Rate how completely the tools kept each client's and each session's information isolated from every other client and session

The mean rating was 4.6 and the median 5.0 (n = 791 respondents). Ratings of 3 or below, the distressed range, made up 30.85% of the answers. The rating was associated with all four study measures. Higher ratings came with less cancellation (r = −0.63, 95% CI −0.71 to −0.52, p < 0.001, n = 150 engagements), higher customer satisfaction among active engagements (r = 0.70, 95% CI 0.57 to 0.79, p < 0.001, n = 94 engagements), a higher engagement indicator mean computed without the indicator (r = 0.87, 95% CI 0.83 to 0.91, p < 0.001, n = 150 engagements) and lower intent to leave (r = −0.66, cluster bootstrap 95% CI −0.71 to −0.60, p < 0.001 with standard errors clustered by engagement, n = 745 respondents). After Holm correction, ratings differed by engagement status and lifecycle stage (Welch F 38.67, Holm p < 0.001, omega squared 0.378), respondent role (Welch F 9.35, Holm p = 0.001, omega squared 0.015) and vendor type (Welch F 4.78, Holm p = 0.002, omega squared 0.023).

All 26 written responses came from C177, the comment field on the AI tools. Eighteen were negative, five positive and three neutral. Respondents described information from another customer, mandate or session appearing in their work, usually caught by the respondent before it went out. No written outcome test reached significance.

Ratings of 3 or below, the distressed range, accounted for 244 of the 791 answers (30.85%). The engagement mean fell in the distressed range in 33 of 150 engagements.

The primary responses numbered 26, all written in answer to C177, “Add any comments on the agentic AI tools that the ratings in this section do not capture.” No response from any other question addressed the indicator, so every response was primary.

Seventeen responses described information from another customer, mandate, team, user or session appearing in the respondent's work. The examples were concrete: a customer name from another mandate in a draft letter, comparables from another customer's folder in a valuation model, a fee quoted from another customer's account in a reply, another customer's holdings in a draft reply, a custodian account number carried into the next session, a cost centre belonging to another organization in a forecast pack, and a leave balance from another employee's ticket in an answer. One respondent noticed a misplaced name only because the address was in the wrong country.

Seven responses described catching the information before it went out, and two respondents added that nobody else would have noticed. Six said the respondent reported an occurrence, and two of those said nothing came back. Five responses described a working practice adopted to keep customers or cases apart: a new session for every customer, a fresh workspace closed at the end of each day, one browser tab per employee, one break at a time, or no use of the tool for balances. A further respondent said they could not confirm that one entity's session was sealed off from another because the logs would not show it.

The five positive responses said that customer information stayed separate, in one case after the team checked for it, and placed their concerns in logging, false completion and invented policy references.

Of the 26 responses, 18 (69.23%, 95% CI 50.00% to 85.71%) were negative and 1 (3.85%) carried the strongest negative score of −10. Positive responses numbered 5 (19.23%), and 3 were neutral. The mean written score was −1.9 (95% CI −4.0 to 0.6, n = 26 responses), and the median was −5.0.

Of the 26 responses, 26 came from respondents who also rated the indicator. Among them, the written score rose with the rating (Spearman rho 0.49, 95% CI 0.13 to 0.77, p = 0.011, n = 26 responses). The written responses corroborated the rating. Of the 18 negative responses with a rating, 14 (77.78%) came from respondents whose rating was in the distressed range.

Respondents who wrote about the indicator and gave a rating numbered 26, with a mean rating of 2.8. The 765 respondents who rated the indicator without writing about it had a mean of 4.6. The difference, −1.9 (95% CI −2.6 to −1.1), excluded zero. Ratings in the distressed range made up 65.38% of the ratings from respondents who wrote (17 of 26) and 29.67% of the ratings from respondents who did not (227 of 765).

The respondents who wrote came disproportionately from the distressed end of the scale. The written evidence therefore gives more weight to the respondents who rated the indicator worst.

Every figure in the segment tables is at respondent level. Segments with fewer than 30 respondents carry no interval. Segments with fewer than 10 report counts only.

Segment families recorded at engagement level were tested on engagement means. Respondent role and vendor type were tested on respondents. Where an expected count fell below 5, the test of the distressed share used Fisher's exact test for a two-by-two table, marked exact, or a Monte Carlo test, marked simulated. After Holm correction, ratings differed by engagement status and lifecycle stage (Welch F 38.67, Holm p < 0.001, omega squared 0.378). Active engagements in Transition/Cut-Over rated highest (mean 5.5, n = 235 respondents) and canceled engagements lowest (mean 3.2, n = 299 respondents). Ratings also differed by respondent role (Welch F 9.35, Holm p = 0.001, omega squared 0.015). Portfolio directors rated highest (mean 5.4, n = 31 respondents) and individual contributors lowest (mean 4.4, n = 608 respondents). Ratings also differed by vendor type (Welch F 4.78, Holm p = 0.002, omega squared 0.023). Consulting firms rated highest (mean 5.1, n = 110 respondents) and BPO specialists lowest (mean 4.2, n = 186 respondents). Outsourcing model, institution type, process family and class of work showed no difference.

Written sentiment did not differ by engagement status and lifecycle stage across the 24 engagements with a written response (Kruskal-Wallis H 2.87, df 3, p = 0.412). Fewer than two groups held 10 or more responses, so the groups are reported as counts only: active engagements in Build-Out (2), active engagements in Transition/Cut-Over (7), active engagements in Production (2) and canceled engagements (15).

No test across respondent roles was run, because fewer than two roles had 30 or more written responses. Individual contributors wrote 20 and engagement managers wrote 6 responses, reported as counts only.

