SEPTEMBER 2023 RELEASES
September 2023 Releases
The RAG Playbook: Structuring Scalable Knowledge Bases for Reliable AI Agents
The knowledge base for your AI agent isn’t just a dusty FAQ archive – it’s the brain fuel that powers accurate, on-point conversations. Building a knowledge base (KB) for a human support team is one thing, but doing it for an LLM-powered AI agent is a different game entirely. What works for humans– long articles with anecdotes, context, and even a dash of marketing fluff–can utterly confuse an AI agent.
The result? Sometimes it’s harmless, like the time a customer asked for your cancellation policy and the bot earnestly replied with your company’s mission statement, three paragraphs about how you value innovation, and a cheerful “Have a great day!” Other times, it’s not so harmless. Think confident but incorrect therapeutic advice or an unauthorized “discount” it just invented on the spot; not exactly what you want from your brand’s digital representative.
In this guide, you’ll see exactly how to format and structure a knowledge base so your AI Agent stays accurate, relevant, and on-script. We’ll break down why traditional human-centric KBs fall short, and show how smart formatting can reduce hallucinations, improve accuracy, and keep your AI Agent’s behavior on the rails.
We’ll cover best practices like single-topic chunking, concrete instructions, explicit outcomes, and explain what to put in the prompt vs. the KB. And don’t worry if “chunking” or other terms sound like AI-speak now. By the time you finish this guide, you’ll know exactly what they mean and how to put them to work, without needing to wade into the full Retrieval-Augmented Generation (RAG) playbook. Let’s dive in!
.webp)
Why Human-Centric Knowledge Bases Fall Short for AI Agents
A knowledge base written for human consumption often contains knowledge clutter that doesn't bother a human reader, but can wreak havoc on an AI agent’s responses. Before we get to solutions, let’s identify the biggest content culprits:
- Irrelevant Context and Fluff: Human-facing articles often start with a narrative or extra background – “At Acme Co., we value our customers…” – before the useful info. An AI agent retrieving this may latch onto the fluff and respond with irrelevant platitudes. Unnecessary context dilutes the useful facts. The AI might even mistake a marketing blurb for an answer and go off on a tangent. Case: From Compliance Guide to Chaos Gremlin. NYC’s small-business chatbot told entrepreneurs it was fine to do illegal things (like taking workers’ tips), then doubled down with outdated wage info. It’s the content-equivalent of quoting a press release when the user needed the statute.
- Assumed Prior Knowledge: Many documents assume the reader has some context. For example, an internal guide might say “follow the usual refund process” without detailing it. A human agent knows the “usual process,” but an AI agent has no memory of knowledge outside what’s in the retrieved text. Missing context forces the AI to guess, which is a recipe for hallucination. If the KB chunk doesn’t spell something out, the model may fill the gap with its own invention, as in the case of Air Canada’s website chatbot. The AI confidently invented a bereavement-refund process that didn’t exist, so convincing that the tribunal made the airline compensate the passenger.
- Conflicting or Redundant Instructions: Large knowledge bases often accumulate overlapping content. You might have two docs with slightly different refund policies, or an old procedure lurking alongside a new one. Human agents can figure out which one is current; an AI might randomly pick or (worse) mash them together. This leads to unpredictable, inconsistent answers. Having too much redundant or conflicting info can confuse the AI and make its behavior less predictable. In one case, an AI was given two sources on whether a fee is required – it sometimes said “no fee” and other times “a fee applies,” purely based on which chunk vector search was fetched first.
- Ambiguous Language: Human writers might use polite or vague phrasing like “you may want to consider doing X” or “Try Y if possible”. This subtlety can baffle an AI agent that craves clear directives. Ambiguity in knowledge content can increase the chance of the AI choosing an incorrect interpretation. For instance, if a policy note says “late cancellations might incur a charge”, the AI might just omit the fee entirely. Without concrete wording, the model’s guesses can go astray.
The Impact: Hallucinations and Off-Script Behavior
When a knowledge base has the issues above, an AI Agent’s accuracy and reliability plummet. The model might confidently fabricate answers or instructions that were never in your company policy, simply because it thinks that’s what you implied. And despite common prejudices, these hallucinations stem not from the model “being goofy,” but from bad or poorly retrieved data – essentially, the AI is being “gaslit” by a messy knowledge base. The good news: by reformatting how knowledge is structured, we can ground the AI Agent in the right context and cut down on hallucinations dramatically.
