An access-control leader cut Level 1 support queries by 30%
One agent grounded in a cleaned-up knowledge base now answers support teams, SaaS users, and resellers, powering 10,000 conversations a month.
Global scale, a broad portfolio, and a flood of Level 1 requests
A global leader in cloud-based access control and smart building solutions was feeling the strain of its own growth. A broad portfolio spanning access control, video, mobile credentials, smart home integrations, and third-party applications meant support agents combed extensive technical documentation for complex cases while a flood of repetitive Level 1 requests, login resets, basic configuration, competed for the same hours. For enterprise and property management customers, every delay in access control carries real cost: downtime, risk, and frustrated occupants.
Make the documentation AI-ready first, then deploy outward
The work started with the knowledge, not the bot. The company's technical content was locked in large, fragmented documentation spread across systems, hard for people to search and nearly impossible for AI to interpret consistently. Rapidflare and the customer ran a content-first program to make it AI-ready, then piloted the agent with a small group of support engineers before expanding it outward: into Slack, into the SaaS application itself, and finally onto the partner portal.
A third fewer Level 1 tickets, across three surfaces
Level 1 support queries fell by 30%, freeing agents for the complex cases. SaaS users resolve entry-level issues in the product, in the moment, which shows up in satisfaction and retention. Resellers on the partner portal get immediate, accurate product knowledge, lifting sales effectiveness and revenue. Together the deployments now power more than 10,000 conversations every month.
Making the documentation AI-ready first
The real blocker was never the model. It was that accurate answers were scattered through large, overlapping, sometimes conflicting documents. Before any agent went live, Rapidflare and the customer:
- Mapped and consolidated content into a single system of record
- De-duplicated and cleaned the data, removing outdated and conflicting entries
- Classified material as internal or external, so access stays secure and relevant
- Closed knowledge gaps surfaced by real AI query outcomes, resolving ambiguities the documents had carried for years
That structured foundation is what lets one agent answer consistently across very different audiences, from a support engineer mid-ticket to a reseller mid-deal.
From pilot to partner portal
Adoption followed the same discipline. A small group of support agents piloted the assistant inside their daily workflow and became its first advocates. From there the agent moved to Slack for the wider team, then into the SaaS application where end users self-serve entry-level issues, and finally onto the partner portal, where resellers ask product questions in the flow of selling. Each surface runs on the same knowledge base, with access scoped to the audience.