Executive Action Plan
By Phil Hatch · Akholi · Full Study Report · October 2026
The findings of the study translate into actions for the executives of financial institutions that buy agentic AI outsourcing. Each action rests on a finding reported earlier in this report and names the Part in which that finding appears. No action introduces a new figure.
The actions address the conditions a financial institution can prepare before it signs and the controls it can require during delivery. The study supports an audit and not a legal conclusion, and the actions are framed as the scope of that audit and of the contracts that follow it.
On this page
2 sectionsSeven Immediate Actions
The seven actions respond to three groups of findings. The lowest-rated failure categories rested with the financial institution, the joint management of the two firms and the vendor. The controls meant to govern agent work existed in name only or were bypassed. The delivery team sustained the engagements by manual alteration, discussed with vendor management and almost never with the financial institution.
The actions run in the order a financial institution meets them. The first two apply before a contract is signed. The third and fourth apply to the terms and the reporting the contract governs, and the last three to the controls that operate while agents are in production.
| Action | Agentic AI outsourcing is new | Performance is poor | No failure category performs well | Manual alteration sustains the engagements | Call to audit |
|---|---|---|---|---|---|
| 1.1 Gate every agentic contract behind a financial institution readiness standard | Part 1 | Part 2 | |||
| 1.2 Establish co-management before the contract is signed | Parts 2 and 3 | ||||
| 1.3 Audit the design of vendor reporting rather than its arithmetic | Part 1 | Part 5 | Part 5 | ||
| 1.4 Restructure commercial terms around outcomes and exit | Part 2 | Part 1 | |||
| 1.5 Treat human oversight as a control function | Parts 2 and 3 | Part 5 | |||
| 1.6 Listen directly to the vendor delivery team | Part 4 | Parts 4 and 5 | |||
| 1.7 Require a manual alteration register and test it | Part 5 | Part 5 |
Each marked cell names the Part in which the supporting finding is reported. The grid holds no figures.
1.1 Gate Every Agentic Contract Behind a Financial Institution Readiness Standard
Financial institution readiness was the lowest-rated of the five failure categories, and about half of its answers were distressed (Part 2). The delivery teams rated the financial institution’s data, legacy systems, processes, staff, management of the engagement, provisioning of access and business case as the weakest conditions of the engagement. When that preparation failed, its tasks moved to the delivery team.
A financial institution can test its own readiness before it signs. The standard should require data with a named owner and a validation step, and systems and environments that give an agent a way in. It should require processes documented to the depth an agent needs, and identities issued for agents in place of borrowed logins. A named decision-maker for the engagement and a business case fixed before work begins complete the standard.
The standard converts a finding about the past into a gate for the future. A contract that proceeds without it places the financial institution’s unfinished preparation on the delivery team, where the study found it absorbed by hand (Part 5).
1.2 Establish Co-Management Before Contract Signature
Co-management was the second-lowest of the five failure categories, and the written responses described joint structures that existed in name only (Parts 2 and 3). The two firms kept two lists of the same risks and issues, met in forums that decided nothing and left the authority to stop an agent unassigned. Where a named person on each side held that authority, respondents described an agent stopped in minutes.
The joint structures should be agreed before the contract is signed. The agreement should name the decision rights for each step of the shared process and the boundary of what an agent may do without a person. It should establish one list of risks and issues held by both firms, and a named person on each side, reachable at every hour the agents operate, with authority to pause an agent.
The cost of these structures is small and the written evidence describes their absence as costly. Respondents proposed the same remedies in their own words, one list, one name against each step and one person with the authority to stop the agent (Part 3).
1.3 Audit the Design of Vendor Reporting Rather Than Its Arithmetic
Reporting alteration was named in most of the written answers on how the vendor’s reports and dashboards were built (Part 5). Respondents described activity counted in place of outcomes, human-corrected work counted as the agent’s, failures removed from the base of a reported rate and open cases that never aged. A report built this way can be arithmetically correct and still misrepresent the engagement.
An audit of vendor reporting should test what the measures count and what they omit. It should ask which cases are excluded from each rate, whether output a person corrected is counted as the agent’s, how long open cases have been open, and which baselines and samples were chosen. The tools’ own reports of success should not stand as evidence that a run succeeded (Part 3).