Best Practices for Structuring an AI-Ready Knowledge Base
Designing your knowledge base for AI agents requires a more disciplined, machine-friendly approach to content. Here are the top best practices to adopt:
- Single-Topic Chunking: Break your knowledge into bite-sized pieces, each focused on a single topic or question (a single-topic chunk). Avoid giant all-in-one documents. Smaller, well-scoped chunks ensure that when a retrieval occurs, the AI gets just the specific information it needs – nothing more, nothing less. As a rule of thumb, think a few paragraphs (or even a few sentences) per chunk. Regal’s platform, for example, automatically uses semantic chunking to split sources. This granular approach keeps retrieval precise and prevents irrelevant text from piggybacking along.
- Concrete Instructions (No Filler): Write the content like you’re instructing a particularly literal intern – clear and direct. Use imperative language for procedural info specific to the AI Agent (e.g. “Call custom action “Submit” to save changes” instead of “You could then click submit”). Remove the “fluff” that you might include for a human’s reading pleasure. The AI doesn’t need motivational quotes or company history in the middle of a how-to. By sticking to concrete steps or facts, you reduce ambiguity. The goal is that any sentence the AI might quote is immediately useful and actionable. If the agent needs to explain something to a user, give it the exact phrasing or numbers it should use. For example, instead of a blurb that says “our plans are affordable”, give a factual statement: “Plan A costs $30 per month, Plan B costs $50 per month.”
- Explicit Outcome Statements: Wherever applicable, state the desired outcome or resolution in the knowledge chunk. This helps the AI understand the point of that information. For instance, after listing steps for a cancellation process, add a line like “Outcome: Appointment is canceled and a confirmation email is sent to the customer.” This explicitly tells the model what the end state should be. Outcome statements function as a sanity check – they remind the AI “this is what you’re trying to accomplish.” If the user’s question is resolved by that chunk, an outcome statement ensures the agent’s answer includes confirmation of that resolution. It also helps prevent the agent from stopping short or going on an unrelated tangent. Essentially, you’re baking a mini-summary of “why does this info matter?” into the content.
- Contextual Reminders and Applicability: Give context about when or to whom a chunk applies, especially if your knowledge base covers nuanced scenarios. For example, if a policy differs by region or customer type, don’t make the AI infer that – state it outright. A chunk title or first line might be “Return Policy (California customers)” or “Procedure: Password Reset for Mobile App (Android)”. This way, the retrieval system can grab the chunk that matches the user’s context, and the AI Agent will see in the content itself who/what it’s for. Contextual labels reduce the chance of the agent using the right answer in the wrong situation. In Regal’s RAG system, AI agents can pull region-specific data when appropriate – e.g., if a contact is in California, the agent fetches the California-specific info first, keeping the conversation compliant and on-target. Our customers achieve this by structuring their KB into clearly scoped chunks (by product, region, scenario, etc.) and adding one-line intros like “Use this info when <these_conditions> are true.”
- Reduced Ambiguity: Write in a consistent style and use the same terminology that users and the AI Agent will use. If your product is called “AcmePhone Pro” in one doc and “AcP Pro” in another, choose one and stick with it (or at least mention the alias in the same chunk). Avoid idioms or internal slang that the model might not interpret correctly. If a process must never be done a certain way, say “Do not XYZ” rather than “XYZ is not recommended” (which the model may soften or ignore). The key is to leave no room for the AI to wonder what you mean. One trick: read the content as if you have no other context – would you 100% understand what to do or say? If not, clarify it. This also means removing any conflicting guidelines: ensure each piece of knowledge gives a single, unambiguous source of truth on that topic. If two chunks even appear to disagree (e.g. one says “usually do A” and another says “in some cases do B”), add context or combine them into one explicit decision-tree chunk. Your AI Agent can’t do critical thinking on clashing instructions – it will just pick one side. Cleaning up these ambiguity and consistency issues will greatly improve the agent’s accuracy.
By following these practices, you transform your knowledge base from a messy wiki into a lean, mean answering machine. A well-structured KB means the RAG system can grab the right facts quickly, and the LLM can trust what it sees. It’s the difference between an agent floundering through irrelevant paragraphs, versus confidently citing a single crystal-clear paragraph that directly answers the question. As a bonus, a slimmed-down, relevant knowledge chunk keeps the token count low and speeds up responses – no more prompt bloating with unnecessary text. In other words, you get faster, more accurate answers grounded in your authoritative content.
Prompt vs. Knowledge Base – Knowing What Goes Where
One common question we get is how to split duties between the AI agent’s prompt and the knowledge base. Both are essential, but they serve different purposes for your AI agent:
- The Prompt is the script and personality. Use it for behavior, tone, decision rules, and short, always-needed info (like disclaimers, FAQs, or guardrails). Think of it as the “how to act” layer.
