Agile testing in the insurance industry: higher quality, less effort, faster releases

The insurance industry is under intense pressure to change. Digital customer services, new products and regulatory requirements must be implemented at an ever-faster pace. At the same time, the workload involved in quality assurance is increasing significantly.

This is because every premium adjustment, every change to the claims process and every enhancement to an existing system must be tested reliably. Added to this are regulatory requirements, such as those imposed by DORA and BaFin. Test results must therefore not only be technically sound, but also documented in a way that is traceable and audit-proof.

Many insurers face an additional problem: there is a shortage of experienced subject-matter testers, business analysts and test managers. At the same time, the business departments are already working at full capacity. They are therefore often unable to take on additional testing tasks with the necessary depth, speed and continuity.

Against this backdrop, a new generation of quality assurance solutions is emerging: agent-based testing with ‘adesso test agents’.

Why traditional testing approaches are reaching their limits

Insurance companies often operate highly complex IT landscapes. Legacy systems, portal solutions, partner platforms, cloud applications and business processes must work together seamlessly. Product changes, migrations, release changes, integration projects and regulatory requirements necessitate extensive regression testing. As a result, testing requirements are constantly growing, whilst release cycles are becoming shorter and, at the same time, the traceability of quality assurance must increase.

Skills shortages and overburdened business departments

A key problem is the shortage of experienced business testers, business analysts and test managers. In the insurance industry in particular, in-depth domain knowledge is required to correctly assess tariffs, contract logic, claims processes, benefit claims, regulatory requirements and special business cases. This knowledge is often concentrated in the hands of just a few key individuals.

At the same time, the specialist departments are often already working at full capacity. They are responsible for product changes, operational processes, customer enquiries, regulatory implementations and project work all at the same time. As a result, additional testing tasks often cannot be undertaken with the necessary depth, speed and continuity. The result is bottlenecks in quality assurance, delayed approvals, incomplete test coverage and an increasing reliance on individual experts.

Test automation alone does not solve the problem

Traditional script-based test automation is often seen as the answer to these bottlenecks. In practice, however, it only solves part of the problem. Automated tests can speed up recurring workflows, but they require stable processes, clear requirements, suitable test data and maintainable test scripts. In dynamic insurance projects, however, business rules, user interfaces, interfaces and data structures change regularly.

This results in additional maintenance work: test scripts must be adapted whenever there is a business or technical change, test data must be laboriously prepared, and special business cases must be added manually. Test automation therefore does not automatically reduce the need for business expertise. On the contrary: without business evaluation, prioritisation and maintenance, it can even create new complexity and merely shift existing bottlenecks to the maintenance of the automation itself.

Outsourcing provides only limited relief

Outsourcing can supplement a lack of testing capacity in the short term, particularly for clearly defined and well-documented tasks such as standardised regression tests or repeatable test steps.

However, this is often insufficient when dealing with complex insurance-specific issues. External teams require a deep understanding of tariffs, contract logic, claims processes, regulatory requirements and specialist exceptional cases. This knowledge must still be provided, contextualised and updated by in-house experts.

This creates additional coordination work: specialist departments must answer queries, evaluate test results and make technical decisions. If they are already overburdened, outsourcing tends to shift the bottleneck rather than actually resolving it.

The key point remains: outsourcing can free up capacity, but it does not replace the specialist knowledge and oversight required in complex insurance projects.

The actual conflict of objectives

This gives rise to a fundamental conflict of objectives: on the one hand, releases are to be delivered faster, more frequently and more reliably. On the other hand, there is often a lack of staff capacity and subject-matter expertise to ensure the necessary test coverage using traditional methods on a long-term basis. Neither manual testing nor traditional test automation, nor pure outsourcing, fully addresses this problem. What is therefore needed is an approach that makes better use of specialist knowledge, reduces routine tasks and specifically eases the burden on business departments and test teams.

Agent-based testing: the next step in quality assurance

Many insurers are already using AI assistance systems modelled on ChatGPT, which are operated internally and protect confidential data. The next step is now to utilise these capabilities specifically in quality assurance.

‘adesso test agents’ addresses precisely this need, creating an AI-supported working environment for quality assurance in which specialised agents take on repetitive, time-consuming and structuring tasks. However, business departments, test managers and business analysts are not replaced; instead, they retain control at all times.

At the heart of the approach is an agent-based framework. This means: There is not just a single AI function that generates text at the touch of a button. Instead, several specialised AI agents work together, dividing the tasks between them. An example:

• One agent can analyse requirements and generate user stories from them.
• Another derives test cases from these user stories.
• A further agent assists with the creation of bug tickets.
• Yet another helps with the preparation or execution of tests.

A central orchestrator coordinates this collaboration. It ensures that the agents do not work in isolation, but interact within a traceable process. In this way, a verifiable test scope can be developed step by step from a user story, a business concept or a regulatory requirement: first, relevant information is recognised; then business risks are identified; next, test cases are proposed; and finally, results are prepared for review, documentation or further processing.

An example illustrates the benefits: if a new rule is introduced in a claims process, under which certain claims may be automatically approved, several questions must be examined:

• What prerequisites must be met?
• What exceptions are there?
• What data is required?
• Which roles are authorised to approve claims?
• Which cases must still be checked manually?

