclosure of chatbot identity before the machine"-customer conversation reduces purchase rates by more than 79.7% luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C3 - B And it costs conversation as well as sales: "the chatbot disclosure substantially decreases call length" luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C4 - B The mechanism is perception rather than measured performance: "they are curt and purchase less because they perceive the disclosed bot as less knowledgeable and less empathetic" luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C5 - B A subjective perception effect despite objective competence; late disclosure mitigates luo-tong-fang-qu-2019-chatbot-disclosure: held as Articles in Advance, part-only; B until the issue edition ·
LUO19-C6
Grades: A, verified against the printed page of the primary source · B, primary source, text layer only · C, authoritative secondary · D, reported.
Related exhibits
-
Four cuts of data you already have
From the essay What AI actually changes in revenue operations
-
Candidate shared numbers, and how each one fails
From the essay The one number a commercial team should share.
-
The six-month rollout review
From the essay The sales tooling minefield, twenty-four years on