With general use of open-source generative AI steadily ticking up, most of us are over the initial awe of getting any response at all and the questions of where exactly the responses are coming from and whether they’re trustworthy are getting louder.
Take, for example, the case of the “rogue stat.”
With sellers, marketers, entrepreneurs, and LinkedIn gurus running to ChatGPT for low-effort content, random, questionable statistics are more pervasive in business rhetoric than ever. Since these kinds of stats are common in the older blog posts that populate the internet and were used to train open-source LLMs like ChatGPT and Perplexity AI, they’re also common in the responses that come from those tools and even harder to fact check.
Here’s how it goes: You’re scrolling LinkedIn and an influencer you like posts a stat that catches your eye. It links to a blog post as the source so head over there to find out more. When you go to the blog post, that stat is linked to an older blog post. You follow this pattern for at least 4-5 clicks through what feels like a maze of old blog posts and analyst reports from 12 years ago only to eventually land on a 404. Now you’re questioning your trust in that stat or even that influencer.
This is the experience you don’t want your customers to have with your content, in whatever form it takes — blog posts, slide decks, social posts, etc.
What if you could make sure you and your team don’t pull rogue stats from generative AI and only get answers that are accurate and relevant to your audience? A capability called Knowledge Scopes for generative AI makes that possible. Here’s a look at how Knowledge Scopes work and why it matters to both know and control where your AI tool is pulling its responses from.
It’s 2025. Do you know where your AI data is coming from?
We know chatbots like ChatGPT are trained on vast quantities of online data and can now crawl the internet for up-to-date responses. But we’ve seen that they’re not always reliable. Most of us have experienced AI hallucinations and inaccurate results when using ChatGPT, like in the case of the rogue stat.
In fact, many open LLMs (large language models) now come with warnings to check responses for errors.
But even that’s not good enough when you’re using AI for business. You need to be absolutely certain that you’re giving your customers accurate and reliable information. You need to be able to trust your sources. So you need to know exactly where your AI data is coming from.
Broadly, there are two classes of data that business AI tools use: internal and external.
Internal sources include proprietary company datasets such as specific company documents, training materials, and internal databases. Internal sources are limited to your company’s knowledge base, CRM, and content management system.
They may be further constrained by access permissions for people at different levels of the company. But you know that the data has been vetted and is aligned with your company’s goals, so you can feel more confident in its accuracy.
External sources can include everything from web searches, social media, and market research reports to Large Language Models (LLMs), third-party datasets, and government publications.
These sources are clearly much more diverse and expansive, opening up vastly more knowledge for AI tools and platforms to draw from. However, you have less control over those sources and fewer guarantees that they are reliable, accurate, or in line with your business priorities and values.
And we’ve all seen how that ends up, right?
Clearly, internal and external sources each have their benefits and drawbacks.
As a business, your best option is to draw from both wells. If you limit yourself to a single source, you face trade-offs between accuracy and scope, reliability, and range.
On the other hand, multiple sources help you triangulate and cross-verify your data, making for more reliable and relevant outputs.
For example, Bigtincan’s GenieAI lets you configure your Knowledge Scopes so you control what sources your AI tools will pull from. Choose a mix of:
- Company content (or a subset): This is data from sales one-pagers, solution briefs, win/loss reports, case studies, product information, playbooks, training, your CRM, and so on. This helps ensure that AI provides contextually relevant information, tailored to your organization’s needs.
- Analytics inside the platform: AI tools can analyze data from user interactions and content performance to generate new insights, inform decision-making, and identify areas for improvement.
- Large Language Model (LLM) general knowledge: LLMs train AI tools on vast pools of general knowledge, which is great for generating content and looking beyond the company-specific context. LLMs can be adapted and fine-tuned to determine what data they include, giving you a level of control over what your AI knows.
- Approved web searches: While you don’t want your AI tools pulling any old information from the internet, you can set certain parameters so your AI can search certain trusted websites. This gives your AI access to live online data that is credible, relevant and adheres to company policies.
Why you should manage your AI’s data sources
Despite — and because of — AI’s seemingly limitless potential and power, it’s important to set some boundaries. Managing your AI’s data sources, or Knowledge Scopes, isn’t just about making sure you’re working with accurate information. Every dataset introduces potential biases and assumptions that can easily lead you to make flawed predictions, irritate or misinform your customers, or land on the wrong side of the regulations.
By managing where your AI is pulling from:
You only share correct information with customers
Curated data makes sure that your AI delivers meaningful insights and accurate information. Without proper control, for example, your sales reps might quote outdated prices or promise features that no longer exist, damaging customer trust and possibly losing deals. Taking charge of your AI’s sources reduces those errors.