A Requirements Agent can derive such questions from the requirement, highlight any ambiguities and, once the review is complete, generate a user story. A Test Case Agent can use this to create positive and negative test cases, for example, a claim within the approval limit, a claim just above the limit, a case with missing documents or a case with conflicting information.

Subsequently, another automation agent can execute this test case autonomously, log the results as required and, if necessary, instruct a defect agent to create a defect ticket.

Throughout this process, the human always retains control and has the option to intervene at any time.

Regulatory requirements and ‘human-in-the-loop’

Full automation of the testing process by AI agents is currently neither technically feasible nor desired from a regulatory perspective. Human oversight is crucial, particularly in regulated insurance processes and in AI-supported decisions with potential implications for customers.

The EU AI Act explicitly requires a “human-in-the-loop” approach for high-risk systems – a category which often includes insurance systems:

• Humans must review the AI’s results,
• be able to intervene in AI-supported workflows,
• and be able to correct or halt decisions.

Agent-based testing takes this framework into account: AI agents take on the time-consuming, repetitive tasks, whilst humans retain responsibility and can intervene to steer the process at any time.

This does not result in autonomous AI that carries out quality assurance independently – effectively in the background – but rather a human-controlled, AI-supported SDLC workbench. It supports the entire software development and testing process, standardises recurring tasks, makes domain knowledge more readily usable and increases the scalability of quality assurance. Particularly where business departments are overburdened and testing capacity is in short supply, ‘adesso test agents’ can help to convert existing knowledge more efficiently into testable artefacts.

How agent-based testing specifically improves insurance projects

Agent-based testing supports insurers particularly where business complexity, limited testing capacity and high documentation requirements converge. The key benefits are:

• Faster and more structured derivation of test cases
  AI agents can analyse requirements, business concepts or regulatory guidelines and derive initial proposals for test cases, test steps and test data from them. This means that business domain testers do not have to start from scratch, but can review, refine and prioritise the prepared results.
• More consistent test artefacts
  Test cases, acceptance criteria and defect tickets can be created according to standardised templates. This reduces variations in language, level of detail and structure – a key benefit for reviews, audits and collaboration between multiple teams.
• Better utilisation of existing knowledge
  Information from business concepts, wikis, SharePoint repositories, test management systems or defect histories can be incorporated into test design in a more targeted manner. This ensures that known edge cases, dependencies and common sources of error are better taken into account.
• More targeted adaptation in the event of changes
  When requirements, user interfaces or interfaces change, agents can highlight which test cases are affected and where regression effects may arise. This makes test maintenance more transparent and less haphazard.
• Reduced workload for documentation-intensive tasks
  Recurring tasks such as structuring test cases, summarising test results or preparing defect tickets can be accelerated. Business departments and test teams gain more time for evaluation, prioritisation and approval.
• Greater traceability for test management and compliance
  Agent-based workflows can document results, assumptions and intermediate steps in a more structured manner. This facilitates reviews and supports audit-proof traceability in regulated insurance processes.

Architecture and integration of “adesso test agents”

‘adesso test agents’ is designed as a flexible framework:

• Cloud-neutral: Can be deployed in various cloud environments (e.g. Google Cloud, Azure, AWS).
• LLM-agnostic: Depending on the complexity of the task, data protection requirements and costs, different LLMs are used – such as models from the Claude series, Gemini or OpenAI.
• Seamless integration: Integration with existing test management tools and workflows, e.g. via Jira/Xray. The agents can learn to work with different tools.

adesso deliberately opted for its own agent-based testing solution rather than relying on existing products. The aim was to create a solution that meets both our own requirements and the specific needs of our customers in regulated industries.

Unlike many agent-based testing tools available to date, “adesso test agents” focuses on high-performance agent-based test execution that:

• operates entirely without coded scripts,
• simulates genuine user behaviour using mouse and keyboard controls (“computer-based approach”) and
• can be flexibly integrated into a wide variety of test toolchains.

Increased productivity without compromising quality

‘adesso test agents’ is already being used in initial pilot projects – for example, at a major German insurer as part of a proof of concept.

This demonstrates particular potential in the structured derivation of test cases, the standardisation of test artefacts and the faster preparation of technical reviews.

adesso is also deploying the test agents in internal projects and gathering further experience to continuously optimise the framework.

Conclusion

Insurers face the challenge of further developing increasingly complex IT landscapes at an ever-faster pace. At the same time, skilled staff are becoming scarcer, regulatory requirements stricter and quality expectations higher.

“adesso test agents” offer a new approach to this challenge. Rather than replacing people, AI agents support business departments, business analysts and test teams in their time-consuming and documentation-intensive routine tasks, whilst complying with regulatory requirements.

In concrete terms, this means:

• For business departments: less manual preparatory work, fewer queries, more structured requirements
• For test managers: better test coverage, faster test case creation, traceable documentation
• For IT/Delivery: shorter release preparation times, more robust quality assurance
• For Compliance: greater audit readiness, documented decision-making processes

The result is a practical quality assurance process that operates more quickly, is more scalable and more consistent – leaving experts with more time for what really matters: the technical assessment and safeguarding of business-critical insurance processes.