The measures should also show the human work behind the agent. A report that records the corrections made by hand gives the financial institution a view of the engagement that a report of agent activity alone does not.
1.4 Restructure Commercial Terms Around Outcomes and Exit
The vendor’s commercial model priced agentic work on the human effort it was meant to remove, and respondents described engagements still measured on seats and hours (Part 2). Licenses written for named human users were consumed by agents, and no respondent described a tested move to a second model provider. The commercial terms rewarded headcount on engagements built to reduce it.
The contract should price the outcomes the agents deliver and the human effort they still require, stated separately. The licensing of agent work with each third-party provider should be settled before the agents run. The contract should also require an exit plan, tested before production, for the loss of the primary model or cloud provider.
The written evidence describes human effort behind the agent kept out of the reporting under terms set on human effort. Terms set on outcomes, with the human effort priced openly, give the vendor no reason to keep that effort out of view.
1.5 Treat Human Oversight as a Control Function
Human review in these engagements was sized for the output of people and received the output of agents (Part 2). Review points approved agent output without examining it, change boards met weekly or monthly against agents that changed daily, and drift between releases had no owner (Part 3). The control existed in the process documents and did not operate on the output.
Human oversight should be designed, staffed and tested as a control function. Review capacity should be set from the volume of agent output and not from the headcount of the old process. A review point should record what the reviewer examined, and a sample of approvals should be re-performed to test that scrutiny took place.
Change control should run at the tempo of the agents. A change to an agent, by either firm, should reach the other firm before it takes effect, and behavior that drifts between releases should have a named owner who watches for it. A control that exists in documents and does not operate in production is the failure the study describes.
1.6 Listen Directly to the Vendor Delivery Team
The individual contributors who perform the engagement’s tasks and make nearly every manual alteration are the people the financial institution hears from least (Parts 4 and 5). Their alterations were discussed with vendor management and stopped there. The 13 disclosures to the financial institution that the study recorded were all made by engagement managers.
The financial institution should establish a direct channel to the individual contributors on its engagements. The channel can be a periodic confidential survey of the delivery team or a standing review of the delivery team’s own logs. It can also be an audit that interviews the analysts who handle the data, the process and the output. Contact with the vendor’s management alone did not reach the alterations the study measured.
The delivery team’s account is also the earliest warning available. Respondents described intent to leave, fatigue and knowledge held only in people, in written answers that no existing channel brought to the financial institution (Part 4).
1.7 Require a Manual Alteration Register and Test It
Manual alteration was described on all 150 engagements in the study, and it was seldom recorded and almost never disclosed to the financial institution (Part 5). The study cannot establish that any alteration breached a regulatory or legal obligation, and the alterations took place inside engagements that respondents described as governed by such obligations.
The financial institution should require a register of every manual alteration of its data, its processes, the agents’ outputs and the reporting it receives, kept by the vendor and disclosed to the financial institution. Each entry should record what was changed, by whom, why and on whose authority. Work done by hand for a third party should be recorded in the same register.
The register should be tested against the engagement’s own records. An audit should sample agent outputs, data loads and reported figures and trace each correction made by hand to its entry. The register then gives the financial institution what the study found missing, a record of the human work that sustains the engagement.
Immediate Executive Actions
All three categories rated weakest in this study, the financial institution’s own readiness, co-management with the vendor, and vendor capability, are within the financial institution’s control or reach. Manual alteration sustains these engagements and follows from those weaknesses. Within 12 weeks the board can know how every engagement is run, contain the critical issues, and approve a funded plan for the year. A 52-week program, set out in Table 28, begins with one appointment.
