Guardian Life Insurance

Wrote the Rules for What the System Could Record, Reuse, and Learn

Founding designer on Guardian's new advisor platform: defined the IA that supports two regulated domains. Designed data entry and reconciliation, eight transactional workflows, and the advisor-facing side of the two AI systems.

Founder Designer

Enterprise Replatforming

10 Months

Wealth Management + Insurance

Data Reconciliation

Agentic Workflows

Guardian Life Insurance

Wrote the Rules for What the System Could Record, Reuse, and Learn

Founding designer on Guardian's new advisor platform: defined the IA that supports two regulated domains. Designed data entry and reconciliation, eight transactional workflows, and the advisor-facing side of the two AI systems.

Founder Designer

Enterprise Replatforming

10 Months

Wealth Management + Insurance

Data Reconciliation

Agentic Workflows

Guardian Life Insurance

Wrote the Rules for What the System Could Record, Reuse, and Learn

Founding designer on Guardian's new advisor platform: defined the IA that supports two regulated domains. Designed data entry and reconciliation, eight transactional workflows, and the advisor-facing side of the two AI systems.

Founder Designer

Enterprise Replatforming

10 Months

Wealth Management + Insurance

Data Reconciliation

Agentic Workflows

Guardian Life Insurance

Wrote the Rules for What the System Could Record, Reuse, and Learn

Founding designer on Guardian's new advisor platform: defined the IA that supports two regulated domains. Designed data entry and reconciliation, eight transactional workflows, and the advisor-facing side of the two AI systems.

Founder Designer

Enterprise Replatforming

10 Months

Wealth Management + Insurance

Data Reconciliation

Agentic Workflows

The company

Platform replatforming

with a new AI layer.

The company

Platform replatforming

with a new AI layer.

The company

Platform replatforming

with a new AI layer.

The company

Platform replatforming

with a new AI layer.

Guardian sells wealth management and insurance products. They want to build a new unified platform that combines 2 regulatory domains, creates a centralized database, and implements new AI features.

The legacy workflows were fragmented across 8 disconnected tools, making them complicated and time-consuming, with a high error rate. Advisors were getting frustrated.

Impact (pilot)

- 47%

Application errors (NIGO)

- 47%

Application errors (NIGO)

- 63%

Approval time

- 63%

Approval time

→

MVP targeted for July 2026; Release 1.1 in test with about 1,000 advisors.

→

Workflows rebuilt on the record: 6 end to end, 2 partial.

→

Multi-product sales reduced from 15 steps and 4 tools to 8 steps in 1.

Highlights

Turned data reconciliation tasks into rules that built trust.

Highlights

Turned data reconciliation tasks into rules that built trust.

Highlights

Turned data reconciliation tasks into rules that built trust.

Highlights

Turned data reconciliation tasks into rules that built trust.

the real problem

Unreliable data was behind

every problem on the platform.

the real problem

Unreliable data was behind

every problem on the platform.

the real problem

Unreliable data was behind

every problem on the platform.

the real problem

Unreliable data was behind

every problem on the platform.

In the discovery workshops, no business owner could narrate their own process end-to-end. The legacy workflows had never been designed; they were just hacked together. This created a messy database, a high error rate, and limited the implementation of the new AI layer.

No accurate client view. A client could exist in different tools with different phone numbers, addresses, and even birthdays. Updating in one tool doesn’t mean updating in all tools.

No insights worth trusting. Guardian had tried AI suggestions in earlier tools. All failed because of the messy database.

No automated workflows. Each transactional workflows involved multiple tools with theirs own order and inputs.

No error checks until the end. Verifications ran at the end of the application process. Some errors were caught after submission, which increased the Home Office’s workload and approval time.

Six stages, six different pain points, one problem.

Project goal

Maintain accurate and trustworthy records that the new platform can depend on.

Project goal

Maintain accurate and trustworthy records that the new platform can depend on.

Project goal

Maintain accurate and trustworthy records that the new platform can depend on.

Project goal

Maintain accurate and trustworthy records that the new platform can depend on.

the solutions

Confirm once upstream, reuse everywhere downstream

the solutions

Confirm once upstream, reuse everywhere downstream

the solutions

Confirm once upstream, reuse everywhere downstream

the solutions

Confirm once upstream, reuse everywhere downstream

01.

Hold each conflict as a decision

and ask at the moment it matters.

Data conflicts or disagreements are unavoidable, and it’s not a one-time task.

01.

Hold each conflict as a decision

and ask at the moment it matters.

Data conflicts or disagreements are unavoidable, and it’s not a one-time task.

01.

Hold each conflict as a decision

and ask at the moment it matters.

Data conflicts or disagreements are unavoidable, and it’s not a one-time task.

01.

Hold each conflict as a decision

and ask at the moment it matters.

Data conflicts or disagreements are unavoidable, and it’s not a one-time task.

Caught and sorted at every sync.

The system sorts each conflicts into an to-do-item or just an alert.

Caught and sorted at every sync.

The system sorts each conflicts into an to-do-item or just an alert.

Caught and sorted at every sync.

The system sorts each conflicts into an to-do-item or just an alert.

Caught and sorted at every sync.

The system sorts each conflicts into an to-do-item or just an alert.

Prioritize by consequence.

Not all conflicts required immediate attention. High-priority ones reach the to-do list, and the rest will become an alert that stay on the client record page.

Prioritize by consequence.

Not all conflicts required immediate attention. High-priority ones reach the to-do list, and the rest will become an alert that stay on the client record page.

Prioritize by consequence.

Not all conflicts required immediate attention. High-priority ones reach the to-do list, and the rest will become an alert that stay on the client record page.

