Respondents on all 150 engagements reported manual alteration as part of delivery. Their accounts of limited disclosure and incomplete recording raised concerns about how fully reported performance represented the work being done. Manual alteration, the study’s term for the delivery team’s changes to data, processes, outputs, or reporting and its performance of third-party work by hand, often involved several methods on the same engagement. Respondents described corrections that sustained delivery but were not always visible in engagement records or reports to the financial institution. Their unfavorable assessments of regulatory training added a concern about preparation for work subject to legal and regulatory requirements. Together, these findings describe a gap between the work respondents reported performing and the visibility and preparation surrounding that work.
The five manual alteration methods are manual alteration of financial-institution-provided data, manual alteration of financial institution processes, manual alteration of AI outputs, reporting alteration, and third-party work done by hand. They are distinct practices, may overlap, and were measured differently. The occurrence counts combine direct True answers with written descriptions of a method on the respondent’s engagement. Related ratings describe frequency, capability, knowledge, or controls rather than independently establishing occurrence.
The study collected the accounts of 850 members of vendor delivery teams on 150 engagements between January and June 2026. Its cross-sectional design recorded each engagement once. The results describe this study set and do not establish industry prevalence. The accounts do not independently establish what the financial institution knew or whether a legal or contractual obligation was breached.
Scroll horizontally to view the complete figure.
Respondents on all 150 engagements reported manual alteration as part of delivery. | ||||||||||||||||||||||||||||
Exact chart data
|
Note: An engagement is counted for a method when at least one respondent answered True to its yes-or-no question or described that method in a written answer about the engagement. Manual alteration of data, processes, and AI outputs uses both sources. Reporting alteration and third-party work done by hand use written answers. An engagement may be counted under several methods. The base is 150 engagements.
Respondents reported manual alteration of financial institution processes on 138 of the 150 engagements (92.00%) and manual alteration of financial-institution-provided data on 133 (88.67%). They reported reporting alteration on 126 (84.00%), manual alteration of AI outputs on 122 (81.33%), and third-party work done by hand on 74 (49.33%).
Several methods were usually reported on the same engagement. Respondents described all five on 51 engagements (34.00%) and four on 65 (43.33%). They reported a single method on eight engagements.
Across the structured and written answers, 738 of the 850 respondents reported at least one method (86.82%). The method-based compilation included 2,410 written answers from 725 respondents on all 150 engagements. In 475 answers from respondents on 131 engagements, manual alteration was described as occurring every cycle or on most days. These are overlapping answer counts, not counts of separate incidents.
“i do this almost every day now. i inform my team lead please, not the client, therefore it stays internal. it is exhausting but i see no other way.” -Individual Contributor, CXM, Corporate and back-office functions
These findings describe the delivery team’s reported experience. They do not establish what the financial institution knew or whether a reported practice breached a contractual or regulatory obligation. Manual correction may have been approved or necessary to maintain delivery. Neither its occurrence nor its absence from a report independently establishes misconduct.
This is the one area I would hold up as an example — the monitoring we added after last year's provider upgrade scare has held, and the model vendor has not surprised us once. What has not improved is our own memory of the manual work. The people who used to fix awkward LP questionnaires by hand have moved on, and I am not certain we could go back to that if we had to.”
PORTFOLIO DIRECTOR · CXM, Corporate and back-office functions