AI has reached a tipping point in accessibility and proficiency, and Sales and Marketing teams are suddenly engaging with AI in new ways—many for the first time.
They’re using ChatGPT, for instance, to write their B2B prospecting emails, personalization and all. It takes a human five minutes or thirty or an hour to do some prospect research on LinkedIn before they can even get started. AI? It can do it all almost instantaneously. Or can it?
Not all AI tools are created equal. They’re not all trained to do the same thing. A chatbot like ChatGPT might write a heck of an email, but the accuracy of that email, especially when it comes to personalization, may not be as dialed in. As fascinating (and remarkable) as ChatGPT is, it also can’t build you an accurate target account list.
There are, however, AI tools that Sales, Marketing and RevOps teams can (and should) use to run faster and generate quality revenue. The key is to make certain you’re using the right tools to produce accurate and powerful results.
With the lessons learned from AI chatbots like ChatGPT and Google Bard, here is our Rev primer for revenue teams that require up-to-date and spot-on results from the leading edge of AI.
AI sources what is available—and it needs quality data
AI can improve lead generation efficiency and speed. So revenue teams absolutely can implement it—just not indiscriminately.
Not all AI is up to the task for what revenue teams need. It can get the details wrong. It did so, infamously and publicly, when Google Bard provided incorrect information about the James Webb Space Telescope.
It’s true that generative AI’s ability to create smooth, coherent, plausible text is impressive. Afterall, that’s what it was trained to do: generate realistic-sounding answers, not necessarily accurate ones. The details can be off. In Sales and Marketing, those details matter.
Teaching a robot to love
Here’s an illustrative thought experiment: Pretend that (like a shocking amount of people) you tasked an AI chatbot with writing a love poem for your partner. Could you pass it off as your own?
If you and your partner just met, maybe you could. But if you know each other well, you can’t get away with cold-calling AI to write a love poem. Straight-up, it won’t sound like you.
But it might, if you adjust the parameters: Give the AI chatbot all the love poems you’ve written over the years, then ask it to write a poem in that style. You just might pull it off, because the AI has greater access to relevant data.
Data quantity matters too
This is part of why Google Bard dropped the ball on the James Webb Space Telescope data: it’s a very recent news phenomenon, so there simply isn’t much information for the AI chatbot to source its knowledge from. It wrote an answer without sufficient context. If the question had been about the Hubble, Bard’s odds of nailing the answer would have improved dramatically. But, again, its model is designed to pick the most likely words and phrases, not what’s true.
To move the needle on your GTM functions, your AI-driven systems need to be used for what they were trained to do and have access to enough data. Otherwise they will point you in the wrong direction, or just come off sounding… well, off.
Revenue teams require up-to-date data
More data doesn’t just give your AI a better knowledge base to draw from; it improves the AI’s performance—more so than more processing elements. It’s a lot like how a bigger brain doesn’t make a person smarter so much as more experience does.
Historical bias causes revenue issues
Yet AI runs into the historical bias problem referenced above with the telescopes: history often overwhelms recent info in AI processing. This is a problem for revenue teams that rely on immediate, relevant data to make decisions both accurately and fast. The lack of data created about today or yesterday can’t stop you from taking action now.
That’s one thing that limits generative AI models like ChatGPT. Put to work for a Sales team, it would miss some timely events and milestones about your target accounts to draft accurate email copy. And if you asked ChatGPT to build you a target account list? Well, it might do a reasonable job of finding a few good targets given enough context about what you’re selling. But generating an accurate list of hundreds or thousands of companies is a fundamentally different task—and absolutely requires specific, up-to-date information. Data that’s even a few months behinds will leave you in the middle of another James Webb Space Telescope situation.
LLMs improve accuracy
So, if you’re looking to AI to build your target account list you need to look beyond the hype of generative AI and start looking at AI that’s using large language models (LLM) in ways that leverage up-to-date information that’s most relevant to B2B targeting. That’s how you’ll get outputs that have a high level of accuracy in the details.
LLMs are the deep-learning algorithms that identify relevant data and synthesize it into useful form. Despite the name, they’re not just used for language processing applications like chatbots—they’re used in many other cases, like building aiCPs, or AI-driven customer profiles, that help revenue teams identify the exegraphics behind their best customers and find other accounts that share those traits and fit their ideal customer profile.
Short version? AI that uses an LLM can make sense of data scraped from the web, including up-to-date information about what’s happening with millions of companies, then analyze it to understand how those companies are executing their mission. It can compare this to how your current best customers are running their business, in order to provide real-time insights into the accounts that you should be targeting.
It’s a sorting problem for revenue teams
The headline-grabbing generative AI used by chatbots relies on, essentially, solving search problems. They are a natural extension of what Google does today, where you ask a question and expect even just one good result.
In B2B, targeting is a different search problem: you want AI not to find you just one result, but to find all the results and then stack rank them in terms of how good a fit they are for what you’re trying to sell. Oh, without missing any, and without diluting the results with poor-quality targets.
In short, revenue teams require AI capable of solving sorting problems in addition to search problems.
To solve those problems accurately requires the right data, normalized in a way that an automated process can digest it. And it has to be up to date. These are significant challenges for LLMs. And LLMs are neither cheap nor easy to create; a company is not going to rebuild LLMs every day to account for new data.
This is why not all AI is created equal, and revenue teams in particular have to be selective in what will create not only fast content, but the right content. The best-fitting AI tools can identify the right information and assess relevant context around the organizations you should be selling to.
Final thoughts: Presentation matters
Let’s face it: the accuracy of AI chatbots is kind of a novelty. It’s fun to see what responses we get, but we’re all too used to finding the right answers on Google to be wowed by, essentially, an impressive search function.
We’re much more moved by the ability of AI to respond like a human in real time—faster than any person could—and sometimes even better. The ability to synthesize information, to comb those billions of data sources and come up with answers that come across more eloquently than any of us ever sounded in a job interview (or a prospecting email), is why revenue teams are so intrigued by technology like ChatGPT.
Presentation matters. No one would let AI write prospecting emails if it talked like a robot. But in the end, substance matters more than presentation. AI saves revenue teams time and resources, no doubt. But relying on the best-fit AI, AI that produces accurate results on B2B prospects, will be the real differentiator for revenue teams in this new landscape.