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.

Across the study set, higher ratings came with less cancellation (r = −0.63, 95% CI −0.71 to −0.52, Holm p < 0.001, n = 150 engagements). The association held after Holm correction in fourteen of the 20 segment levels tested.

Customer satisfaction is the percentage recorded against each engagement. The primary estimate uses the 94 active engagements. A descriptive contrast uses all 150 engagements.

Across the active engagements, higher ratings came with higher customer satisfaction (r = 0.70, 95% CI 0.57 to 0.79, Holm p < 0.001, n = 94 engagements). The association held after Holm correction in fifteen of the 16 segment levels tested.

The engagement indicator mean is the mean of an engagement's ratings across the rated indicators, measured at engagement level (150 engagements). The indicator is one of the rated indicators. Each comparison therefore uses the mean recomputed without it.

Higher ratings came with a higher engagement indicator mean (r = 0.87, 95% CI 0.83 to 0.91, Holm p < 0.001, n = 150 engagements). The association held after Holm correction in twenty-four of the 24 segment levels tested.

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. It is measured at respondent level on up to 798 respondents, with the engagement-level form reported alongside.

Respondents who gave higher ratings reported lower intent to leave (r = −0.66, 95% CI −0.70 to −0.62, Holm p < 0.001, n = 745 respondents). The association held after Holm correction in thirty-six of the 37 segment levels tested.

Two comparisons set the engagements with at least one negative response against all other engagements. Of the 134 other engagements, 126 had no written response addressing the indicator. These two comparisons therefore do not isolate written sentiment. Of the 16 engagements with a negative response, 8 (50.00%) were canceled, against 48 of the 134 other engagements (35.82%). The difference did not reach significance (Fisher exact p = 0.285). The comparison of customer satisfaction was not computed, because one of the two groups held too few active engagements.

Two correlations test written sentiment directly. Across the 24 engagements with a written response, the engagement mean written score was not associated with cancellation (point-biserial r 0.19, p = 0.371). Across 25 respondents, the respondent mean written score was not associated with intent to leave (Spearman rho −0.11, p = 0.607). Of the three comparisons, none reached significance at 0.05.

The four associations were compared in pairs on their engagement-level forms, each oriented so that a higher value is the better outcome, across the 150 engagements with all four measures. After Holm correction, five of the six pairs differed. The association with the engagement indicator mean was the strongest and the association with cancellation the weakest. The associations with customer satisfaction and intent to leave did not differ (Holm p = 0.913).

Answers clustered within engagements (ICC 0.61). Clustering reduced the 791 respondent answers to the equivalent of about 218 independent answers. Respondent-level intervals therefore used the cluster bootstrap. Respondent-level tests of intent to leave used standard errors that allow for clustering by engagement.

No association was equivalent to zero. Each of the four differed from zero. The detection floor shown is the study floor at 150 engagements. Intent to leave was tested on its effective sample size.

Each change was computed before rounding and may differ by 0.01 from the difference between the two printed values.

All 22 checks held. A check holds when its test falls on the same side of 0.05 as the primary test. Where the primary test falls below 0.05, the check must also keep the sign of the primary correlation. The thin-base check removes the engagements in any level of outsourcing model or process family with fewer than 30 engagements. Institution type is not used for that check, because every one of its levels has fewer than 30 engagements.

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. The indicator is one of the rated indicators, and the engagement indicator mean was recomputed without it. The correction changed r by 0.00, from 0.87 to 0.87.

Each deviation was computed before rounding and may differ by 0.01 from the difference between the two printed values.

The comparison covered 150 engagements. The indicator followed the study's typical pattern on all four measures, and no deviation held after Holm correction.

Among the 71 Part 2 indicators with 30 or more written responses, the indicator's share of negative responses, 69.23%, ranked 61st from the highest, against a median of 80.82%. Its mean written score, −1.9, ranked 68th from the most negative, against a median of −4.7.

The indicator most often present in the same responses was C157, “Rate the factual reliability of the AI tools' outputs on this engagement before any human review.”, in 8 of the 26 responses (30.77%). C176, “Rate the quality of the tools' logs, traces, and usage telemetry for detecting and diagnosing failures.” (5 responses), C165, “Rate how completely the tools captured material facts located anywhere in long documents, including mid-document content.” (4 responses), C169, “Rate how safely the tools handled sensitive data, with no exposure through outputs, logs, or connected systems.” (4 responses) and C172, “Rate how accurately the tools' own reports of success or failure matched the actual outcomes of their work.” (3 responses) followed.

The most frequent theme was “Raised it, nothing came back”, in 3 responses (11.54%). The second was “Occurs every cycle or most days”, in 2 responses.

Close to a third of all ratings of how completely the tools kept information isolated between customers and sessions fell in the distressed range (244 of 791 answers, 30.85%). The rating was associated with all four study measures, and every association held under each of the 22 robustness checks. Canceled engagements rated far below the three active stages (mean 3.2, n = 299 respondents).

The written evidence describes a confidentiality control that rested on individual attention. Leakage was found by the person reading the output, and the remedies were working habits rather than changes to the tools. For a financial institution handling customer holdings, account numbers and salary details, each such occurrence is a potential data incident, and two of the respondents who reported one heard nothing back.

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.

Tests by respondent role and by vendor type treated respondents as independent. The section on clustering reports how far clustering within engagements reduced the information in respondent-level figures.

Customer management executives gave 2 answers, a base of anecdote strength.

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 (26 responses on 24 engagements). Intervals for written sentiment therefore used a bootstrap by engagement.

The written responses number fewer than 30, so their shares and the written-sentiment tests rest on a small base.