At a time when B2B marketers are constantly facing tighter budgets and fiercer competition, new developments in generative AI can sound too good to be true.
Faster content creation? Access to actionable data? More accurate segmentation and lead scoring? It’s no wonder that 61% of marketers are already using AI in their operations, according to Influencer Marketing Hub’s 2023 report.
But some leaders — especially those on IT and legal teams — aren’t as ready to go all-in on AI. They have real concerns about its risks, including the accuracy of AI-generated content, privacy concerns, and the potential for bias and discrimination. And marketers need to be prepared to address these objections head-on.
Why AI is worth the risk for B2B lead generation
Before we dig into the risks, let’s recap why generative AI is quickly becoming indispensable for B2B marketers. Here’s why a strategic approach to AI could be well worth the potential risks:
AI enhances your B2B marketing data
Manually collecting and refining B2B marketing data — including demographic, firmographic, and intent data — has been a labor-intensive and expensive endeavor for marketing teams for years. But modern AI tools can streamline and automate these processes. This makes it much easier for marketing teams to more precisely segment and target their audience with relevant campaign messages.
AI-powered models also enable predictive analytics for B2B marketing, which uses your historic data to develop predictions about how prospects will behave as they move throughout their buying journey. This allows you and your team to focus on leads and accounts that are most likely to convert — and improve campaign performance in real-time by predicting future outcomes based on very early results.
AI supercharges your lead scoring
Traditionally, lead scoring has attempted to quantify a lead’s readiness for a sales handoff, often based on subjective opinions and assumptions about which behaviors indicate sales readiness. With AI-powered lead scoring, your approach becomes truly data-driven. Machine learning models analyze behaviors preceding conversions, and identify patterns that can be used for dynamic, real-time lead scoring. In this context, AI models evaluate the cumulative actions of a lead instead of treating each one — like a content download or a webpage visit — as an isolated behavior.
AI facilitates fast, personalized content creation
New generative AI models are capable of producing high-quality content, both written and visual, in a matter of seconds. While AI-produced content still needs human oversight, as we’ll discuss shortly, this ability to ideate, prototype, draft, or edit existing content at scale is a huge game-changer — especially for marketing organizations with smaller teams or smaller budgets.
The same tools can be used to take campaign assets and personalize them for different audience segments in moments. For example, you can give a tool like ChatGPT an ebook you’ve created for a C-level audience, and ask it to adapt the tone and copy to speak directly to individual contributors or mid-level managers instead. Or, you can harness the core messaging of a nurture sequence and task AI to craft variations aimed at sectors like finance, healthcare, or e-commerce.
While these benefits only scratch the surface of how AI can impact B2B lead generation, they serve as a foundation when weighing the potential pitfalls of AI. As you journey toward AI adoption, these points will be important for discussions with your stakeholders.
Common AI concerns in B2B marketing — and how to overcome them
Now, let’s dive into how marketing leaders can proactively acknowledge these concerns. We’ll leave the existential worries to sci-fi movies, and focus on the most pressing objections around AI adoption in B2B marketing — what happens, how it occurs, and how marketing teams can avoid negative impacts.
Inaccurate content and AI hallucinations
Large language models (LLMs) — like ChatGPT’s GPT-4 or Meta’s LLaMa — generate text that plausibly appears like a human could have written it. These models have been trained on enormous quantities of data to understand the meaning and nuance of language. And they function, at the simplest level, by predicting the next likely word in a sentence. As the Financial Times describes them, “LLMs are not search engines looking up facts; they are pattern-spotting engines that guess the next best option in a sequence.”
That means that LLMs can, occasionally, predict the wrong words and generate completely inaccurate information. This happens because LLMs are designed to deliver responses and answer prompts, even if they don’t have the precise context or reference points needed to fulfill requests with total accuracy. Like an eager-to-please intern reluctant to reveal their lack of expertise, LLMs will predict their way through a prompt — and give no indication they don’t actually know what they’re talking about.
These are known as “AI hallucinations.” Some documented cases include:
- A New York law firm was fined for a ChatGPT-assisted court filing containing fabricated legal cases and citations.
- Men’s Journal faced scrutiny over an AI-generated article on low testosterone that included factual errors and poorly supported claims about men’s health.
- Google faced ridicule (and lost $100B in market value) after its launch materials for Bard, its AI chatbot, included an animation of the LLM confidently sharing an inaccurate statement about the James Webb telescope.
For marketers, this means AI-generated content could include factual errors or made-up information — not ideal for building trust in your brand and credibility with your audience.
