Rate the efficiency of the tools' token use, considering retries, loops, and reprocessing needed to complete work

The mean rating was 4.6 and the median 5.0 (n = 788 respondents). Ratings of 3 or below, the distressed range, made up 32.36% of the answers. The rating was associated with all four study measures. Higher ratings came with less cancellation (r = −0.66, 95% CI −0.74 to −0.56, p < 0.001, n = 150 engagements), higher customer satisfaction among active engagements (r = 0.67, 95% CI 0.54 to 0.77, 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 = −0.63, cluster bootstrap 95% CI −0.69 to −0.57, p < 0.001 with standard errors clustered by engagement, n = 740 respondents). After Holm correction, ratings differed by engagement status and lifecycle stage (Welch F 46.88, Holm p < 0.001, omega squared 0.426), respondent role (Welch F 8.88, Holm p = 0.002, omega squared 0.018) and vendor type (Welch F 4.21, Holm p = 0.006, omega squared 0.021).

Of the 24 written responses, nearly all from C177, the comment field on the AI tools, 18 were negative and 5 positive. Respondents described tools that retried or looped on the same task many times, with no counter to show it, so that cost and volume grew out of sight. Customer satisfaction was higher among the active engagements with a negative written response, the reverse of the direction in the ratings, on a base of 13 engagements. The other written outcome tests did not reach significance.

Ratings of 3 or below, the distressed range, accounted for 255 of the 788 answers (32.36%). The engagement mean fell in the distressed range in 33 of 150 engagements.

The primary responses numbered 22, all written in answer to C177, “Add any comments on the agentic AI tools that the ratings in this section do not capture.” The remaining 2 responses, 8.33% of the total, were secondary and came from 1 other questions. Respondents who wrote both a primary and a secondary response numbered 0. Negative responses made up 72.73% of primary responses (16 of 22) and 2 of the 2 secondary responses.

Fifteen responses described the tool retrying, re-running or looping on the same task, file or job. The counts given were specific: a receipt tried eight or nine times before the tool gave up, a break case going round five or six times, nine retries on one submission, a batch re-run four times with no explanation, and a job that looped eleven times and looked the same as one that went through clean. One agent retried a failed refund call through the night with no one rostered to catch it, and another kept re-running the same match on a custodian file for four hours.

Seven responses described the retries or their cost as invisible until late: found by watching runtime rise, by a cost alert, at month end, or never seen by finance leads at all. One respondent described processors handling more volume than came in because the tool retried itself. Four responses described teams splitting runs into smaller passes, batching at night or running the work twice on purpose, and two described paying for the whole run again when a job failed partway, in one case three or four times a day. Two responses came from C135, on the employer's capability, and placed the loops in how agents were run and staffed.

The five positive responses described token use as efficient or improved, with less repetition of the same step, and placed their concerns in testing, logs and setup. One respondent reported that a looping job was caught by a cost alert and the cost problem was fixed after it was raised.

Of the 24 responses, 18 (75.00%, 95% CI 56.52% to 91.67%) were negative and 2 (8.33%) carried the strongest negative score of −10. Positive responses numbered 5 (20.83%), and 1 was neutral. The mean written score was −3.1 (95% CI −4.8 to −1.2, n = 24 responses), and the median was −5.0.

Of the 24 responses, 24 came from respondents who also rated the indicator. Among them, the written score did not track the rating (Spearman rho 0.11, 95% CI −0.31 to 0.51, p = 0.593, n = 24 responses). Of the 18 negative responses with a rating, 5 (27.78%) came from respondents whose rating was in the distressed range.

Respondents who wrote about the indicator and gave a rating numbered 24, with a mean rating of 4.8. The 764 respondents who rated the indicator without writing about it had a mean of 4.6. The difference, 0.2 (95% CI −0.7 to 1.2), included zero. Ratings in the distressed range made up 29.17% of the ratings from respondents who wrote (7 of 24) and 32.46% of the ratings from respondents who did not (248 of 764).

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 46.88, Holm p < 0.001, omega squared 0.426). Active engagements in Transition/Cut-Over rated highest (mean 5.5, n = 236 respondents) and canceled engagements lowest (mean 3.2, n = 299 respondents). Ratings also differed by respondent role (Welch F 8.88, Holm p = 0.002, omega squared 0.018). Portfolio directors rated highest (mean 5.6, n = 31 respondents) and individual contributors lowest (mean 4.4, n = 605 respondents). Ratings also differed by vendor type (Welch F 4.21, Holm p = 0.006, omega squared 0.021). Consulting firms rated highest (mean 5.2, n = 106 respondents) and BPO specialists lowest (mean 4.1, 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 23 engagements with a written response (Kruskal-Wallis H 4.32, df 3, p = 0.229). Fewer than two groups held 10 or more responses, so the groups are reported as counts only: active engagements in Build-Out (6), active engagements in Transition/Cut-Over (8), active engagements in Production (3) and canceled engagements (7).

No test across respondent roles was run, because fewer than two roles had 30 or more written responses. Individual contributors wrote 15 and engagement managers wrote 9 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 = −0.66, 95% CI −0.74 to −0.56, Holm p < 0.001, n = 150 engagements). The association held after Holm correction in thirteen 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 = 0.67, 95% CI 0.54 to 0.77, Holm p < 0.001, n = 94 engagements). The association held after Holm correction in thirteen 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 = −0.63, 95% CI −0.67 to −0.59, Holm p < 0.001, n = 740 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 133 other engagements, 127 had no written response addressing the indicator. These two comparisons therefore do not isolate written sentiment. Of the 17 engagements with a negative response, 4 (23.53%) were canceled, against 52 of the 133 other engagements (39.10%). The difference did not reach significance (Fisher exact p = 0.290). Customer satisfaction among active engagements differed between the two groups (mean 70.46% on 13 engagements against 63.07% on 81, Welch t 2.26, p = 0.037).

Two correlations test written sentiment directly. Across the 23 engagements with a written response, the engagement mean written score was not associated with cancellation (point-biserial r 0.13, p = 0.556). Across 23 respondents, the respondent mean written score was not associated with intent to leave (Spearman rho 0.16, p = 0.480). Of the four comparisons, one 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 intent to leave the weakest. The associations with customer satisfaction and intent to leave did not differ (Holm p = 0.438).

Answers clustered within engagements (ICC 0.57). Clustering reduced the 788 respondent answers to the equivalent of about 229 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, 75.00%, ranked 49th from the highest, against a median of 80.82%. Its mean written score, −3.1, ranked 60th 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 4 of the 24 responses (16.67%). C160, “Rate the end-to-end reliability of the tools on multi-step tasks, from first step to finished output.” (3 responses), C161, “Rate the predictability of the tools' token consumption for a given volume of work.” (3 responses), C165, “Rate how completely the tools captured material facts located anywhere in long documents, including mid-document content.” (3 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 “Comparison to a prior engagement”, in 3 responses (12.50%). The second was “Normalised into the process”, in 2 responses.

About a third of all ratings of the efficiency of the tools' token use fell in the distressed range (255 of 788 answers, 32.36%). 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 places the inefficiency in retry behavior that the tools do not report. A loop leaves no trace in the log beyond a finished run, so its cost reaches the financial institution's budget before anyone can see where it arose. The reversed customer satisfaction contrast rests on a small base and does not change the direction of the ratings.

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 (24 responses on 23 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.