A use case tested on the current process and on Duna side by side, compared on STP rate, first-time right, and time per case
Hande Kahraman

Hande Kahraman

On this page

Financial institutions are experimenting with agentic AI faster than they’re putting it into production, and many projects stall before they deliver value. According to EY, 52% of banks surveyed had piloted agentic AI, but just 16% had fully deployed use cases.

Gartner predicts that more than 40% of agentic AI projects across industries will be canceled by the end of 2027 due to rising costs, unclear business value, and lack of risk controls.

Banks can reduce that risk by testing before they scale. Duna’s AI readiness assessment is the starting point, helping teams identify where AI can improve compliance operations. A proof of concept (PoC) then tests the most viable option against existing processes before the bank commits to a larger investment.

Leaders planning a PoC for an agentic client lifecycle management platform need to decide what they want to prove and how they’ll know if the technology delivers.

Choose a use case worth testing

A PoC works best when it tests a specific use case at any stage of the customer lifecycle. For example, a bank testing business onboarding could narrow the scope by business type, geography, or the technology being tested.

Strong use cases usually meet five criteria:

  1. End-to-end: Covers the full journey from onboarding to the analyst’s decision

  2. Baseline: Has existing data on conversion, speed, cost, and customer experience

  3. Sufficient volume: Includes enough cases during the test to measure results

  4. Contained scope: Focuses on one customer segment and one product

  5. Existing process: Covers processes the team already runs, so results can be measured against existing operations

Identify what success looks like

Before testing begins, compliance and operations leaders agree on the current performance baseline with their vendor. They also set business outcomes and metrics.

For a business onboarding PoC, metrics could include straight-through processing (STP) rate, first-time-right rate, or average active onboarding time per customer. Teams choose two or three metrics that match their priorities, along with the results that would justify further investment.

A PoC should show how improvements in individual tasks affect the overall process. For example, AI might help analysts review cases faster. But if the goal is to free up time for high-risk cases, the pilot also needs to establish how many cases analysts can handle and how much of their time goes toward routine reviews.

Scope the pilot

Leaders can choose between two PoC approaches, depending on what they want to achieve.

A targeted pilot addresses a specific problem by replacing one component, such as the customer onboarding interface, while using Duna’s policy engine to orchestrate the relevant rules. The bank keeps its other systems in place and tests how much the change improves that part of the process.

An end-to-end pilot tests a complete solution for a defined use case, such as onboarding customers in a new country.

For a business onboarding PoC, Duna’s end-to-end journey has six steps:

  • Start: Customer begins onboarding online

  • Pre-fill: Trusted sources populate the online form with company and shareholder details

  • Ask: Only missing information is requested

  • Prepare: Case is prepared with a risk review for analyst evaluation

  • Decide: Analyst reviews and decides

  • Hand off: Result is sent to the system of record

For each step, the bank and vendor agree on the configuration for the PoC. This includes which data sources the policy engine will use, what rules trigger additional questions or checks, and when a case needs analyst review. They also define what information passes back to the system of record and in what format.

Test against the current process

An effective test runs the new platform alongside the current process. In the onboarding example, the institution could run an A/B test, using the new platform to onboard some applicants and the usual process for others.

Real customer data lets teams compare how the two approaches perform. Dummy data can show how the technology works, but makes it harder to compare results.

Leaders also need to choose which existing systems to connect during the PoC. Testing individual capabilities without those connections might be useful, but it won’t show how the platform works with the bank’s other systems. The closer the PoC comes to the intended deployment, the more useful its findings will be. See how Duna integrates with banks’ existing systems.

Testing the full process shows how well the steps work together, including the handoffs between automated processes and humans. It can also reveal where improvements in one area create additional work elsewhere.

The test period needs to run long enough to produce useful results and support the decision to move forward.

Keep the project team small

Pilots are most effective when the vendor and the institution collaborate closely, supported by a lean team that can keep the project moving forward. The bank’s working group should include key decision-makers, with risk and compliance teams involved from the start so their concerns are addressed early.

A promising pilot can attract more stakeholders who want to get involved. But bringing more people into the working group can make it harder to move forward. Keeping the core team small helps the pilot stay on schedule.

Make the case with a focused PoC

Compliance and operations leaders don’t need a major implementation to see if a client lifecycle management platform delivers. By running a limited pilot in their own operations, leaders can see how it performs before committing to a full deployment.


