The vendor manipulates the dashboard and performance reports given to the customer

Every rating is read on the goodness view, the reversed scale on which a higher figure is the better outcome. The mean rating was 4.7 and the median 5.0 (n = 786 respondents). Ratings of 3 or below on the goodness view, the distressed range, made up 29.90% of the answers. The rating was associated with all four study measures. Higher ratings on the goodness view came with less cancellation (r = −0.61, 95% CI −0.70 to −0.50, p < 0.001, n = 150 engagements), higher customer satisfaction among active engagements (r = 0.62, 95% CI 0.48 to 0.73, p < 0.001, n = 94 engagements), a higher engagement indicator mean (r = 0.79, 95% CI 0.72 to 0.84, p < 0.001, n = 150 engagements) and lower intent to leave (r = −0.62, cluster bootstrap 95% CI −0.68 to −0.54, p < 0.001 with standard errors clustered by engagement, n = 739 respondents). After Holm correction, ratings differed by engagement status and lifecycle stage (Welch F 31.80, Holm p < 0.001, omega squared 0.358), respondent role (Welch F 7.22, Holm p = 0.009, omega squared 0.012) and vendor type (Welch F 3.92, Holm p = 0.010, omega squared 0.018).

Of the 465 written responses, 464 were negative. The 454 responses to C055, on how the vendor manipulates performance reports and dashboards, described measures that counted work as done when the agent acted, omitted the cases a person had to correct, and blended or reclassified failures until the headline figure looked sound. The single positive response said the indicators reported to the financial institution were not restated afterward, while the vendor's internal margin dashboard overstated utilization.

Ratings of 3 or below on the goodness view, the distressed range, accounted for 235 of the 786 answers (29.90%). The engagement mean fell in the distressed range in 33 of 150 engagements.

The primary responses numbered 454, all written in answer to C055, “How does the vendor manipulate performance reports and dashboards?” The remaining 11 responses, 2.37% of the total, were secondary and came from 4 other questions. That secondary share ranked 5th 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 9. Negative responses made up 99.78% of primary responses (453 of 454) and 100.00% of secondary responses (11 of 11).

The 454 responses to C055 described the ways a reported figure departed from the work behind it. In 156 responses, the measure set aside human or manual work: corrections, overrides, rework or reopened cases that the agent's figures did not carry. Cases, queues or failures excluded from the measure, or routed to a separate queue, bucket or tracker that never reached the report, appeared in 131. In 73 responses, a task was counted as resolved, complete or reviewed the moment the agent acted, whether or not the work was right. Results blended, taken as a mean, rolled up or folded into a composite figure appeared in 70, and failures reclassified or logged under another category in 49.

A baseline, target, definition, denominator or threshold set or changed so the figure held appeared in 44 responses. In 34 responses, the figure measured the agent against itself, its own prior run or its own confidence, and in 21 against a curated, scripted or test sample in place of live work. A dashboard described as green while the work behind it failed appeared in 23. Access, provisioning and agent identities were the subject of 49 responses. The themes most often marked were proxy substitution (112 marks), category conflation (52), denominator exclusion (50) and scope restriction (38).

Seven responses to C156, on joint management, described the reporting in the first person: our dashboard counted a question as closed when the bot replied, the agent's confidence threshold was tuned before each review of the financial institution and reset afterward, and data from the last cycle that met its goals was loaded into the dashboard. A response to C210 said the coverage figure in the report came from an official test suite that no one used, so everyone reporting it knew the figure was meaningless.

Of the 465 responses, 464 (99.78%, 95% CI 99.32% to 100.00%) were negative and 127 (27.31%) carried the strongest negative score of −10. Positive responses numbered 1 (0.22%), and no response was neutral. The mean written score was −6.3 (95% CI −6.6 to −6.1, n = 465 responses), and the median was −5.0.

Of the 465 responses, 437 came from respondents who also rated the indicator. Among them, the written score rose with the rating (Spearman rho 0.10, 95% CI 0.00 to 0.19, p = 0.046, n = 437 responses). The written responses corroborated the rating. Of the 436 negative responses with a rating, 209 (47.94%) came from respondents whose rating was in the distressed range.

Respondents who wrote about the indicator and gave a rating numbered 428, with a mean rating of 3.4. The 358 respondents who rated the indicator without writing about it had a mean of 6.3. The difference, −2.8 (95% CI −3.2 to −2.5), excluded zero. Ratings in the distressed range made up 46.96% of the ratings from respondents who wrote (201 of 428) and 9.50% of the ratings from respondents who did not (34 of 358).

