Scaling Guidewire teams is not getting easier

The standard response to delivery pressure is to add capacity, and that carries constraints of its own. Guidewire-certified professionals are limited in supply and cost more each year. Ramp-up on a carrier’s specific products, configuration and integration landscape requires several months. Larger teams also require more coordination, which absorbs part of the capacity added.

Adding people does not change the underlying economics of delivery. When a large share of the work remains manual, a larger team simply performs more manual work at a higher overall cost. The pattern is visible in budgets as well: industry benchmarks place close to 70% of enterprise IT spend on running and maintaining what already exists, against 30% or less available for change, according to industry IT spending benchmarks.

“

Adding people does not change the underlying economics of delivery.

Every SDLC stage still depends on human effort

Across Guidewire engagements we have assessed, approximately 60% of delivery effort is manual. The effort is distributed across the full lifecycle, rather than concentrated in one stage.

• Analysis – Resources read requirement documents, trace the impact across products, rules and integrations, and write stories by hand. The same assessment is repeated each time a similar change enters the backlog, because the knowledge sits with individuals rather than in reusable form.

• Development – Gosu, PCF and rule changes are written manually, and they frequently repeat configuration patterns the team has already implemented for another product or state. Each repetition consumes a full development and review cycle.

• Testing – Test cases are authored and maintained by hand, and regression cycles are executed repeatedly across releases. Each new release adds to the testing burden, making validation a larger component of the overall delivery effort.

• Documentation and support – Documentation is produced after the work is complete, which makes it slow to write and quick to fall out of date. In production support, an engineer reads logs, reproduces the defect and traces it through configuration and integration code before a fix is written.

Manual effort limits scalability

This effort profile produces four consequences:

• Cost per change remains high – Effort is the unit of pricing, so every enhancement carries the full cost of analysis, development, testing and documentation, regardless of how similar it is to work already completed.

• Timelines extend – Regression cycles and documentation sit on the critical path. Both expand as the estate grows, so release duration increases even when the scope of each release does not.

• Release velocity reaches a limit – Teams arrive at a cadence they cannot improve further, because the constraint is human throughput rather than tooling or platform capability.

• Capacity applied to unimportant work – Senior resources spend a substantial share of their time on repeatable tasks, and architecture decisions, complex defect resolution and delivery improvement are deferred until that capacity becomes available.

All these effects compound. Change requests enter the backlog faster than the team can clear them, priorities move into later releases, and the business sizes its plans against what delivery can absorb.

AI is now entering the delivery lifecycle

AI is moving into the delivery lifecycle itself rather than sitting beside it as a personal productivity tool. It is being applied across analysis, development, testing, documentation and production support. Guidewire’s own product direction points the same way: its Qusar release introduces an Agentic Framework on Guidewire Cloud Platform, along with Developer Assistants that Guidewire states help teams ship features more than 40% faster than generic coding assistants and reduce development effort by up to 60%, according to Guidewire.

This reinforces the fact that true value depends on context. General-purpose assistants provide limited benefit on a Guidewire estate because they do not understand Gosu, PCF, the data model, or the integration patterns a carrier has developed over several years. AI grounded in these can produce configuration, test cases, documentation and defect analysis that a team uses directly. The result is higher throughput from the same team.

The question worth asking

Every insurer running Guidewire has already invested in a modern platform. The question is whether the delivery model around it is equally modern, or whether it continues to spend budget and time on work that no longer requires manual execution.

If 60% of Guidewire delivery effort is manual today, what should a less effort-intensive, AI-accelerated delivery model look like?