Rate the quality of the tools' logs, traces, and usage telemetry for detecting and diagnosing failures

The mean rating was 4.6 and the median 5.0 (n = 796 respondents). Ratings of 3 or below, the distressed range, made up 32.04% of the answers. The rating was associated with all four study measures. Higher ratings came with less cancellation (r = −0.65, 95% CI −0.73 to −0.55, p < 0.001, n = 150 engagements), higher customer satisfaction among active engagements (r = 0.66, 95% CI 0.52 to 0.76, p < 0.001, n = 94 engagements), a higher engagement indicator mean computed without the indicator (r = 0.85, 95% CI 0.79 to 0.89, 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 = 749 respondents). After Holm correction, ratings differed by engagement status and lifecycle stage (Welch F 44.03, Holm p < 0.001, omega squared 0.410), respondent role (Welch F 5.25, Holm p = 0.037, omega squared 0.011) and vendor type (Welch F 4.99, Holm p = 0.001, omega squared 0.024).

Of the 101 written responses, 82 were negative and 18 positive, and 44 carried the strongest negative score of −10. Respondents described logs that showed a start and a finish, or a result, and nothing of the steps between, so that failures had to be reconstructed by hand. Engagements with a negative written response were canceled more often, a more negative engagement mean written score came with cancellation, and respondents whose written responses were more negative reported higher intent to leave.

Ratings of 3 or below, the distressed range, accounted for 255 of the 796 answers (32.04%). The engagement mean fell in the distressed range in 36 of 150 engagements.

The primary responses numbered 80, 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 21 responses, 20.79% of the total, were secondary and came from 6 other questions. That secondary share ranked 34th from the highest among the 71 Part 2 indicators with 30 or more written responses, against a median of 18.92%. Respondents who wrote both a primary and a secondary response numbered 4. Negative responses made up 77.50% of primary responses (62 of 80) and 95.24% of secondary responses (20 of 21).

Twenty responses described logs or traces that recorded only the request, the final state or the result: a start time and an end time and nothing between, an action without the reason or the content the agent had read, a step marked finished on a load that was half missing, only the last of several runs kept, and an audit trail that held the agent's final output but not its intermediate steps. Twenty-one responses described no logs at all, or nothing useful in them, when something went wrong. One respondent said the platform should have been stopped on that ground alone, and another, on a pre-production engagement, said the gap was tolerable there and would not be in production.

Ten responses described work redone or a run reconstructed by hand because the trace could not show what happened: two days rebuilding a release from the team's own notes, the sequence of a failed access request rebuilt before the respondent could answer the financial institution, a whole batch or posting redone, and a run rebuilt by hand to show the error was not the respondent's own, a task that filled most of the respondent's week. Three respondents contrasted this with an earlier program where a failed run left a trace they could read. The responses to C055 added two practices that hid the gap: audit findings that could not be traced to an agent action logged as resolved with a generic note, and a curated weekly summary for the steering committee that smoothed over observability gaps the engineering team was raising internally.

The eighteen positive responses described logs good enough to find a failed step in minutes, several against an earlier engagement or the previous year, and one said a trace now ended disputes with the financial institution's operations lead. Several of these respondents rated logs well while placing their concern in drift, cost or false completion, and one said the trace existed but the team had no access to it.

Of the 101 responses, 82 (81.19%, 95% CI 73.33% to 88.57%) were negative and 44 (43.56%) carried the strongest negative score of −10. Positive responses numbered 18 (17.82%), and 1 was neutral. The mean written score was −5.2 (95% CI −6.3 to −4.1, n = 101 responses), and the median was −5.0.

Of the 101 responses, 98 came from respondents who also rated the indicator. Among them, the written score rose with the rating (Spearman rho 0.53, 95% CI 0.37 to 0.67, p < 0.001, n = 98 responses). The written responses corroborated the rating. Of the 79 negative responses with a rating, 44 (55.70%) came from respondents whose rating was in the distressed range.