- The Knowledge Base is the agent’s memory. Use it for detailed, dynamic, or frequently updated content (like policies, product manuals, or regional variations). Think of it as the “what to say” layer.
The general rule of thumb is: keep prompts lean (stable rules and flow), and put the rest in a structured KB so the agent can fetch facts on demand. This avoids prompt bloat while keeping responses accurate and scalable.
For a deeper dive on context engineering and when to use prompts, KBs, or custom actions, check out our full guide.
Knowledge Bases That Make Your AI Agent (and You) Look Smart
A well-structured knowledge base is the secret sauce behind highly effective AI agents. By eliminating irrelevant context, breaking knowledge into focused chunks, and writing instructions that are unambiguous and outcome-oriented, you ground your AI in facts and context, dramatically reducing the odds of it veering off into nonsense or error. In turn, your customers get faster, more accurate answers. Your AI Agent isn’t playing detective to figure out what the policy really means; it’s confidently quoting the correct, up-to-date information you gave it.
To recap, when building a knowledge base for LLM-powered agents, remember:
- Format is king – Structure your content for retrieve-ability. Clear headings, short paragraphs, and markdown bullet points (if applicable) all help the vector search do its job and help the AI parse the info quickly.
- Keep prompt and knowledge in their lanes – Use the prompt for your agent’s personality and always-needed logic, and the KB for the details that can be fetched on demand. This keeps your AI’s “mind” uncluttered and focused.
- Continuously curate – An AI knowledge base isn’t “set and forget.” Monitor your agent’s outputs; if you catch a weird answer, trace it back to the source chunk. You might find an opportunity to edit that chunk for clarity or split it further. Remove outdated info promptly to avoid misinformation. Essentially, tend to your AI knowledge base like a garden – prune the bad, nurture the good.
- Test in realistic scenarios – Use Regal’s agent testing tools to simulate queries and see what it pulls from the KB. This will show you whether it’s grabbing the right chunks and phrasing answers correctly. If not, tweak the content or add more guidance (either in the KB description or the prompt) until the AI consistently does the right thing.
By investing effort upfront in building a clean, AI-optimized knowledge base, you’re setting your virtual agent up for success. You’ll get predictable, reliable agent behavior, faster and more accurate answers, and conversations that stay on-message and on-brand – all at scale, without having to stuff an ever-growing list of facts into your prompt. It’s a win-win: the AI Agent is happier (less confused), and your customers are happier (more satisfied with the help they receive).
Finally, remember that knowledge bases and AI agents are a partnership. The smartest AI won’t shine if it’s fed garbage info – garbage in, garbage out. But with a well-built knowledge base, even a relatively small LLM can outperform a larger one that’s unguided, because it’s grounded in the truth of your data. You’re the expert on your business; by translating that expertise into a structured knowledge base, you make your AI agent an expert too.
Play around with these tips and watch your AI agent’s IQ (and CSAT scores) soar. And if you need a little help crafting the perfect AI knowledge base, schedule a demo with us!
Frequently Asked Questions
What is Retrieval-Augmented Generation (RAG)?
Retrieval-Augmented Generation, or RAG, is a technique that improves AI responses by retrieving relevant information from a knowledge base before generating an answer. Instead of relying only on training data, the AI pulls context from documents, guides, or internal data sources, which helps produce more accurate and up-to-date responses.
Why do traditional knowledge bases perform poorly with AI agents?
Traditional knowledge bases are often written for humans, not machines. They typically contain long documents, mixed topics, and unclear instructions. When AI systems retrieve this type of content, they may struggle to identify the most relevant information, which can lead to inaccurate responses or hallucinations. Structuring knowledge in smaller, clearly defined chunks helps AI agents retrieve and use the right information
How should a knowledge base be structured for RAG systems?
A RAG-ready knowledge base should organize information into clear, single-topic sections with concrete instructions and defined outcomes. Structured content allows the retrieval system to surface the most relevant context for a question. This approach improves the reliability of AI-generated responses and helps scale knowledge bases for production AI systems.
What types of content should be included in an AI knowledge base?
AI knowledge bases can include product documentation, support articles, internal policies, troubleshooting guides, and FAQs. These resources give AI agents reliable context during conversations and help them answer questions accurately by retrieving relevant information at runtime.
How does a well-structured knowledge base reduce AI hallucinations?
Hallucinations often occur when AI systems lack reliable context or retrieve irrelevant information. A well-structured knowledge base improves retrieval accuracy by organizing content clearly and ensuring that AI agents pull information from trusted sources. This makes responses more grounded in real documentation and reduces the likelihood of incorrect answers.