Learn how Duna can help you plan and run an effective PoC for client lifecycle management. Book a PoC discussion.

Financial institutions are experimenting with agentic AI faster than they’re putting it into production, and many projects stall before they deliver value. According to EY, 52% of banks surveyed had piloted agentic AI, but just 16% had fully deployed use cases.

Gartner predicts that more than 40% of agentic AI projects across industries will be canceled by the end of 2027 due to rising costs, unclear business value, and lack of risk controls.

Banks can reduce that risk by testing before they scale. Duna’s AI readiness assessment is the starting point, helping teams identify where AI can improve compliance operations. A proof of concept (PoC) then tests the most viable option against existing processes before the bank commits to a larger investment.

Leaders planning a PoC for an agentic client lifecycle management platform need to decide what they want to prove and how they’ll know if the technology delivers.

Choose a use case worth testing

A PoC works best when it tests a specific use case at any stage of the customer lifecycle. For example, a bank testing business onboarding could narrow the scope by business type, geography, or the technology being tested.

Strong use cases usually meet five criteria:

  1. End-to-end: Covers the full journey from onboarding to the analyst’s decision

  2. Baseline: Has existing data on conversion, speed, cost, and customer experience

  3. Sufficient volume: Includes enough cases during the test to measure results

  4. Contained scope: Focuses on one customer segment and one product

  5. Existing process: Covers processes the team already runs, so results can be measured against existing operations

Identify what success looks like

Before testing begins, compliance and operations leaders agree on the current performance baseline with their vendor. They also set business outcomes and metrics.

For a business onboarding PoC, metrics could include straight-through processing (STP) rate, first-time-right rate, or average active onboarding time per customer. Teams choose two or three metrics that match their priorities, along with the results that would justify further investment.

A PoC should show how improvements in individual tasks affect the overall process. For example, AI might help analysts review cases faster. But if the goal is to free up time for high-risk cases, the pilot also needs to establish how many cases analysts can handle and how much of their time goes toward routine reviews.

Scope the pilot

Leaders can choose between two PoC approaches, depending on what they want to achieve.

A targeted pilot addresses a specific problem by replacing one component, such as the customer onboarding interface, while using Duna’s policy engine to orchestrate the relevant rules. The bank keeps its other systems in place and tests how much the change improves that part of the process.

An end-to-end pilot tests a complete solution for a defined use case, such as onboarding customers in a new country.

For a business onboarding PoC, Duna’s end-to-end journey has six steps:

  • Start: Customer begins onboarding online

  • Pre-fill: Trusted sources populate the online form with company and shareholder details

  • Ask: Only missing information is requested

  • Prepare: Case is prepared with a risk review for analyst evaluation

  • Decide: Analyst reviews and decides

  • Hand off: Result is sent to the system of record

For each step, the bank and vendor agree on the configuration for the PoC. This includes which data sources the policy engine will use, what rules trigger additional questions or checks, and when a case needs analyst review. They also define what information passes back to the system of record and in what format.

Test against the current process

An effective test runs the new platform alongside the current process. In the onboarding example, the institution could run an A/B test, using the new platform to onboard some applicants and the usual process for others.

Real customer data lets teams compare how the two approaches perform. Dummy data can show how the technology works, but makes it harder to compare results.

Leaders also need to choose which existing systems to connect during the PoC. Testing individual capabilities without those connections might be useful, but it won’t show how the platform works with the bank’s other systems. The closer the PoC comes to the intended deployment, the more useful its findings will be. See how Duna integrates with banks’ existing systems.

Testing the full process shows how well the steps work together, including the handoffs between automated processes and humans. It can also reveal where improvements in one area create additional work elsewhere.

The test period needs to run long enough to produce useful results and support the decision to move forward.

Keep the project team small

Pilots are most effective when the vendor and the institution collaborate closely, supported by a lean team that can keep the project moving forward. The bank’s working group should include key decision-makers, with risk and compliance teams involved from the start so their concerns are addressed early.

A promising pilot can attract more stakeholders who want to get involved. But bringing more people into the working group can make it harder to move forward. Keeping the core team small helps the pilot stay on schedule.

Make the case with a focused PoC

Compliance and operations leaders don’t need a major implementation to see if a client lifecycle management platform delivers. By running a limited pilot in their own operations, leaders can see how it performs before committing to a full deployment.


Learn how Duna can help you plan and run an effective PoC for client lifecycle management. Book a PoC discussion.