TT PSC helped a cybersecurity product team reduce integration delivery time from approximately 2–3 months to around one month. By making reusable AI skills the default way to implement new integrations, the team enabled experienced engineers to coordinate multiple workflows in parallel, with automated testing and engineering review built into the process.

At a Glance

Industry: Cybersecurity / Cloud Infrastructure

Challenge: A growing number of similar but non-identical integrations across the organization’s Public Key Infrastructure (PKI) ecosystem created repetitive engineering work within a complex Certificate Lifecycle Management platform.

Approach: Reusable AI skills and agents became the default way to perform repeatable integration tasks, with engineers orchestrating the workflow and reviewing results.

Engineering Result: Integration delivery time decreased from approximately 2–3 months to around one month, and experienced engineers can coordinate multiple integrations in parallel.

Impact: Greater engineering capacity to expand product compatibility while retaining established quality and security controls.

The Engineering Challenge

As a Certificate Lifecycle Management (CLM) platform built to automate key processes across the Public Key Infrastructure (PKI) ecosystem, the product requires integrations with Certificate Authorities (CAs) and a wide range of infrastructure, cloud, and network systems.

Although many connectors followed a similar overall structure, each third-party system introduced different APIs, supported operations, authentication mechanisms and edge cases.

This meant developers repeatedly had to analyze documentation, prepare implementation structures, write similar code, create tests and produce supporting documentation.

The challenge was not simply to generate code faster. It was to identify which repeatable parts of the workflow could be supported by AI without compromising the engineering standards, security controls and technical ownership required for a cybersecurity product.

From Hackathon to Engineering Capability

TT PSC identified an opportunity to apply AI more broadly to the integration development process early in the engagement. Before introducing changes, the team first focused on understanding the product domain, engineering requirements and existing development workflow.

Based on this understanding, TT PSC initiated an AI hackathon bringing together its AI Champions and engineering team with the client’s Product Group Manager, Product Owner, Architect and AI Champion. The workshop combined practical AI experience from R&D and customer projects with the client’s product knowledge to identify where AI could support analysis, development, test planning and test automation.

The hackathon produced the first working framework for an AI-assisted integration workflow. This became the starting point for further development, refinement of reusable AI skills, expansion of automated testing and gradual incorporation of the approach into the real engineering process.

Engineering bottleneck → Hackathon → Working prototype → R&D leadership demo → Iterative improvement → Integration into the engineering workflow

The value came not from using a generic coding assistant, but from combining AI with product knowledge, repeatable engineering patterns and established development controls.

From AI-Assisted Tasks to an AI-First Integration Workflow

The initial prototype used AI to assist engineers with individual development tasks. As the team refined the approach, it codified product knowledge, implementation patterns and engineering standards into reusable skills and made them the default way to develop new integrations.

For new integrations developed through this workflow, engineers use AI skills to generate the integration code rather than writing it themselves. They initiate and guide the process through the relevant AI skills, review the generated implementation and test results, and identify what needs to change. Those corrections feed into subsequent iterations until the integration meets the required acceptance criteria and passes engineering review.

This marked the transition from AI-assisted tasks to an AI-first integration workflow. AI skills perform the implementation work, while engineers direct execution, evaluate results and retain ownership of technical decisions, quality and security. The same model enables an experienced engineer to coordinate several integration workflows in parallel.

How the AI-First Integration Workflow Works

The workflow covers selected stages of connector development, with engineering review and feedback guiding each iteration: Research & context preparation → AI-generated implementation → Unit testing → Test planning → E2E verification → Documentation

Reusable AI skills execute repeatable activities across the integration workflow: analysing third-party documentation, preparing implementation context, generating integration code, creating test scenarios and producing supporting documentation. A dedicated testing skill prepares scenarios for manual verification, which also provide the basis for automated checks executed through Playwright. Together with unit and E2E testing, this reduces repetitive verification work and enables broader validation of integration behavior.

A key enabler of the workflow is a knowledge base built according to the Open Knowledge Framework (OKF) standard, which serves as the authoritative source of domain and implementation knowledge for connector development. The knowledge base is bi-directional: AI skills and agents consume it as the source of truth when generating implementations, reviews and tests, while findings, implementation insights and newly discovered patterns identified during development can be fed back into the knowledge base. This creates a continuously evolving repository of organizational knowledge that improves consistency, reduces repeated mistakes and strengthens future integration development efforts.