Treat your customers like royalty
Ready to see Regal in action?
Book a personalized demo.
Thank you! Click here if you are not redirected.
Oops! Something went wrong while submitting the form.
Founded in 2020, Regal is an enterprise voice AI agent platform for contact centers. Regal helps businesses build, deploy, and manage autonomous AI agents across sales, support, and operations teams.
Latest Blog Posts
.webp)
Pinpoint Critical Knowledge Base Gaps with Regal Improve
Customer conversations are inherently dynamic, and Regal Improve’s Knowledge Base Coverage tracks changes in retrieval performance over time and at scale. This makes it easy to verify that KB updates increase coverage and if your documentation is keeping up with evolving customer requests.
November 2025 Releases
November's releases make it easier to scale your transfer-heavy call flows, with new tools that align dialer pacing to downstream agent capacity and streamline complex handoffs. You’ll also see improved post-call workflow accuracy, with reliable data extraction and disposition outcomes driven by LLM configurability and earlier validation.
.webp)
Teaching Voice AI to Behave: Prompting Principles Every Parent Already Knows
Building a great AI voice agent isn’t so different from parenting a five-year-old: clarity, examples, and structure make all the difference. This post breaks down proven prompting principles—from positive framing and persona design to conversational guardrails—that help AI agents sound natural, stay on task, and follow directions every time.
Announcing Regal + Rime: Human-Like Voices at Enterprise Scale
Regal is partnering with Rime to deliver ultra-realistic, emotionally nuanced AI voices for enterprise contact centers. Discover new voice options that capture human qualities like pacing, resonance, and subtle expression, and learn how Rime’s technology powers conversations that build trust at scale.
%20(1).webp)
Announcing Regal + Cartesia: Expanding Voice Options for Enterprise AI
Regal is partnering with Cartesia to deliver ultra-low latency, high-fidelity AI voices for enterprise contact centers. Explore new voice options, hear them in action, and learn about Cartesia's industry-leading voice AI offerings.
%20(1).webp)
.webp)
%20(1).webp)
Beyond CCaaS: How to turn your Contact Center into a Revenue Center
Why are so many contact centers stuck being cost centers? Why do they keep looking for cost reductions instead of revenue increases? Bad data and technology decisions are at the core of this issue. We need "marketing automation" (aka unified customer profiles, event-driven systems and a/b testing) like Regal.io in the contact center to empower teams to start looking for incremental revenue instead.
The Next Step Forward in AI will come from Consumer Businesses
Consumer businesses are implementing AI-enabled customer experiences without any change in behavior on the part of consumers, which is leading AI to become common with more speed than past transitions like the internet and the smartphone. And it means that the next step in AI will not come from the LLM providers. The next big step forward in AI rests in the hands of consumer businesses.
Innovating on Patient Care | B2C Sales Podcast Episode 7
Explore Eric Hauser's remarkable career journey from GovTech to healthcare disruption at Cadence, highlighting the transformative power of innovation, collaboration, with a special focus on the pivotal role of Regal in driving patient engagement and outcomes.
How SoFi, Perry Health & Allstate Personalize CX at Scale
Discover how Regal.io's AI-powered personalized outreach solutions are revolutionizing outbound sales and customer experience across industries like healthcare, finance, and insurance in our latest eBook, "Modernizing Outbound Contact Centers: How to Treat Millions of Customers like One in a Million."
.jpeg)
Introducing Collaboration-Native Features to CCaaS
Combining collaboration functionality into CCaaS workflow tools invites more cross-functional users from a company to participate in designing the end-customer experience, leading to better omni-channel orchestration and customer outcomes. Learn more about Regal's collaboration features.
How Regal.io is Turning the $40B CX Industry on Its Head
Jon Heaps, former VP, Channel at Observe.ai, Talkdesk, and inContact interviews Alex Levin, Co-Founder & CEO of Regal.io, about the history of the contact center industry and some of the key challenges teams making outbound calls face as customers demand more online experiences.
New SMS & Branded Caller ID Rules Are Enhancing Customer Experience
Are you using Branded Caller ID for outbound calls or SMS? New SMS and Branded Caller ID regulations rolled out in Q2 2023 that you MUST KNOW ABOUT. The new regulations require that every company register their SMS campaigns and Branded Caller ID campaigns before being allowed to send texts or brand calls.
Announcement: Regal Conversation Intelligence
Regal’s Conversation Intelligence drives higher conversion rates for B2C sales teams. We use both the traditional QA/coaching tools, and brand new conversational triggers that allow you to update customer profiles and send automated follow up messages based on what is said in a conversation.
