I or my management discussed AI output alteration with the client

Of the 224 respondents who answered, 1 (0.45%) answered True and 223 (99.55%) answered False. No association with the study measures was tested, because fewer than two answers were assigned to at least ten engagements.

The share answering True did not differ by any segment family after Holm correction.

Of the 50 written responses, 43 were negative. At least 22 of the 43 negative responses said the financial institution was not told, or that the correction stayed within the vendor team. The single positive response came from a respondent who had told both the financial institution and their own leadership.

Of the 224 respondents who answered, 1 (0.45%) answered True and 223 (99.55%) answered False. The non-answers were I do not know (25) and I prefer to not answer (16). Each engagement was assigned the answer given by the largest number of its respondents. No engagement was assigned True, and False was the assigned answer on 86 engagements, “I do not know” on 10, “I prefer to not answer” on 3, and 51 engagements were tied or unanswered and carry no assigned answer.

The primary responses numbered 49, written in answer to C216, C214 and C204, the written-answer questions of the indicator's survey section. The remaining 1 responses, 2.00% of the total, were secondary and came from 1 other questions. That secondary share ranked 7th from the highest among the 13 Part 5 indicators with 30 or more written responses, against a median of 2.00%. Respondents who wrote both a primary and a secondary response numbered 0. Negative responses made up 85.71% of primary responses (42 of 49) and 1 of the 1 secondary responses.

In 22 of the 43 negative responses, the respondent said the financial institution was not told, never saw the corrections, or that no one outside the team knew. In 21, the team lead or manager was told or knew. Six responses said the respondent did not know what the financial institution was told, most after telling the team lead.

Of the 50 responses, 43 (86.00%, 95% CI 76.00% to 94.23%) were negative and 36 (72.00%) carried the strongest negative score of −10. Positive responses numbered 1 (2.00%), and 6 were neutral. The mean written score was −7.8 (95% CI −8.9 to −6.7, n = 50 responses), and the median was −10.0.

Each share is computed on the respondents who answered in that row. Rows with fewer than 10 respondents report counts only.

Segment families recorded at engagement level were tested on engagements, and respondent role and vendor type on respondents. The share answering True did not differ by any segment family after Holm correction.

Written sentiment did not differ by engagement status and lifecycle stage across the 39 engagements with a written response (Kruskal-Wallis H 2.45, df 3, p = 0.485). Fewer than two groups held 10 or more responses, so the groups are reported as counts only: active engagements in Build-Out (5), active engagements in Transition/Cut-Over (8), active engagements in Production (6) and canceled engagements (31).

No test across respondent roles was run, because fewer than two roles had 30 or more written responses. Individual contributors wrote 48 responses, engagement managers 2.

Cancellation rate is the share of engagements canceled by the study date, measured at engagement level (150 engagements, 56 canceled). Canceled engagements were rated after cancellation. The association is therefore a descriptive contrast and not a forecast of cancellation. Each test compares the engagements by their assigned answer, within the study set and within each segment.

Of the engagements assigned False, 43 of 86 were canceled. No test was computed, because fewer than two answers were assigned to at least ten engagements.

Customer satisfaction is the percentage recorded against each engagement. The primary estimate uses the 94 active engagements.

Among active engagements, customer satisfaction was 60.1% (n = 43 engagements) where respondents most often answered False. No test was computed, because fewer than two answers were assigned to at least ten engagements.

The engagement indicator mean is the mean of an engagement's ratings across the rated indicators, measured at engagement level. The indicator is not a rated indicator and does not enter the mean.

The engagement indicator mean was 3.5 (n = 86 engagements) where respondents most often answered False. No test was computed, because fewer than two answers were assigned to at least ten engagements.

Intent to leave is the answer to C019, “How likely are you to leave your current job in the next 12 months?”, rated 1 to 10 by up to 798 respondents. For the indicator, the comparison uses the mean intent to leave of each engagement.

The engagement mean of intent to leave was 7.6 (n = 86 engagements) where respondents most often answered False. No test was computed, because fewer than two answers were assigned to at least ten engagements.

Two comparisons set the engagements with at least one negative response against all other engagements. Of the 115 other engagements, 111 had no written response addressing the indicator. These two comparisons therefore do not isolate written sentiment. Of the 35 engagements with a negative response, 22 (62.86%) were canceled, against 34 of the 115 other engagements (29.57%). The difference reached significance (Fisher exact p < 0.001). Customer satisfaction among active engagements did not differ between the two groups (mean 60.38% on 13 engagements against 64.69% on 81, Welch t −1.37, p = 0.188).

Two correlations test written sentiment directly. Across the 39 engagements with a written response, the engagement mean written score was not associated with cancellation (point-biserial r 0.19, p = 0.238). Across 48 respondents, the respondent mean written score was not associated with intent to leave (Spearman rho 0.03, p = 0.847). Of the four comparisons, one reached significance at 0.05.

Fewer than two associations were tested, so no comparison across the measures applies.

Answers clustered within engagements (ICC 0.43). Clustering reduced the 265 respondent answers to the equivalent of about 160 independent answers. Respondent-level intervals therefore used the cluster bootstrap. Respondent-level tests used standard errors that allow for clustering by engagement.

No association was tested, so no equivalence test applies. The equivalence bound is an omega squared of 0.090, the study's threshold for a negligible association expressed as variance explained.

No robustness check could be computed, because no primary test was computed.

Each of the four measures was computed from data the indicator does not contain. Each result is therefore a finding and not an arithmetic consequence.

A categorical indicator has no signed correlation with the study measures and therefore no profile against the study norm. The comparison by answer appears in the four association sections.

Among the 13 Part 5 indicators with 30 or more written responses, the indicator's share of negative responses, 86.00%, ranked 9th from the highest, against a median of 98.03%. Its mean written score, −7.8, ranked 3rd from the most negative, against a median of −6.3.

The indicator most often present in the same responses was C212, “I discussed the AI output alteration with internal management”, in 45 of the 50 responses (90.00%). C211, “I have altered AI outputs” (28 responses), C157, “Rate the factual reliability of the AI tools' outputs on this engagement before any human review.” (27 responses), C159, “Rate the consistency of the tools' outputs when given the same input and instructions more than once.” (8 responses) and C036, “What stage is the engagement in?” (5 responses) followed.

The most frequent theme was “Financial institution was not told”, in 31 responses (62.00%). The second was “Told vendor-side management”, in 28 responses.

Of the respondents who answered, 0.45% said they or their management had discussed the AI output alteration with the financial institution (one of 224). No engagement was assigned the answer True, so no association with the study measures could be tested.

The written evidence matches the answers. Respondents described the correction of AI outputs as routine work, checked line by line within the team, often known to the team lead and seldom raised with the financial institution.

The evidence is the reported experience of the vendor's delivery teams, collected in a single cross-section. The associations reported here are not causal. The answer is the respondent's own report of an alteration and not a record of it.

Each engagement carries the answer given by the largest number of its respondents, and tied engagements are excluded from the engagement-level tests.

Shares of written sentiment are shares of responses and not of the study set of respondents. The written score takes five fixed values, so its mean and median are coarse. Responses clustered within engagements (50 responses on 39 engagements). Intervals for written sentiment therefore used a bootstrap by engagement.

The written responses are counted by sets of words stated in each rule. A response that described a practice in other words is not counted, so each count is a floor.

With one respondents answering True, the indicator supports a description of how rarely the AI output alteration was discussed and no comparison by answer.