How many agentic AI outsourcing engagements have you worked on across your career, including this one?

Respondents typed the number of agentic AI outsourcing engagements they had worked on across their careers, including the current one. The mean answer was 1.8 and the median 2.0, from 1 to 5 (n = 799 respondents). For 371 respondents (46.43%), the current engagement was the first.

The engagement mean of the count was not associated with cancellation (r = −0.05, 95% CI −0.21 to 0.11, p = 0.553, n = 150 engagements), customer satisfaction among active engagements (r = −0.02, 95% CI −0.22 to 0.18, p = 0.840, n = 94 engagements) or the engagement indicator mean (r = 0.07, 95% CI −0.09 to 0.23, p = 0.385, n = 150 engagements), and at respondent level it was not associated with intent to leave (r = −0.03, cluster bootstrap 95% CI −0.10 to 0.04, p = 0.370 with standard errors clustered by engagement, n = 751 respondents). Each association was equivalent to zero, and every robustness check held. Engagement managers and portfolio directors had worked on more such engagements than individual contributors.

Eight written responses addressed the indicator, and seven named how many engagements of this kind the respondent had worked on.

The count is a typed whole number and not a rating, so no distressed range applies. The table groups values in bands of 0.4, so the bands that hold no whole number are empty. One engagement was the most common answer (371 respondents), followed by two (238) and three (151). Four or more came from 39 respondents. No respondent answered I do not know or I prefer to not answer. The engagement mean of the count ran from 1.0 to 2.7.

All eight responses were secondary. They came from 3 questions outside the indicator's own survey section (C135 with 4 responses, C114 with 2 responses and C156 with 2 responses). Seven named how many engagements of this kind the respondent had worked on, from one to four. Of these, four named three or more. Those with more engagements behind them used them as a comparison: one had never seen a financial institution this unready to start, one had seen quality assurance for agent output built after the work was sold on four accounts, and two contrasted the engagement with earlier ones where someone could decide on the spot or where ownership of a correction was clear. One respondent who had seen a single engagement had received no sanctions training and learned the lists from a colleague. Two respondents on their second or third engagement described the route from a working prototype to governed production as thin or unwritten, and one had written most of it on the live engagement.

Of the eight responses, one was negative and the others neutral. With fewer than 10 responses, the distribution is reported as counts only.

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, and none differed after Holm correction. Respondent role and vendor type were tested on respondents. The count differed by role (Welch F 32.31, Holm p < 0.001, omega squared 0.117): individual contributors had a mean of 1.7 (n = 617), engagement managers 2.5 (n = 150) and portfolio directors 2.4 (n = 31). Vendor type showed no difference after Holm correction.

Cancellation rate is the share of engagements canceled by the study date, measured at engagement level (150 engagements, 56 canceled). Canceled engagements were described after cancellation. The association is therefore a descriptive contrast and not a forecast of cancellation.

Across the study set, the engagement mean of the count was not associated with cancellation (r = −0.05, 95% CI −0.21 to 0.11, Holm p = 1.000, n = 150 engagements). The association held after Holm correction in none 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, the engagement mean of the count was not associated with customer satisfaction (r = −0.02, 95% CI −0.22 to 0.18, Holm p = 1.000, n = 94 engagements). The association held after Holm correction in none 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 count is not a rated indicator and does not enter the mean.

The engagement mean of the count was not associated with the engagement indicator mean (r = 0.07, 95% CI −0.09 to 0.23, Holm p = 1.000, n = 150 engagements). The association held after Holm correction in one of the 24 segment levels tested, the credit card and payments issuing engagements (r = 0.84, Holm p = 0.015, n = 12 engagements), a base under 30 among many tests. At respondent level, with standard errors clustered by engagement, respondents with more engagements behind them gave slightly higher ratings of their own (r = 0.08, cluster bootstrap 95% CI 0.01 to 0.14, p = 0.019, n = 799 respondents).

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.

The count was not associated with intent to leave (r = −0.03, cluster bootstrap 95% CI −0.10 to 0.04, p = 0.370 with standard errors clustered by engagement, n = 751 respondents), nor was its engagement-level form (r = −0.03, 95% CI −0.19 to 0.13, n = 150 engagements). The association held after Holm correction in none of the 37 segment levels tested.

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, none of the six pairs differed. Every association was weak.

The count did not cluster within engagements (ICC −0.05): respondents on the same engagement typed different values. The design effect was 1.00, and the effective n equaled the 799 answers. Respondent-level tests still used standard errors clustered by engagement.

All four associations were equivalent to zero. No result was inconclusive. The detection floor shown is the study floor at 150 engagements.

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.

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 departed from the study's typical pattern on all four measures. The typical indicator is a rating with a strong association with each measure, with study medians from 0.60 to 0.84. The count is not a rating, and its associations were weak.

The indicator most often present in the same responses was C126, on the maturity of the employer's playbook for taking agentic use cases from concept to governed production, in three responses. The most frequent theme was “Comparison to a prior engagement”, in four responses, followed by “Raised it, nothing came back” in three.

For most respondents the current engagement was their first or second of its kind, and engagement managers and portfolio directors had worked on more than individual contributors. The count was not associated with cancellation, customer satisfaction, the engagement indicator mean or intent to leave, and each result was equivalent to zero. At respondent level, respondents with more engagements behind them rated the engagement slightly higher, a small association that may reflect role.

The written responses show the use of that experience: respondents with several engagements behind them judged the current one against earlier ones. The written evidence rests on eight responses, a base of anecdote strength.

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. Canceled engagements were described after cancellation.

The count is the respondents' typed report and not a record of staffing.

The count differs by role, and the analysis does not adjust for role, so the respondent-level association with the respondents' own ratings may reflect role rather than the count.

Counts of written responses describe the respondents who wrote and not the study set of respondents. With eight responses, the written evidence supports counts only and no statistic.