You stay compliant with global data regulations
Using AI in business comes with risks as well as opportunities. From data privacy and security to transparency and the potential for bias, you need to make sure your AI tools don’t let you run afoul of regulations such as the European Union’s GDPR and California’s CCPA.
When sales reps are handling sensitive or legally controlled data (such as product safety information or private customer data) they need to get the facts right. With controlled sources for their AI queries, you can make sure they’re always compliant with regulations and legal requirements.
💡 Bigtincan’s SecureGLP helps ensure that our GenieAI suite of tools guarantees security compliance against global standards, even as online compliance and risks evolve.
How AI platforms with customizable Knowledge Scopes provide better results
Whether you use AI to find up-to-date product information, answer customers’ queries, design training videos, or even roleplay sales scenarios, your go-to-market teams need a flexible platform that delivers relevant content from reliable sources.
Here are 3 ways that a customizable platform helps you manage your AI sources:
Tailoring content and access to information
Configurable Knowledge Scopes let you pick and choose the sources you want your AI tools to bring information from — whether that’s company- or team-specific documents, analytics, general LLMs, or trusted websites.
By managing data access, you can create specialized knowledge bases for various users or departments or use cases. For example, executives might need high-level analytics for strategy, while it’s more important for sales reps to have current product information and customer insights so they can answer questions on the fly.
Proprietary sources like customer databases and sales records can be integrated into AI datasets, while still keeping them secure. And by enabling access to approved online sources, you can enhance your AI tools with accurate, up-to-the-minute data.
Balancing general and specific knowledge
Any good salesperson should be familiar not only with their product and their industry but also with general sales principles and techniques and the wider market they operate in. Finding a balance between broad scope and relevant knowledge is key.
Sales reps should have a well-rounded AI knowledge base at their fingertips, but also one that suits their needs. An AI tool that can easily combine relevant company-specific data with relevant general information gives your sales teams the depth and breadth of knowledge they need for effective, personalized sales.
Eliminating the info dump
Users shouldn’t have to sift through information that’s not appropriate for their roles or their clients. By training your AI tools on what’s relevant, you can make them more context-aware.
Knowledge scopes in the Bigtincan platform allow for custom configurations that determine the exact combination of sources Genie gets its answers from. These scopes can include subsets of company content, Bigtincan Analytics, general knowledge from large language models (LLM), or web searches of specified approved sites.
This makes responses to AI search and assistant prompts and queries more targeted, focused, and relevant, which makes for more effective sales enablement and better customer service. And, by customizing the user settings and permissions, you can also protect more sensitive data.
How to configure Knowledge Scopes for better AI
Here are the key steps for managing your AI data sources:
1. Figure out where your company actually needs AI help
- Start by identifying the key areas where AI can add value. Ask:
- Do your sales teams need quick answers to customer questions?
- Do they need to practice and perfect their sales techniques?
- Do you want to generate training videos or sales scripts?
- For each function, identify the key internal and external data sources you’ll need for the most appropriate content.
2. Set up content criteria
- Establish the criteria you want to apply to your sources’ content. For example:
- The timeframe of the data you want to include (do you have a range of dates in mind?)
- Any business objectives or values you want target
- Sensitive details you want to redact or anonymize from internal data
- Trusted websites or searches
- Put in place any security or privacy measures needed.
- Map out and apply permissions for each source.
3. Train users and encourage adoption
- Educate your teams on using the AI platform.
- Highlight the benefits for their own roles and for their customers.
- Sensitize them to security issues and train them on using AI effectively and responsibly.
How to choose the right AI tool for your business: questions to ask yourself
Data sourcing
- What internal and external data sources do we need?
- How do we know the data is high-quality and reliable?
- Have we vetted third-party data providers?
- How often is data updated?
Data management
- How is our data stored, transferred, and disposed of?
- What are our data retention policies and responsibilities?
- Do we have a mitigation and action plan in the case of AI data breaches?
- How do clients request that we delete their data?
Privacy
- Do we have protections for sensitive customer and company data?
- Do we have compliance procedures for relevant laws and regulations (e.g., GDPR, CCPA)?
- Have we included opt-out procedures?
Transparency
- Have we documented our AI decision-making processes?
- What performance metrics and evaluation processes are needed?
- Do we have accurate audit trails?
Ethical considerations
- Have we documented our ethical principles and guidelines around AI use?
- How do we minimize and mitigate bias in our AI models?
- How and how often should we conduct ethical audits of our AI systems?
Gain more control over your AI sources
To get the best out of your AI tools, you’ve got to put the right stuff in. With a customizable AI platform like Bigtincan’s GenieAI suite, you can configure your data sources to meet the various needs of your sales teams and customers.
Whether you just need internal company details or you want to draw on the whole worldwide web of information, you can control where your AI tools look for answers. And with customized permissions, you can ensure the right people can access the data they need.