| Weeks | Actions | Delivered to the board |
|---|---|---|
| Week 1 | Appoint an independent audit lead with a written mandate and budget. Name the executive accountable for corrections. Preserve all records. Dual approval on agent changes. CEO message to all staff. | Mandate, appointments, preservation instruction, interim controls |
| Week 2 | Form the audit team, with Internal Audit as observer and an independent third-party advocate for the vendors. Notify every vendor and require preservation. Collect and reconcile the engagement list. | The complete engagement list |
| Weeks 3 to 4 | Rate and rank every engagement. Define the core group and apply interim restrictions. Open the confidential channel to vendor and financial institution staff. Questionnaire due at the end of Week 4. | Inventory, ranking, core group, interim restrictions, response rate |
| Weeks 5 to 8 | Face-to-face lifecycle audit of each core engagement. Review every other engagement against staff returns. Identify the critical issues and prepare the emergency action plan. | Evidenced findings, emergency action plan, decisions requested |
| Weeks 9 to 12 | Fix or contain every critical issue, each closure retested. Prepare the funded first-year plan and the standing oversight function. | Closure record, funded first-year plan |
| Weeks 13 to 52 | Board approves the plan and decides on every core engagement. Complete the plan, review the remaining engagements, train both teams, and validate every closure. | Quarterly reports, a decision on every engagement by Week 52 |
2.1 Week 1: Appoint the Audit Lead
In the first week the board appoints one senior executive to lead an audit of every agentic AI outsourcing engagement. The appointee holds full autonomy and authority over the audit, is independent of every executive who sponsors or runs an engagement, and reports to the board’s risk committee. A written mandate grants access to every record, every person on either side, and every vendor site, and the authority to impose interim controls. Any refused request goes to the committee chair. A second executive, named by the CEO, is accountable for the corrections the audit will require.
Four instructions issue in the same week. Counsel orders preservation of every engagement record on the financial institution’s side, including agent configurations, changes to rules and prompts, the agents’ outputs before and after correction, manual work records, and the calculations behind every report. Management freezes every change to an agent’s rules or configuration without recorded dual approval and names a person on each side with authority to stop each agent. Legal and Compliance determine the supervisory notifications required in each jurisdiction and the handling of privilege. The CEO states to all staff the purpose of the review, the duty to cooperate, and the protection of anyone who reports in good faith.
Every week the audit lead reports in writing to the committee chair, and at the end of the fourth, eighth, and twelfth weeks to the committee. Any critical finding reaches the chair on confirmation. By the end of the first week the committee holds the signed mandate, both appointments, the preservation order, and the interim controls.
2.2 Week 2: Form the Audit Team and Notify the Vendors
In the second week the audit lead forms the team from Compliance, governance, and regulatory affairs, Legal, Human Resources, Procurement, outsourcing management, Risk, Technology, and Finance. Internal Audit attends as observer and validates every closure. An independent third-party firm joins as advocate for the vendors and operator of the confidential channel. Conflicts of interest are cleared before work begins.
Each vendor receives formal notice of a comprehensive performance and compliance audit. The notice orders preservation of all work materials, invokes the financial institution’s audit and access rights, and requests the engagement’s contracts, change logs, agent configurations, manual work records, staffing, and the roster of the delivery team. The notice states that the audit gives the vendor a formal mechanism to raise obstacles on the financial institution’s side with evidence. Each vendor undertakes in writing that no member of its staff will suffer for taking part.
A list of all outsourcing engagements is compiled the same week from the procurement and outsourcing registers, the contracts, the payment records, and the technology and model inventories. Pilots, transition work, subcontractors, and third-party tools are included.
2.3 Weeks 3 to 4: Rank Every Engagement and Open the Confidential Channel
In the third and fourth weeks the team reviews every engagement from the financial institution’s own records and ranks each on financial exposure and the level of legal and regulatory obligation it is subject to. Impact on the financial institution’s account holders, the autonomy and volume of the agents, and the manual work already visible weigh in the ranking. The ranking method is recorded and given to the committee. From the ranking the audit lead defines the core group of high-risk engagements that receive the immediate audit and, where warranted, a corrective action plan. Interim restrictions apply to any core engagement whose exposure cannot wait.
Through the third-party firm the team opens a confidential interview channel to every member of every delivery team, whatever their seniority or role, and to the financial institution’s own staff on the engagements. Participation is confidential and has no consequence. Identities remain with the third party, and findings reach the financial institution in aggregate.
Questions cover what was manually altered, when, why, by whom, and whether the financial institution was told. A second set covers legal and regulatory compliance and the reports the financial institution receives. Improvements to the engagement are sought across the financial institution’s side, the vendor’s side, co-management, the AI tools, and the third-party ecosystem. Each roster arrives in the third week, and answers are due by the end of the fourth.