How to address AI hallucinations
There’s a simple way to overcome this AI risk: human oversight. Any content that you use ChatGPT or a similar generative AI tool to create should be fact-checked and confirmed by a knowledgeable team member. Plain and simple, this step needs to be proactively included in your AI workflows and content production processes.
While there are many LLMs out there for marketing teams to use, ChatGPT is certainly one of the most popular and widely adopted tools so far. And it has one clear limitation: it was trained on publicly available information on the internet, up until January 2022.
That means ChatGPT can’t tell you who won the Grammy for Best New Artist in 2023, what the biggest trends in marketing were in 2022, or which search terms are the most popular in your industry right now.
For marketing teams, this can be problematic if you try to use the tool for time-sensitive tasks like SEO research. And if you rely on ChatGPT for lead generation strategies or brainstorming, you could be missing out on the latest developments and context.
How to address outdated information
First, determine the training cut-off for your generative AI tools, whether you use a direct solution (like ChatGPT or Bard) or a product with an integrated third-party LLM (like most AI writing platforms).
Then, determine which use cases make sense for gen AI and when you’ll need human intervention. For example, if you’re writing a history of your industry, AI can help you build a timeline – up until the model’s training cut-off date. In those cases, you’ll need to complete your research the old-fashioned way.
Data privacy risks
Most LLMs continually train and improve based on interactions with users. That means, theoretically, that any information you share with ChatGPT or a similar tool could become part of its training database — and be surfaced to other users in response to their queries.
For many businesses, this poses a security and privacy risk. Financial services companies could have a compliance nightmare on their hands if their customers’ personally identifiable information ended up in an LLM training database. Healthcare watchdogs are concerned AI may weaken HIPAA-compliant data protection practices — like the deidentification of personal information — through its ability to detect patterns in datasets stripped of listed identifiers. These concerns have prompted several leading financial, tech, and telecommunications companies (including Goldman Sachs, Apple, and Comcast) to ban the internal use of ChatGPT and similar tools.
For marketing teams across industries, compliance with privacy regulations like GDPR and CCPA may be jeopardized if customers’ personal data is shared with AI tools.
How to address data privacy and AI
Marketing teams should be aware of the privacy policies of any AI tools or platforms that include AI integrations. Especially if user data will be included in its future training sets, everyone using those tools should know not to share any information that shouldn’t be available to the public. (In other words, if you wouldn’t post it on social media, don’t enter it into ChatGPT.)
Consult with your legal team about your organization’s specific concerns and codify these guidelines into your AI workflows and policies.
Bias and discrimination
Finally, AI tools are trained on massive datasets, typically scraped from existing content across the internet. Anyone who has spent time on the internet knows that content includes a lot of biased and troublesome information.
That bias naturally seeps into the output of AI. For example, an Asian MIT graduate asked an AI image generator to transform her original photo into a professional LinkedIn photo. The tool responded by giving her lighter skin and blue eyes. And researchers found that certain ad campaigns promoting STEM job opportunities were designed to be gender-neutral, but ended up reaching far more men — because the bidding algorithm optimized cost-effectiveness, and young women were more expensive to target.
Lead generation teams can see bias come into play when using AI tools to segment audiences or score leads — especially if you serve a global market, but your AI training data is mostly based on North American or European customers. And LLM-generated content might include cultural insensitivities or lack examples and perspectives that resonate with diverse audiences.
How to address bias and discrimination
Once again, human oversight is key, especially when it comes to AI-generated content. Editors should review all output for potentially offensive or non-inclusive language.
Additionally, when using AI to support activities like lead scoring or audience segmentation, you’ll want to make sure the data your models are trained on is truly representative of your broadest customer base. Otherwise, if AI tools are only fed narrow results (like information about a specific user group), you could end up in a feedback loop that reinforces and amplifies existing biases in your data.
Don’t miss out on the benefits of AI-driven strategies
These concerns around using AI for marketing may be real, but they don’t need to be deal breakers. AI-driven strategies can supercharge B2B team performance by improving data accuracy, powering predictive analytics, enhancing lead scoring, and speeding up content creation.
For most organizations, those benefits easily justify the added legwork of developing clear AI usage policies and conducting internal training. Marketing leaders should be prepared to meet with their legal and IT colleagues to make the case for AI and take a proactive approach to combating objections. The potential payoff is well worth it.
Get the latest content from DemandScience in your inbox.
In B2B marketing, accurate data is essential for success. However, data quality can decline over time due to…Read More
Account-Based Marketing (ABM) is a strategic method where marketers communicate with a defined set of accounts as markets…Read More
Traditional sales and marketing strategies have long focused on lead quantity by casting a wide net to capture…Read More