Respondents who wrote about the indicator and gave a rating numbered 94, with a mean rating of 4.0. The 702 respondents who rated the indicator without writing about it had a mean of 4.6. The difference, −0.6 (95% CI −1.1 to −0.1), excluded zero. Ratings in the distressed range made up 43.62% of the ratings from respondents who wrote (41 of 94) and 30.48% of the ratings from respondents who did not (214 of 702).

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 44.03, Holm p < 0.001, omega squared 0.410). Active engagements in Build-Out rated highest (mean 5.4, n = 165 respondents) and canceled engagements lowest (mean 3.2, n = 304 respondents). Ratings also differed by respondent role (Welch F 5.25, Holm p = 0.037, omega squared 0.011). Portfolio directors rated highest (mean 5.5, n = 30 respondents) and individual contributors lowest (mean 4.5, n = 615 respondents). Ratings also differed by vendor type (Welch F 4.99, Holm p = 0.001, omega squared 0.024). Consulting firms rated highest (mean 5.3, n = 109 respondents) and pure-play agentic AI firms lowest (mean 4.2, n = 69 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 73 engagements with a written response (Kruskal-Wallis H 7.55, df 3, p = 0.056). Of the groups with 10 or more responses, responses from canceled engagements were the most negative (mean −7.1, 52 responses) and responses from active engagements in Production the least negative (mean −1.0, 10 responses).

No test across respondent roles was run, because fewer than two roles had 30 or more written responses. Individual contributors wrote 72, engagement managers wrote 21 and portfolio directors wrote 8 responses.

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 = −0.65, 95% CI −0.73 to −0.55, 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 = 0.66, 95% CI 0.52 to 0.76, Holm p < 0.001, n = 94 engagements). The association held after Holm correction in nine 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 = 0.85, 95% CI 0.79 to 0.89, 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 = 749 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 87 other engagements, 77 had no written response addressing the indicator. These two comparisons therefore do not isolate written sentiment. Of the 63 engagements with a negative response, 34 (53.97%) were canceled, against 22 of the 87 other engagements (25.29%). The difference reached significance (Fisher exact p < 0.001). Customer satisfaction among active engagements did not differ between the two groups (mean 66.10% on 29 engagements against 63.20% on 65, Welch t 1.13, p = 0.262).

Two correlations test written sentiment directly. Across the 73 engagements with a written response, a more negative engagement mean written score came with cancellation (point-biserial r −0.34, p = 0.003). Across 92 respondents, a more negative respondent mean written score came with higher intent to leave (Spearman rho −0.41, p < 0.001). Of the four comparisons, three 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 = 0.528).

Answers clustered within engagements (ICC 0.59). Clustering reduced the 796 respondent answers to the equivalent of about 224 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 0.85 to 0.85.

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, 81.19%, ranked 34th from the highest, against a median of 80.82%. Its mean written score, −5.2, ranked 29th 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 33 of the 101 responses (32.67%). C173, “Rate the tools' ability to explain and evidence how each output was produced, at the depth an audit requires.” (33 responses), C172, “Rate how accurately the tools' own reports of success or failure matched the actual outcomes of their work.” (12 responses), C168, “Rate how consistently the agents took only intended and authorized actions when operating connected tools and systems.” (11 responses) and C164, “Rate how reliably the tools retained instructions and constraints given earlier in long tasks and sessions.” (10 responses) followed.

The most frequent theme was “The scale or question does not fit”, in 9 responses (8.91%). The second was “Human rework absorbs the failure”, in 8 responses.

About a third of all ratings of the quality of the tools' logs, traces and telemetry fell in the distressed range (255 of 796 answers, 32.04%). 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 = 304 respondents).

The written evidence places the weakness between the request and the result. Telemetry that records only endpoints cannot locate a failure, so every failure becomes a manual reconstruction and every dispute a matter of recollection. Written sentiment was associated with cancellation in both engagement-level tests and with intent to leave.

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 (101 responses on 73 engagements). Intervals for written sentiment therefore used a bootstrap by engagement.