Prioritize by consequence.

Not all conflicts required immediate attention. High-priority ones reach the to-do list, and the rest will become an alert that stay on the client record page.

Skippability depends on cost.

All conflicts are skippable at sync. However, when It becomes regulatory at the moment, the workflow will not move forward until that is resolved.

Skippability depends on cost.

All conflicts are skippable at sync. However, when It becomes regulatory at the moment, the workflow will not move forward until that is resolved.

Skippability depends on cost.

All conflicts are skippable at sync. However, when It becomes regulatory at the moment, the workflow will not move forward until that is resolved.

Skippability depends on cost.

All conflicts are skippable at sync. However, when It becomes regulatory at the moment, the workflow will not move forward until that is resolved.

02.

Used the same rule

to protect the source of truth.

Data is collected through multiple paths at different stage of a sale. Some are noise, some are required. The system uses the same rules to keep the data accurate. All incoming data is hold as a decision, Advisors need to confirm or verified before adding it into the source of truth.

02.

Used the same rule

to protect the source of truth.

Data is collected through multiple paths at different stage of a sale. Some are noise, some are required. The system uses the same rules to keep the data accurate. All incoming data is hold as a decision, Advisors need to confirm or verified before adding it into the source of truth.

02.

Used the same rule

to protect the source of truth.

Data is collected through multiple paths at different stage of a sale. Some are noise, some are required. The system uses the same rules to keep the data accurate. All incoming data is hold as a decision, Advisors need to confirm or verified before adding it into the source of truth.

02.

Used the same rule

to protect the source of truth.

Data is collected through multiple paths at different stage of a sale. Some are noise, some are required. The system uses the same rules to keep the data accurate. All incoming data is hold as a decision, Advisors need to confirm or verified before adding it into the source of truth.

03.

Validated once,

ahead of time.

Validations were a major problem in the legacy workflows. Results weren’t kept or shared, so every verification landed at the end of the application.

03.

Validated once,

ahead of time.

Validations were a major problem in the legacy workflows. Results weren’t kept or shared, so every verification landed at the end of the application.

03.

Validated once,

ahead of time.

Validations were a major problem in the legacy workflows. Results weren’t kept or shared, so every verification landed at the end of the application.

03.

Validated once,

ahead of time.

Validations were a major problem in the legacy workflows. Results weren’t kept or shared, so every verification landed at the end of the application.

Starts ahead of time.

Advisor can start the verification process at any stage of the the sale, even before starting an application.

Starts ahead of time.

Advisor can start the verification process at any stage of the the sale, even before starting an application.

Starts ahead of time.

Advisor can start the verification process at any stage of the the sale, even before starting an application.

Starts ahead of time.

Advisor can start the verification process at any stage of the the sale, even before starting an application.

Runs in the background.

Starting an application triggers it automatically.

Runs in the background.

Starting an application triggers it automatically.

Runs in the background.

Starting an application triggers it automatically.

Runs in the background.

Starting an application triggers it automatically.

Saved and reused.

Results are saved until they expired.

Saved and reused.

Results are saved until they expired.

Saved and reused.

Results are saved until they expired.

Saved and reused.

Results are saved until they expired.

Every check runs at product selection, before the advisor has invested anything. What used to come back as a rejection after submission now shows up at step two.

04.

Optimized everything

downstream.

A trustworthy database enable the new AI layer. For example, the system prepares the application at the beginning, cuts down the application time by pre-filling, highlights what is missing in a single list, and runs verification in the background at the same time.

04.

Optimized everything

downstream.

A trustworthy database enable the new AI layer. For example, the system prepares the application at the beginning, cuts down the application time by pre-filling, highlights what is missing in a single list, and runs verification in the background at the same time.

04.

Optimized everything

downstream.

A trustworthy database enable the new AI layer. For example, the system prepares the application at the beginning, cuts down the application time by pre-filling, highlights what is missing in a single list, and runs verification in the background at the same time.

04.

Optimized everything

downstream.

A trustworthy database enable the new AI layer. For example, the system prepares the application at the beginning, cuts down the application time by pre-filling, highlights what is missing in a single list, and runs verification in the background at the same time.

closing

The same rule runs through

the human and AI layers.

closing

The same rule runs through

the human and AI layers.

closing

The same rule runs through

the human and AI layers.

closing

The same rule runs through

the human and AI layers.

That’s what makes the trust loop work. Advisors make final decisions by confirming data before adding it to the record and approving applications before submission. In return, the system warns and suggests using the confirmed inputs. Every suggestion can be traced back to the advisors themselves. It makes taking recommendations, or even delegating more to the system, easier.

And that’s the trust loop.

Then the system learns the same way. Businesses set thresholds to get more insights into certain behavior, and the system captures the patterns and reports back with valuable insights that might turn into institutional changes in the future. That’s the win-win situation we are building towards.

Impact (pilot)

- 47%

Application errors (NIGO)

- 47%

Application errors (NIGO)

- 63%

Approval time

- 63%

Approval time

→

MVP targeted for July 2026; Release 1.1 in test with about 1,000 advisors.

→

Workflows rebuilt on the record: 6 end to end, 2 partial.

→

Multi-product sales reduced from 15 steps and 4 tools to 8 steps in 1.

Let's connect!

Interested in working together? Reach out and let's chat.
jessica.ling.design@gmail.com

Let's connect!

Interested in working together? Reach out and let's chat.
jessica.ling.design@gmail.com

Let's connect!

Interested in working together? Reach out and let's chat.
jessica.ling.design@gmail.com

Let's connect!

Interested in working together?
Reach out and let's chat.
jessica.ling.design@gmail.com