Enterprise onboarding in 60% less time

The Challenge

The company draws on a large number of contractors, consultants, and seasonal workers as essential team members. The manual process of onboarding and offboarding the extended workforce spanned multiple departments (HR, technology, legal, and hiring managers) and enterprise software systems. While the process was painful for the staff involved, it complicated potential activities such as missing necessary equipment or access to systems and spaces needed to perform their duties.

The team identified 4 areas they wanted to improve:

• Visibility: where is a given person in the process, and who has the next steps?
• Avoiding rework: the existing process involved a complex web of interlocking requirements, and if it was not done correctly, it would require starting over again from scratch
• Efficiency: could they lower the overhead to only the essential tasks - for example, background checks or equipment provisioning
• Legal risk: new workers must be correctly vetted and approved before they start work

The Solution

An AI Agent implemented with Swamp, connected to Google Gemini Enterprise via the Agent-2-Agent protocol, managing the onboarding process. A manager requests to onboard a new employee, and a graphical wizard navigates them through providing all of the required data. Multiple Swamp workflows then orchestrate placing all the right information in each enterprise system. Each step in the process was visualized in Asana, with any manual tasks or approvals assigned directly to the department that can process it.

How it works

They began by describing the existing process using an agentic coding harness and used Swamp to help plan and design an end-to-end solution. Swamp then used adversarial evaluation planning techniques to guide the harness, breaking down the problem into discrete, deterministic components.

A key step in the process was how they used Swamp and the integrations it developed to reverse-engineer existing system workflows, such as ServiceNow, so that they could be deterministically orchestrated. Swamp would take in all the data about the employee to be onboarded and their specific requirements, then ensure all the proper fields and variables were populated.

Finally, they used the harness to build a stand-alone AI agent that used the process they built with Swamp as a back-end, connecting the underlying implementation to their custom front-end in Gemini Enterprise.

Benefits

The resulting solution delivered on every area the team hoped for:

• Visibility: it delivered full visibility of work through the process via Asana, allowing everyone involved, from the line manager to the legal team, to see where a given request was in the process.
• Avoiding rework: by collecting all the information up front and validating it with Swamp models, incomplete or impossible requests never enter the system; instead, they become prompts for more data in Gemini Enterprise.
• Efficiency: A 60% reduction in the total time to onboard a new worker by automating everything that can be automated and streamlining handoffs and notifications when it cannot.
• Legal risk: by removing variability from the onboarding process, the company ensured that things like background checks or other legal requirements would never be delayed.

The entire project, from initial scoping to proof of value, was completed in two weeks - with the bulk of the time spent in requirements gathering.

Lessons Learned

Trusting Swamp to handle implementation allowed them to focus on the final outcome for their stakeholders, rather than the implementation details. Swamp allowed them to trust working with AI across systems and delivering deterministic solutions consistently and securely. The combination of agentic harness and Swamp created end-to-end user-acceptance testing in sandbox environments. As a result of functional and system testing working so well, they focused on the Swamp workflow definitions rather than the underlying code. More time was focused on ensuring they delivered a great experience for their internal customers.

Futures

Using Swamp to streamline the existing enterprise onboarding processes serves as a blueprint for how to up-level existing business processes into the AI era. They are confident they can now go to any stakeholder in the company looking to automate a business process with AI, and they have a very high degree of confidence they know exactly what it will take to deliver.

Conclusion

Using Swamp to build AI Agents that implement complex business process automation delivered rapid results that matched the existing environment. The time saved by allowing Swamp to build for the desired outcome allowed for a deeper understanding of the internal customer requirements, ensuring the final outcome exceeded their expectations.