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 31.80, Holm p < 0.001, omega squared 0.358). Active engagements in Production rated highest (mean 5.8, n = 96 respondents) and canceled engagements lowest (mean 3.4, n = 302 respondents). Ratings also differed by respondent role (Welch F 7.22, Holm p = 0.009, omega squared 0.012). Portfolio directors rated highest (mean 5.4, n = 22 respondents) and individual contributors lowest (mean 4.6, n = 613 respondents). Ratings also differed by vendor type (Welch F 3.92, Holm p = 0.010, omega squared 0.018). Pure-play agentic AI firms rated highest (mean 5.1, n = 65 respondents) and BPO specialists lowest (mean 4.3, n = 182 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 125 engagements with a written response (Kruskal-Wallis H 6.02, df 3, p = 0.111). Of the groups with 10 or more responses, responses from canceled engagements were the most negative (mean −6.4, 246 responses) and responses from active engagements in Build-Out the least negative (mean −6.2, 77 responses).

The role test compared individual contributors and engagement managers, the roles with 30 or more responses. Written sentiment did not differ (Kruskal-Wallis H 3.65, p = 0.056, n = 450 respondents): the mean written score was −6.5 for individual contributors (371 responses) and −5.9 for engagement managers (88 responses). Portfolio directors 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 on the goodness view came with less cancellation (r = −0.61, 95% CI −0.70 to −0.50, 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 on the goodness view came with higher customer satisfaction (r = 0.62, 95% CI 0.48 to 0.73, Holm p < 0.001, n = 94 engagements). The association held after Holm correction in fourteen 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 does not enter the mean.

Higher ratings on the goodness view came with a higher engagement indicator mean (r = 0.79, 95% CI 0.72 to 0.84, Holm p < 0.001, n = 150 engagements). The association held after Holm correction in twenty-three 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 = −0.62, 95% CI −0.66 to −0.57, Holm p < 0.001, n = 739 respondents). The association held after Holm correction in thirty-five of the 37 segment levels tested.

Two comparisons set the engagements with at least one negative response against all other engagements. Of the 25 other engagements, 25 had no written response addressing the indicator. These two comparisons therefore do not isolate written sentiment. Of the 125 engagements with a negative response, 55 (44.00%) were canceled, against 1 of the 25 other engagements (4.00%). The difference reached significance (Fisher exact p < 0.001). Customer satisfaction among active engagements differed between the two groups (mean 62.31% on 70 engagements against 69.29% on 24, Welch t −2.44, p = 0.020).

Two correlations test written sentiment directly. Across the 125 engagements with a written response, the engagement mean written score was not associated with cancellation (point-biserial r −0.12, p = 0.179). Across 426 respondents, the respondent mean written score was not associated with intent to leave (Spearman rho 0.00, p = 0.951). Of the four comparisons, two 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, four 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 cancellation and intent to leave did not differ (Holm p = 0.108). The associations with customer satisfaction and intent to leave did not differ (Holm p = 0.210).

Answers clustered within engagements (ICC 0.62). Clustering reduced the 786 respondent answers to the equivalent of about 217 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.

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 13 Part 5 indicators with 30 or more written responses, the indicator's share of negative responses, 99.78%, ranked 2nd from the highest, against a median of 98.03%. Its mean written score, −6.3, ranked 6th from the most negative, against a median of −6.3.

The indicator most often present in the same responses was C052, “The dashboard and performance reports given to the customer are accurate.”, in 460 of the 465 responses (98.92%). C157, “Rate the factual reliability of the AI tools' outputs on this engagement before any human review.” (49 responses), the engagement field recording return on engagement as a percentage (17 responses), C148, “Rate how well joint human review capacity keeps up with the volume of output the agents produce.” (8 responses) and C036, “What stage is the engagement in?” (7 responses) followed.

The most frequent theme was “Proxy substitution”, in 112 responses (24.09%). The second was “Category conflation”, in 52 responses.

About three in ten of all ratings of whether the vendor manipulates the dashboards and performance reports fell in the distressed range on the goodness view (235 of 786 answers, 29.90%). The rating was associated with all four study measures, and every association held under each of the 22 robustness checks. Canceled engagements rated well below the three active stages on the goodness view (mean 3.4, n = 302 respondents).

The written evidence describes reporting shaped to show the agent's activity and not its accuracy. Respondents placed the gap in what the measures counted and omitted, in the work people did behind the agent, and in baselines and samples chosen so the figure held. Engagements with a negative response were canceled more often and recorded lower customer satisfaction. The associations do not establish that the shaping of reports produced the outcomes.

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 1 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 (465 responses on 125 engagements). Intervals for written sentiment therefore used a bootstrap by engagement.

The written responses to C055 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, and a response may fall in more than one set.