Instant, defensible answers from
5,000 pages of government policy
Government of Alberta staff must deliver accurate answers to complex policy and legislative questions under time pressure, but the
information they need is buried across acts, regulations, and lengthy PDFs. We built a citation-based legislative document search AI
that returns answers in seconds, with every response traceable to its approved source.
10%+ process efficiency gains, with frontline staff across 5 business units answering complex policy questions in seconds without escalation
THE CHALLENGE
Government staff are expected to answer questions accurately the first time. In Alberta, that means navigating a sprawling body of legislation — acts, regulations, policy manuals, and procedural documents — where the relevant detail is rarely on page one.
The result was a system that worked, but not efficiently. Staff relied on memory, manual searches through lengthy PDFs, and the institutional knowledge of colleagues who had been doing the work longer. When an answer wasn't obvious, the default was to escalate, pulling in senior staff who had their own work to do.
GoA contact centres were answering approximately 51% of incoming calls. The gap is either more staff, better training, or faster access to the right information. For the Government of Alberta, it was the third. An answer buried in a 400-page policy manual is functionally unavailable under call centre pressure. Frontline staff who couldn't find it in time couldn't use it.
What's more, a wrong answer to a benefits eligibility question, a licensing requirement, or a compliance obligation has consequences that outlast the call. So staff quickly needed answers that were accurate and traceable to an approved source.
THE SOLUTION
GovLab developed and deployed a ForgeAsk regulatory compliance AI assistant, built specifically for legislative research, designed to return answers that are both instant and defensible.
The system uses smart chunking to process policy documents in a way that preserves legal structure, so when a user searches for a specific provision or policy detail, the model returns the relevant section in context rather than a fragment stripped of its meaning. Every answer includes clickable citations that link directly to the source document, so staff can verify what they're reading before they act on it.
Strict source controls ensure ForgeAsk draws only from approved materials. It does not synthesize across unapproved sources or generate answers from general knowledge. What it returns is traceable, bounded, and auditable — qualities that matter in a regulatory environment where accuracy has compliance implications.
Built on enterprise-ready architecture, the system was deployed with minimal disruption to existing workflows. It is now in use across five GoA business units, covering 5,000 pages of policy and documentation.
WHY IT MATTERS
In government, an incorrect answer to a policy question can affect a citizen's access to services, a business's compliance status, or a staff member's professional standing. Accuracy matters, and so does the ability to show your work.
Government knowledge management AI that returns instant answers without citations is a risk. One that returns answers grounded in approved source documents, with every claim traceable, is a tool that frontline staff can actually trust and use.
That traceability is what transforms a knowledge tool from a helpful shortcut into infrastructure, something that can scale across departments, onboard new staff, and reduce the institutional knowledge dependency that makes teams fragile when experienced people leave.

THIS IS WHAT AI LOOKS LIKE WHEN EVERY ANSWER CITES ITS SOURCE.
See more of how we build for frontline staff, or talk to our experts. We'd love to hear about what your team is working on.

