The delivery workforce is junior, dissatisfied, and leaving.

Table 16 · Experience and credentials by role

Meaningful agentic tenure does not exist in this workforce, at any rank. Among the 850 respondents, 467 reported no agentic experience, and 326 reported a single year of experience; the maximum anywhere in the study is 1 year.

Junior in the trade, new to the technology

The workforce is junior in the trade — 77.04% reported fewer than 5 career years, and the median career length is 2. The extreme sits with the Individual Contributors: 51.70% were on their first agentic engagement, 59.74% reported having no prior agentic AI experience, and 9.93% reported having no career experience at all before this specific engagement.

Management holds deep delivery seniority and no agentic seniority. Engagement Managers averaged 10 years of career experience and 0.5 years of agentic work; Portfolio Directors averaged 15.06 years of career experience and even less experience with AI. Signing off on agentic delivery falls to people who learned the outsourcing trade in an era that ended before the technology arrived.

Agentic AI is still new, and there is no solution to the shortage of workers with collective wisdom gained through years of experience. The risk to financial institutions is nonetheless material: those who build, operate, manage, and govern agentic AI projects for the banks are learning on the job.

Certifications and credentials

No credential separates confidence from capability. Across the workforce, 58.71% hold no agentic AI certification, and the only AI certification that does appear is one of 5 generic platform-fundamental courses. Management holds full credentials from the previous outsourcing era, including Six Sigma, ITIL, and project management.

Table 17 · Morale and departure intent by role

Job satisfaction averaged 4.19, and the most common rating was 1; 46.79% rated satisfaction at 3 or below. Departure intent mirrors it: 53.88% rated themselves as likely to leave within a year, and satisfaction and departure intent correlated at r = -0.91 (p < 0.001). 43.29% rated satisfaction at 4 or below and departure likelihood at 7 or higher, and only 16.47% of respondents indicated they were both content and staying.

Feeling wasted drives the departure

While direct management, job stress, and other indicators surfaced, the strongest correlate of departure intent is whether staff believe their employer uses them effectively — a correlation of r = -0.78 (p < 0.001). Respondents noted frustration that their employer is not using them to help define the agentic AI offering strategy and to optimize the vendor playbooks now being created.

Management read the distress as milder

Management read the team’s distress as milder than the people reporting it. Portfolio Directors assessed their teams’ satisfaction 1.69 points above what the delivery layer reported for itself, and understated the team’s intent to depart by 1.26 points.

Respondents’ strongest strategy expectation is that the client will move the work into its own global capability center (GCC): 44.28% rated the move as likely, compared with 19.78% who expected the client to scale its outsourcing within the respondent’s employer.

Table 18 · Strategy expectations and career direction by role

Expectation intensifies inside failing work. Respondents on canceled or high-risk engagements rated the GCC move at 7.06 of 10, with 64.91% rating it 7 or higher; on canceled engagements alone, the mean reaches 7.31.

41.55% of respondents rated themselves as likely to seek employment within a bank’s GCC. The pull concentrates on the population leaving: likely leavers rated GCC preference at 7.45 out of 10, while likely stayers rated it at 2.77.

Section 2.1 (Client Preparation) documents the human backstop: manual repair and undocumented decisions hold these engagements together outside any reported measure. The cost of that dependence lies in what is left when the people leave. An engagement’s operating knowledge has no backup copy, and the error-catching that keeps defects from the client departs with the staff who performed it.

Table 19 · The backstop and its exposure

All 3 continuity conditions ranged from 4.58 to 4.66, with 30.82% to 33.54% of ratings distressed — the engagements do not reliably know where their own human fixes live, and they have not deliberately handed those fixes to anything.

Against those readings sits a 53.88% departure intent. The gradient makes the exposure concrete: every workforce indicator is 2 to 3 points lower for canceled engagements than for active ones. The engagements most dependent on undocumented human knowledge are those whose staff sit closest to departure.

What leaves with the team

Undocumented workarounds and the map of past fixes leave with each departure. For an institution, the conclusion is structural and requires direct mitigation: failure to address team departures or document their methods will lead to a new team forced to reinvent what the previous team refined. Given the volume of undocumented and behind-the-scenes efforts by the respondents, as regularly noted throughout this report, the loss of team members will affect the performance of the outsourcing engagement.