Bain Capital is building toward
agentic IT with Echelon.

Bain Capital onboarded ServiceNow in 2011 and has customized it ever since to meet its growing needs. Over a decade the instance drifted far enough with so little documentation that nobody could say how a given process actually worked. A 3-person team, many relatively fresh to the business, were expected to maintain it while leaders were left with little visibility to make critical decisions.

The platform had become so arcane, the business stopped even asking for changes & extensions, instead beginning to work around it. Work that belonged in a governed workflow moved into inboxes and spreadsheets instead. Every detour was process the ServiceNow platform could no longer see.

The goal was to get Bain to agentic ServiceNow: agents and skills running on the platform, and effective employee chat solutions. The legacy state made that impossible. Service Operations Workspace was switched on and Agent Assist came back blank, plugins sat misconfigured, and a decade of knowledge articles had gone stale.

As Alex Wong puts it: “we have no live chat today in the sense that every chat is either a live human or a ticket. So we have no analysis, no chat, no deflection.” Reaching a clean core meant a complete rebuild, leaving Bain's existing ServiceNow team with an insurmountable mountain of work & no budget to add to the team.

“The platform was so old, I tried to turn on Service Operations Workspace, and Agent Assist, the capability, was blank. It was not even returning any information.”

Agents fail on bad data. Agent Assist drawing on knowledge articles nobody has retired, a Virtual Agent routing against catalog items that no longer match the process, a CMDB whose description field holds something other than the description. That groundwork has to be true before anything agentic is worth switching on.

“The whole point of the refresh and bringing in Echelon,” Alex Wong says, “was so I could quickly evolve the existing processes, the knowledge base and everything, so that we could get to an agentic use case.” Descriptions backfilled across the application estate. Duplicate CIs consolidated and cross-referenced against live incidents. 703 unpublished software models surfaced in a single prompt. Catalog gaps mapped across the hardware lifecycle, and update sets analysed into a prioritised view of the technical debt underneath.

The operating model has moved with it. When an enhancement request lands in the queue, Echelon takes on the bulk of the work. It drafts the user story for the team to polish; Echelon then writes the technical detail of what it intends to build, and only once that is approved does it build. The human sits at every gate rather than doing the work.

“Instead of thinking about where to embed the AI, I want them to start thinking about where to embed the human.”

The platform team no longer moves blind. A decade-old instance that nobody could explain is now something anyone can query, and the answers come from the live instance rather than documentation that was never written. Leaders can see what the platform holds and what it costs, and process owners answer their own questions instead of joining a queue to ask them.

Capacity has changed shape too. What would have meant a hiring plan, a search and a budget request is instead a predictable software line item. The team that was too small to attempt a rebuild is now running one in-house. The audits, backfills and bulk operations that never reached the top of the queue get done.

Most of all, the business is back on the platform. Critical processes that had drifted into inboxes, spreadsheets and chat threads run on ServiceNow again, with approvals, an audit trail and access controls attached to every one of them. Every workflow that returns is one more the platform can see, govern and improve.

That is what AI readiness looks like. As Bain grows, the platform grows with it & documentation remains up to date. Workflows, processes, & core CMDB data remains clean, so AI returns the right answer every time.