Golden Horizons
CAPABILITY · OPS & BACK-OFFICE
AI Knowledge Base
New hires ask questions in plain English and get cited answers from your own SOPs.
Audit this workflow →What it does
Indexes your SOPs, Drive, knowledge workspace, and team chat into a retrieval store. New hires ask questions in plain English and get cited answers. Ramp goes from months to weeks.
The six-to-eight week ramp is almost always a documentation problem wearing a training costume. New hires aren't slow because they lack aptitude. They're slow because the answers to their daily questions live in three places: the head of whoever's been there longest, a folder of SOPs nobody's touched since 2021, and a team chat archive that requires knowing what to search before you can search it. The tenured person becomes a walking helpdesk. The new hire interrupts them ten times a day. Both lose time they don't have.
The agent we build ingests what you already have, including SOPs in document storage, process docs in knowledge workspace, onboarding team chat threads, policy PDFs, and training recordings with transcripts, and indexes everything into a retrieval store purpose-built for retrieval. A new hire types a plain-English question like "What's the process for escalating a billing dispute?" The agent pulls the relevant SOP sections, summarizes the answer in plain language, and cites the source documents with links. No hallucinated policy. No guessing. When the answer isn't in the indexed knowledge base, the agent says so, which is useful signal on its own.
The secondary output is a gap report. Every question the agent can't answer confidently gets logged. After the first thirty days you have a ranked list of the documentation that's missing or out of date, the exact knowledge gaps silently costing you every time someone new joins. Most owners have never seen that list. It turns an invisible problem into a fixable one.
This is not a chatbot bolted onto a folder. The retrieval is tuned to your document structure. The citation format matches how your team references procedures. And when your SOPs update, which they should, the index re-syncs on a schedule you set, so the agent's answers stay current without manual intervention. Your tenured staffer goes back to the work only they can do. The new hire gets answers at 9pm when nobody's around to ask.
Use cases
- A law firm associate in week one asks how to open a new matter in the practice management system. The agent returns the firm's intake SOP, the conflict-check checklist, and the billing code reference, cited by document and section, without pulling a paralegal off a deadline.
- A dental practice's front-desk hire needs the protocol for a patient who requests records before their next appointment. The agent returns the practice's HIPAA release procedure with the correct form name and submission path, no manager required.
- A marketing agency's new account coordinator asks how to set up a client's first monthly reporting package. The agent retrieves the reporting template SOP, the naming convention guide, and the most recent example report from the client folder, all in one response.
- A regulated business's IT engineer onboarding to a security-sensitive project asks which security controls apply to their workstation configuration. The agent surfaces the relevant security controls control references from the firm's compliance binder and the project-specific system plan.
- A property management company's new leasing agent asks what to do when a prospective tenant's income falls short of the standard threshold. The agent returns the qualifying criteria, the exceptions approval workflow, and the required documentation, pulled from the current leasing policy.
- A physical therapy clinic's front-office hire asks how to handle a patient whose insurance authorization lapsed mid-plan-of-care. The agent retrieves the auth renewal SOP, the payer-specific contact list, and the visit-hold protocol, defusing a stressful first encounter with the billing process.
What’s included
- Fixed scope with written acceptance criteria before any build starts
- Customization layer for your brand voice and business rules
- Clean handover with documented runbook and live training
- Monthly ROI report for three months post-delivery
- Source code delivered to your GitHub on handover
What’s NOT included
- Third-party API subscription costs (billed to your accounts)
- Data migration from legacy systems
- Ongoing infrastructure costs after handover
How clients use this
Fixed-scope build with clean handover, documented ownership, and optional support for monitoring, maintenance, and minor changes.
Part of
Used in: Law Firms , Real Estate Agents , Construction Firms , Dental Practices
Questions Knowledge Onboarding clients ask
Which document types and sources can the agent actually index?
PDFs, Google Docs, knowledge workspace pages, Word files, plain-text SOPs, and team chat channel archives with transcripts. If a document lives somewhere with an API or export path, like Drive, knowledge workspace, document repository, knowledge workspace, or a shared network folder, we can pull it in. We don't index images or scanned PDFs without text extraction, and we don't index content you haven't explicitly scoped in. The first step of every engagement is a document audit, so you see exactly what goes in before anything goes live. Audio and video training content can be indexed if you have transcripts, and we can generate those from recordings as a pre-step if needed.
How do you prevent the agent from making up a policy that doesn't exist?
Retrieval-augmented generation means the agent only answers from what's in your indexed knowledge base. It doesn't draw on general internet knowledge for policy questions. When a question doesn't match anything in the index with sufficient confidence, the agent returns a defined fallback: it tells the user the answer isn't in the current documentation and suggests who to ask or where to look. We set that confidence threshold during calibration and tune it against your actual documents before go-live. No system eliminates every error, and we tell every owner that, so the agent should be treated as a first-pass reference, not a replacement for human judgment on high-stakes calls. For the bulk of onboarding questions that are procedural and well-documented, retrieval accuracy runs high.
What happens when our SOPs or policies change?
The index re-syncs on a schedule we configure with you, typically nightly for active Drive or knowledge workspace folders, with a manual re-trigger available when something changes mid-day and you need the agent to reflect it right away. When a document gets updated, the old version's chunks are replaced. When a document gets removed, it drops from the index on the next sync. You don't need to tell the agent which documents changed, because it scans the source for modifications automatically. With support, we handle configuration changes when your folder structure shifts or a new integration needs adding. Without ongoing support, we document the re-sync process and hand it off so your team can manage it independently.
Is this only useful during onboarding, or does it help current staff too?
Both, and most owners end up using it more for current staff than they expected. The onboarding case is the clearest, since new hires carry the most questions, the least context, and the highest interruption cost. But tenured staff use it differently. They don't ask how to do something. They ask where the current version of a thing lives. What's our current fee schedule for commercial properties? Which form does a patient sign for out-of-network billing? What's the updated process since we switched billing platforms? Those questions eat five minutes apiece when someone has to go hunting. The agent returns the answer in thirty seconds with a citation, faster than team chat-searching for it. The gap-reporting function helps everyone too: if a tenured employee's question comes back with no confident answer, that's a documentation gap the whole team shares.
How do we measure whether this is actually working?
We instrument three things from day one. Query volume tells you how many questions the agent fields per week and how much it's actually being used. Answer confidence distribution tells you what share of queries return a high-confidence answer versus a fallback, which shows where the documentation is solid and where gaps remain. Gap-log growth is the running list of unanswered or low-confidence queries, which becomes your documentation backlog. Golden Horizons can deliver a simple monthly summary of those three metrics as part of ongoing support. We don't promise a specific ramp-time reduction, because every team's baseline differs, but owners who look at query volume after thirty days almost always have a clear sense of whether the agent is carrying weight. The gap log alone tends to be the most valuable output in the first ninety days. It's the first time most operators have a concrete, prioritized list of what their documentation is actually missing.