Caller ID vs. CNAM vs. Branded Caller ID: What’s the Difference?
Caller ID, CNAM, and Branded Caller ID all help identify an incoming call and, therefore, increase call answer rates. Find out how they differ and how outbound phone sales teams can utilize each to reach prospects and customers effectively.
How Regal Helps Consumer Lenders Achieve Growth in Tough Times
Our Co-Founder & CTO, Rebecca Greene, spoke with journalist Tony Zerucha at the recent Fintech Nexus USA 2023 event about how consumer lenders are using personalized phone and SMS sales outreach to obtain growth despite high inflation and rising interest rates.
Outbound Call Center: What It Is, How It Works & Best Practices
The fastest growing High Consideration brands have invested heavily in outbound call centers for sales and marketing. With the right technology, staffing and measurement, you too can build a high performing outbound call center to help hit your growth goals.
Meet Regal Call Branding™: Branded Caller ID for 400M+ US Devices
We are excited to announce Regal Call Branding™ including Branded Caller ID. Available on all 400M wireless devices in the US (as well as most of Canada and the UK), Regal Call Branding™ puts you in control of what people see when you call their cell phone – including showing your brand and/or logo.
Q&A with Perry Health’s Scott Chesrown
Get behind-the-scenes insights about Perry Health’s 23% revenue increase with Regal.io. Perry Health’s Scott Chesrown sat down with Regal.io and discussed how they implemented outbound sales technology, integrated new capabilities into their process, and built on their early success.
Drive More Contact Center Revenue With These 6 Holiday Tips
It’s the holiday season and you’d like to hit your December revenue goals. Outbound B2C sales during the holidays can be a major performance driver – if you're prepared and you have the right tools in place. Use these seven tips to get your outbound B2C sales in shape for holiday success.
The Rise of Voice AI Agents: Real-World Deployments Across the Enterprise
In this AI4 fireside chat, Regal.ai COO Sahil Mehta and TaskUs CCO Jarrod Johnson explore how enterprise organizations are deploying voice AI agents across customer experience, support, and sales. They share real-world deployments, discuss quality and ROI trade-offs, and outline best practices for aligning operations and product teams when adopting AI agents at scale.
.png)
.png)
.png)
October 2026 Product Recap
Last month, we shipped releases that improve your contact center operations by extracting business insights with Copilot and reducing wasted agent minutes with our new Voicemail Detection model. The result is more reliable, productive AI agents that can drive revenue for your business.
.png)
How We Built the Voice CX Benchmark
Enterprises need voice agents that hold a natural, human-like conversation while moving each call toward its goal and staying within compliance boundaries. The Voice CX Benchmark scores models on exactly that: how well they progress the call turn by turn, how safely and clearly they communicate, and how quickly they respond.
.png)
Introducing Regal’s Voicemail Detection Model: How We Reduced Voicemails Routed to Agents by 45%
With Regal’s next-generation Voicemail Detection model, you can now classify voicemail, iOS or Google call screeners as separate categories with very high accuracy. This enables screener-specific handling, ultimately reducing missed agent connections.
.png)
September 2026 Product Recap
Last month, we shipped releases that make it easier to build, monitor, and test agents in one interface. Copilot Canvas brings building, editing, and testing into one interface and it can work straight from your files. Multi-state sticky notes let your team document decisions directly on a live agent.
.png)
.png)
August 2026 Product Recap
Last month, we launched features that extend your management of AI agents via MCP and CCaaS platforms. Debug based on agent versions or simulation runs, and manage AI analysis data points via Copilot and Regal MCP. As a Five9 AI Agent Connect partner, Regal now has a dedicated integration API built for Voice AI agents.
.avif)
.webp)
.webp)
.webp)
Dialing Compliance: How We Scaled Outbound Calling Reliably
Behind a fragmented and ever-changing regulatory landscape is a simple goal: protecting people from unwanted or poorly timed outreach and giving them control over how they’re contacted. In practice, that responsibility translates into a set of checks that must be validated.
.webp)
No Need to "Press 1": How AI Phone Agents Go Beyond Automation
AI phone agents aren't just better phone trees. They are autonomous systems that can understand caller intent, retrieve relevant information from your CRM in real time, execute actions, handle objections and clarifications, and complete the conversation without human involvement.
.webp)
.webp)
The PII Question Every Regulated Industry Asks Before Deploying AI Agents
When prospects in healthcare and financial services hear "PII," they often imagine bulk data transfer: the idea that deploying an AI agent means copying a database somewhere. In most modern deployments, that's not what's happening.