Engineers orchestrate the workflow, review generated code and test results, and specify corrections for subsequent iterations. They retain responsibility for acceptance criteria, non-standard scenarios, technical decisions and final approval, while AI skills and testing agents perform the repeatable implementation and verification work.

Engineering Ownership and Quality Gates

Engineers retain ownership of acceptance criteria, architecture, technical decisions and final approval. An optional automatic mode allows the workflow to proceed through most intermediate choices without manual confirmation, while final acceptance remains with the engineering team.

Review and correction of AI-generated outputs typically take around one week and can be shorter when fewer issues are identified or the implementing engineer addresses them before review. The team often runs several review iterations, including reviews supported by agents using different models or focusing on specific aspects such as security and code readiness.

A continuously maintained connector knowledge base provides shared implementation guidance and product context. It supports both implementation and review, but does not replace verification: AI-generated code and tests remain subject to the product’s established engineering controls, including:

  • coding standards and reusable shared components,
  • automated testing and test coverage expectations,
  • linting and static checks,
  • code and security review,
  • E2E verification,
  • engineer approval.

The workflow is usually led by one engineer, with pair programming used where helpful. Less experienced engineers can also run it with support from the team, rather than relying exclusively on a single experienced champion.

Reusable skills embed engineering expectations into the workflow, while engineering review determines whether the integration is ready for release.

Results Achieved So Far

The team has:

  • reduced integration delivery time from approximately 2–3 months to around one month, with engineering records showing development completion typically within 3–4 weeks and, in some cases, around two weeks;
  • enabled experienced engineers to coordinate multiple integration workflows in parallel,
  • extended automated verification, including E2E testing,
  • standardized repeatable integration activities through reusable AI skills.

Additional value beyond new connectors

The value of the AI-DLC approach extended beyond developing new connectors. The team also applied the skills and knowledge gained to add functionality, make changes to the platform and its integration framework, rework connectors originally delivered by Professional Services, and rewrite older connectors implemented in a different technology.

This broader use is an additional engineering outcome: the initiative built reusable capability for both new development and the evolution of existing integrations and platform components. The team’s investment in AI-supported connector development therefore created value beyond the original delivery use case, supporting product enhancement and modernization as well as integration expansion.

The timing figures above relate to integration work; development completion should not be read as production release. Time savings for the other types of engineering work have not been quantified here, nor has the broader impact on product release schedules, customer acquisition or revenue.

Why This Matters for ISV R&D Organizations

For R&D leaders, the value of this initiative is not simply faster code generation. It is a structured, repeatable way to deliver integrations, combining reusable AI skills, product knowledge and established engineering controls.

Following the AI hackathon and subsequent implementation, the client’s VP of R&D recognised the team’s work and its demonstrated ability to deliver connectors consistently within approximately one month. The significance went beyond individual delivery results: the team had established a reusable, partly automated workflow for future integrations, supported by a growing connector knowledge base.

For a product that depends on external systems, this creates additional engineering capacity to expand out-of-the-box compatibility. Supporting more customer environments can help reduce adoption barriers and support future growth. These are potential downstream benefits, however; the impact on customer acquisition, revenue and overall product release schedules has not yet been quantified.

Although this case focuses on connectors for a cybersecurity product, it suggests a pattern that other ISV R&D teams can assess for workflows that are repetitive, governed by established standards and dependent on domain knowledge:

Repetitive engineering workflow → Codified domain knowledge → Reusable AI skills → Engineering review and approval → Repeatable delivery capability

The broader opportunity is to move beyond isolated AI experiments and make AI part of a defined engineering process. Knowledge and implementation practices become reusable across future work, while engineers retain responsibility for technical decisions, quality and release readiness.

From AI Experiment to Engineering Transformation

TT PSC combined domain knowledge, practical experience from its AI Champions and close collaboration with the client to turn an engineering bottleneck into a repeatable AI-first integration workflow. Reusable skills and agents carry out implementation and verification tasks, while engineers guide iterations and retain accountability for product quality.

Is your integration backlog slowing product growth?

Let’s explore how an AI-first engineering workflow could help your team deliver more integrations, expand out-of-the-box compatibility, and maintain your quality and security standards.

Discuss your integration workflow
Schedule a call