How CGI Slashed Manual Work and Scaled Smarter with Copado

The Challenge

Too Much Manual Work, Too Little Team

CGI's Transcend IP platform for child welfare is under constant, active development—new features delivered weekly to State agencies across the U.S. Managing five Salesforce orgs with 100–200 custom objects, no Apex code (only flows and OmniStudio), and strict FedRamp compliance requirements, the team needed to operate at maximum efficiency.

The reality was anything but. "It was a very manual process," explains Forrest Chang, Director of Solution Architecture at CGI. "The architect and product owner would work through design sessions, break up features into multiple user story titles, then work with BAs to explain the needs of each one and assign them to write the acceptance criteria and business specifications."

The numbers told a painful story:

• 23 hours to create a single feature with 10 user stories — PO and Architect: 4 hours each, BA: 15 hours
• 1–2 hours per user story from creation through refinement to dev-ready status
• Multiple rounds of refinement as the product owner re-reviewed functional specs and the architect wrote technical specs

And testing wasn’t any better. "Testing was immediately outdated as soon as new development came through," says Chang.

With Transcend IP under active weekly development, a single full-time QA resource was spending 40 hours a week on manual testing — running through user flows, writing Robot Framework scripts by hand, and still never keeping pace with the rate of change.

"It was impossible and wasteful to have one human resource performing the tests constantly. They had to start over every week with new changes," comments Chang. For a team already stretched thin — ranging from just 1 to 5 people at any given time — this level of manual overhead was unsustainable. The team turned to Copado for help.

The Solution

One Platform. Three Transformations.

CGI didn't adopt Copado in pieces — they went all in, using Copado as the core engine to accelerate their Transcend IP.  This allows them to industrialize the development and documentation of their public sector solutions before they even reach a customer.

When it came to implementation, Copado CI/CD came first, replacing a fragmented process of manual SFDX package deployments through VS Code and the CLI. Then came Copado Robotic Testing and Copado AI, adopted simultaneously to tackle the twin bottlenecks of testing and planning.

Before Copado, CGI's deployment process required manually remembering every changed component and packaging it correctly — a process prone to error and impossible to scale. With Copado's Source Format Pipeline, the team moved to a structured Dev → QA → Prod workflow, attaching user stories to promotions and validating before each environment promotion.

"The new Commit process is a huge improvement," says Chang. "It helps me find the exact metadata I need very quickly." Copado's commit tooling also helps the team navigate the circular reference issues common in Agile Salesforce delivery — where bite-sized user stories can create dependency conflicts when promoted to production. While Salesforce's metadata quirks remain a challenge, Copado makes them manageable.

Ending a Laborious, Manual Testing Cycle with Automation

CGI had already experimented with Robot Framework and Selenium, writing test scripts by hand. But manual scripting was slow, fragile, and couldn't keep pace with weekly releases. 

With Copado Robotic Testing, CGI automated their most critical and complex test scenario: the intake process for a hotline worker receiving a child abuse report — a workflow involving heavy OmniStudio components and bulk record generation.

"We no longer need a manual tester performing the Sisyphean task of testing the same thing over and over again for every change," says Chang. "Copado Robotic Testing has helped us catch defects caused by Salesforce release updates that changed the behavior of integration procedures."

Transforming Planning with Copado AI

When CGI discovered Copado AI, one capability immediately stood out: PlanAgent. "The ability for myself as the architect and our product owner to have a single design session, collaboratively define our feature prompt, and have Copado AI build all the user stories and correctly map them to features, applications, projects, epics, and user story record types in Copado automatically — that was game-changing," says Chang.

The learning curve was minimal. "It's very familiar to anyone who uses other gen AI LLM tools," Forrest notes. The team was proficient within a day or two.

CGI also discovered unexpected advantages. The deep integration between Copado AI and their CI/CD org meant the agents could generate and edit records directly — removing tedious data entry that had previously fallen on the team.

And BuildAgent opened a new frontier. "The BuildAgent can generate decent deployable Salesforce flow metadata from a prompt," explains Chang. "This is superior to anything Agentforce can do today. I've also tried with other general-use LLMs and not been able to even get a deployable flow metadata XML."

Copado’s ReleaseAgent has also played a valuable role for CGI, discovering hard-to-diagnose deployment errors. It helps CGI identify API version conflicts — such as a flexipage:grid component available in Dev preview but not in QA — and even resolved a cryptic OmniStudio error that Salesforce support had flagged with only an error code. When it finds the answer, it saves 2–4 hours of troubleshooting per incident.

The Outcome

Measurable Impact Across the Entire Delivery Pipeline

Copado operates under FedRamp Moderate requirements, providing the process controls CGI needs: tracked deployments with audit logs, pipeline deployment permissions, access controls to pipelines and environments, Salesforce-native SSO/MFA integration, and full git-based version control. Copado AI agents don't have access to real client data — the BuildAgent generates metadata files for manual deployment, keeping sensitive information fully protected.

With Copado CI/CD, Robotic Testing, and AI, the CGI Transcend team’s DevOps is transformed. Before Copado AI, the team’s workflow to create user stories looked like this: Design sessions → Manual story breakdown → BA assignment → Spec writing → Multiple refinement rounds → 23 hours per feature

With Copado AI they were able to optimize their workflow significantly, saving the team 83% of their time: Design session → Feature prompt → AI-generated stories → Review and feedback → 4 hours per feature, no BA needed

"With every new feature we design and release now, the PO and Architect will have a design session and may spend 1-2 hours to go from whiteboarding a feature to having the full feature and 5-20 user stories generated and ready for technical refinement and dev loaded into our CI/CD org," Forrest explains.

The rest of the pipeline remains consistent—using CI/CD with user stories attached to promotions and releases, validating prior to deploying to the next environment, and regression testing. What once took 5–10 minutes of manual effort per regression runs now executes autonomously in under 2 minutes — freeing the team to focus on development while tests run in the background. The 40 hours per week of manual testing? Eliminated entirely. The team now delivers more features, with the same or fewer resources, changing the way the small team works.

The Future

Expanding AI Success with BuildAgent

CGI's journey with Copado AI is just beginning. Looking ahead, Forrest sees significant opportunities to expand AI usage across Copado’s AI agents.  

For CGI, the Copado journey is just beginning. With Copado CI/CD, Robotic Testing, and AI agents all running in concert, the team is now focused on pushing each capability further.

Looking ahead, CGI plans to scale their test library by having business SMEs generate test scripts while running through user flows — removing the need for a dedicated testing resource altogether. The goal: a self-sustaining test automation practice that grows with the product, not against it.

On the AI side, BuildAgent is next in line for expanded use. A large part of Salesforce development involves slow, repetitive manual steps — creating objects and fields, managing permission sets, configuring quick actions. CGI sees BuildAgent as the key to accelerating these tasks. The team is also exploring data seeding capabilities to generate realistic, relationship-aware data for demo and test sandboxes.

And there's a smarter deployment future on the horizon too. Forrest envisions a Copado commit process that's aware of the target org's existing structure — intelligently determining whether a dependency already exists in the target environment, rather than flagging it as missing. This would eliminate a significant source of unnecessary deployment complexity in Agile Salesforce delivery.