Who is the Oracle for your Guidewire implementation?

Guidewire implementations are significant undertakings.

For a large insurer, they can involve PolicyCenter, BillingCenter and ClaimCenter, multiple insurance products, hundreds of integrations, data migration, rating, payments, documents, customer portals, regulatory requirements and extensive configuration.

And inevitably, a significant amount of the programme is spent on testing and quality assurance.

But I think there is a more important question than how much does testing cost? It is, who actually knows what should be tested?

That question is becoming increasingly important as AI starts to transform software testing.

Automating the Test Isn't the Hard Problem Anymore

The testing industry has spent years solving execution.

• Can we automate this journey?

• Can we execute it faster?

• Can we run thousands of tests in parallel?

• Can AI generate the automation for us?

Increasingly, the answer to all of those questions is yes. But that exposes a much harder problem.

What should we test in the first place?

And then:

• What should the expected outcome be?

• What could go wrong?

• What risks need to be tested?

• Which regulatory obligations matter?

• What assertions actually prove the business outcome?

• What changed in this release?

• What does that change mean for this particular insurer?

Those questions require knowledge. And in a large Guidewire programme, that knowledge is fragmented.

Nobody Has the Complete Picture

Guidewire knows Guidewire. The systems integrator understands the implementation. The insurer understands its products, customers and business processes. Risk and compliance teams understand the regulatory obligations and controls. Testers understand testing. Individual SMEs understand years of decisions, exceptions and tribal knowledge.

But very rarely does one person or one system understand all of it.

That fragmentation is one reason enterprise testing becomes so expensive. Teams spend enormous amounts of time gathering requirements, speaking to SMEs, interpreting documents, designing scenarios, writing tests, maintaining automation, analysing failures and establishing whether the software actually did what the business intended. Then the application changes and they do much of it again.

This is the problem we have built Touchstone to solve.

Touchstone becomes the Oracle

When we talk about Touchstone understanding Guidewire, we mean something much more significant than teaching an AI how to navigate ClaimCenter. We are building an Oracle for software assurance.

Touchstone brings together:

• Guidewire knowledge - how the platform, applications, objects, workflows and integrations are designed to work.
  
  

• Testing expertise - how to turn requirements and risks into effective positive, negative, boundary, integration and end-to-end tests.
  
  

• Insurance knowledge - how policy, billing and claims processes work and the business outcomes that matter.
  
  

• Regulatory and risk knowledge - the obligations, controls and risks that need to be considered.
  
  

• Customer knowledge - how this particular insurer has configured Guidewire, its products, integrations, workflows, business rules and operating model.
  
  

Put those together and you have something fundamentally different from an AI test-generation tool. Touchstone can reason about what needs to be true.

Consider a Claim

A traditional automation approach might start with a journey:

Create claim → assign claim → create exposure → reserve → approve payment → close claim.

You can automate that. You can make it run quickly. You can run it a thousand times. But none of those things necessarily proves that the claims process works correctly.

The real objective might be:

A valid claim should be assessed and settled accurately, by appropriately authorised people, in accordance with the policy, the insurer's controls and applicable regulation.

Now the problem becomes much richer.

• Could an invalid policy generate a claim?

• Could someone approve a payment beyond their authority?

• Is the correct coverage being applied?

• Are reserves calculated and updated correctly?

• Can a control be bypassed?

• What happens when an integration fails halfway through the process?

• Are financial records updated correctly?

• Are required customer communications generated?

• Does the audit trail contain the evidence needed to demonstrate what happened?

• What happens in the negative cases?

Those are not questions about automation. They are questions about assurance.

To see how insurers approach testing across PolicyCenter, BillingCenter and ClaimCenter in practice, read our guide to Guidewire testing.

From Objective to Evidence

This is the model behind Touchstone:

Objective → Risk → Control → Scenario → Journey → Assertions → Evidence

The Oracle understands the context. The agents do the work.

They determine the scenarios required, create the journeys and assertions, execute them, maintain them as the application changes and collect the evidence.

The human moves up the value chain.

Instead of spending their time manually creating and maintaining thousands of tests, they provide context, challenge assumptions, review evidence and make the decisions requiring judgement and accountability.

AI proposes. Humans govern.

And That Changes the Economics

This is where I think the conversation about AI and testing often misses the bigger opportunity.

Reducing test execution time is useful. Generating automation faster is useful.

But on a large Guidewire transformation, execution is only one part of the cost.

The bigger human effort sits across understanding requirements, identifying risk, designing tests, creating automation, maintaining it, investigating results, retesting and producing evidence.

Touchstone attacks that entire chain.

The ambition isn't simply:

How can we make testing cheaper?

It is:

How much more assurance can we achieve for every £ and every hour we spend on quality?

If we can simultaneously reduce human effort while increasing coverage, test effectiveness and evidence, the economics of quality change considerably.

The Real Opportunity for Insurers

Guidewire knows how Guidewire should work. Your implementation partner knows how it was implemented. You know how your insurance business needs to work.

Touchstone brings those bodies of knowledge together with testing, risk and regulatory expertise and continuously turns them into the tests, assertions and evidence required to prove that the system still works as intended.

That is the opportunity I find exciting. Not AI replacing testers. Not generating thousands more tests. Not another layer of automation.

An Oracle that understands what good looks like, and an agentic system capable of continuously proving it.

As software changes faster, I think that distinction is going to matter enormously. Because the question for a large insurer isn't really:

“How much testing have we done?”

It is:

“What evidence do we have that our Guidewire operation works as intended?”