2.4 Weeks 5 to 8: Audit the Critical Engagements
In Weeks 5 to 8 the team audits each core engagement in a face-to-face walkthrough with the vendor, across the full lifecycle from the financial institution’s preparation to production. Four areas are reviewed in full.
1. Review of the financial institution’s side covers its data and processes, its organization, its secondary and dependent processes, and its ownership and response rates.
2. Review of the vendor’s side covers its delivery team, its playbook, its controls over manual alteration, and its knowledge of the financial institution’s business and the obligations that govern it.
3. Co-management is reviewed for how the two firms decide, change, escalate, and resolve, and for how long each takes.
4. Technology is reviewed for the agents, their configurations and change history, their monitoring, and the third-party ecosystem beneath them.
Controls are tested as they operate. For the audit period the manual alteration register is reconstructed, and each entry is traced to its cause. Reported performance is reconciled to the work performed, including the human work behind it. Skills on both teams are observed in the work. In parallel, every other engagement is reviewed against the questionnaire returns, and any engagement whose staff report undisclosed manual alteration or compliance concerns joins the core group.
From this evidence the audit identifies the critical issues and the problematic engagements and prepares an emergency action plan. Each critical issue is assigned its containment, its owner on each side, its date, the assessment of harm to the financial institution’s account holders, the notification decision, and the contract remedies. By the end of the eighth week the committee holds the evidenced findings and the emergency action plan and takes the decisions reserved to it.
2.5 Weeks 9 to 12: Correct the Critical Issues and Set the First-Year Plan
In Weeks 9 to 12 management corrects the critical issues found in Weeks 5 to 8. Each correction is retested against production evidence before Internal Audit accepts closure. A revised policy, contract, or attestation is not a closure. Where a critical issue cannot be corrected in these weeks, the accountable executive and the CRO restrict or suspend the activity or begin an exit. No affected use expands while a critical issue remains open.
Weeks 9 to 12 also produce the first-year plan for the board’s approval at Week 13. Four methods make up the plan.
1. Remaining engagements are reviewed in ranked order under the same walkthrough, evidence standard, and confidential channel, and every engagement is decided by the end of the year.
2. Co-management is rebuilt on every continuing engagement. Each side names a decision owner, one joint issue list records agreed response times, agent changes pass through joint control, the manual alteration register is shared, and reports are reconciled to the work performed.
3. On the financial institution’s side, every engagement and material workflow receives a named owner with measured response times, dependent processes are redesigned for agents, and data is reconciled before an agent uses it. Engagement owners, process owners, and their overseers are trained in the preparation, active management, and co-management of an agentic AI engagement.
4. Vendor delivery teams are trained in the financial industry function, the law and regulation that govern it, and the financial institution’s processes, and demonstrate that competence before holding unsupervised responsibility. Vendor staffing is stabilized, playbooks are rewritten, and controls over manual alteration and disclosure are written into the contract.
A standing oversight, control, and compliance function for the agents is also created, under a single executive accountable for AI and reporting to the CRO. Under it sit the inventory of agents, the change control, the monitoring, review capacity sized to the volume of output, the authority to stop an agent, and the obligations mapped to each workflow. Every item in the plan has an owner, a date, a budget, and a closure test applied by Internal Audit. A communication plan for the financial institution’s staff and the delivery teams accompanies it. By the end of Week 12 every critical issue is corrected or contained, and the funded plan is before the committee.
2.6 Weeks 13 to 52: Complete the First-Year Plan
At Week 13 the board approves the first-year plan and decides whether each core engagement continues, continues under added controls, is restructured, is transferred, or exits. From Week 14 management completes the plan as approved. Each quarter the committee receives a report with the manual alteration register, the closure record, the response-time measures, the competence results, and the remaining exposure. Internal Audit validates each closure before it is reported. By Week 52 the board holds a decision on every engagement, a complete and disclosed manual alteration register, controls that pass retesting, and reporting that matches the work performed. Lessons are written into the outsourcing policy and the contracting standards, and the review becomes a standing function for the life of every engagement.
This study found an immature delivery model sustained by manual work that, by the delivery teams’ own account, the financial institution was rarely told about. Every weakness behind it is within the financial institution’s reach. Within 12 weeks the program above brings that work into the open, and within the year it corrects what it reveals. The board can make the first appointment this week.