The drumbeat from thought leaders and executives has been steady and insistent for the last few years: AI. AI. AI. More AI in everything. But as the mandates to build AI products keep coming down from the C-suite, sales teams are running into two show-stopping problems, according to research from Upwork:
- Consumers are wary of AI interactions.
- Most workers have no idea how to use AI to make them more efficient and successful.
On the one hand, Upwork shares that 96% of C-suite executives expect AI to improve productivity, and 37% think their teams are skilled and comfortable with it, with the same number expecting AI to increase worker output. Meanwhile, only 17% of employees feel skilled and confident in their AI use, and 77% say these tools have actually made them less productive.
Choosing the right AI tools to help facilitate the buyer’s journey can be daunting for a busy sales leader. After all, the amount of AI tools on the market has exploded in the past few years alone. A good first step that will immediately supercharge your sales without breaking the budget (or your patience) is an AI sales chatbot trained on your internal data.
Why Chatbots and Why Internal Data?
According to research from 6sense, most B2B buyers are 70% of the way through their decision-making process by the time you hear from them — if you hear from them at all.
In fact, Gartner found that 75% of B2B buyers want a completely rep-free sales experience. More and more buyers are making decisions based on independent research and opting for a self-serve experience any chance they get. They want to get to the information they need with no gatekeeping and as quickly as possible, and they hate having their time wasted.
This isn’t unique to B2B buying either. Users are already turning to AIs to replace traditional search, with Gartner predicting search engine traffic drops of 25% by 2026. GWI found that >30% of Millennials and Gen Zs already use generative AIs for most of their search needs. There’s no reason to suspect things are any different when it comes to making buying decisions.
AI-powered chatbots have the ability to give users more specificity and relevance than standard content; and more context than spec sheets and user manuals. In a real way, they combine the best of these tools into one, presenting accurate, relevant information on the buyer’s terms and with the context necessary to apply it to their situation, all without buyers having to piece it together themselves. Or at least they can.
The problem is that your average AI chatbot has absolutely no idea what makes your particular company or product great. They also have absolutely no preference for your company over your competitor’s. General-purpose
The best AI chatbots are trained on your product (specifically your data), just like your best SDRs and AES.
The good news? There are tons of ready-built solutions that let you take an existing tool, train it on your product and company data, and let it loose to start closing business.
Hiring Your First Salesbot
Before you pull out the company card and sign up with the first chatbot provider, you need to think about what kind of AI you want. If it helps, think of it like making a sales or customer service hire: You wouldn’t pull a random person off the street and offer them a salary without having a job description and going through some interviews first, right?
“Hiring” a sales AI works the same way. The first step is coming up with a job description: What do you want the AI to handle, how do you want it to behave, what level of “seniority” should it have (or, in this case, how in-depth do you want it to be?) Some questions to consider:
- What “position” is the chatbot being hired for?
Are they going to be similar to a BDR who greets prospects, takes information and questions, and passes it all to a salesperson to follow up on? Or will they be able to actively engage with, answer questions from, and generally guide prospects through the buyer journey?
- How much information will the newly hired chatbot have access to?
Given their role, what will the chatbot know and what will they need human help with? Will they be privy to technical specifications and background material, or will they only be trusted with marketing content?
- How will you oversee the chatbot?
What kinds of guardrails will the chatbot work under? What will it do if it runs into trouble? Who will it escalate problems to? Who does the chatbot report to?
- How much training will you need to do to get your new hire up to speed?
Just to make it clear: No matter which solution you pick, you will need to do a lot of training to get your chatbot operational. But whether that training consists of uploading existing documents or coding new machine-learning models is going to depend on which solution you pick: Some are much easier to train than others.
As with any JD, it sounds simple, but as anyone who has ever hired anyone knows “the devil is in the details.” Figuring out a plan can be difficult, but doing it in advance will save time, money, and headaches later in the process.
Meet the Candidates
You’ve got your JD, now it’s time to see who’s applied and do some interviews: sometimes literally! There are many solutions available on the market, each with pros and cons. Lets meet the candidates:
NLP-Powered/AI Search:
NLP stands for “natural language processing,” and is the simplest form of proto-AI available. In short, it’s just a better kind of search that uses AI to figure out what the searcher was asking.
Pros: Relatively simple to build — often can just replace your existing site search feature with a plugin.
Cons: Not really a chatbot, has minimal personalization, and everything that’s searched has to be exposed to the broader internet (more or less). Some can get rather pricey or difficult to fine-tune if working with large, complex databases.
Examples:
NLP-Interpreted Legacy Chatbot
Like a legacy chatbot, you still have to define all the possible ways a conversation can go by specifying acceptable questions and their answers. Unlike legacy chatbot, a coat of AI allows them to sound natural and better match buyer questions to workflow answers.
Pros: Closer to “true” AI. Can be made fairly personal with a complex ruleset. Doesn’t require public access to all data, and can be easily controlled to only answer specific questions and provide specific information.
Cons: Still relatively simplistic, though it can be made to sound like a natural conversation. Relies on extensive labeling of information and manually defined rules. Easy to stump if it encounters something you didn’t plan for.
Examples:
- Dialogflow
- Rasa
- Tidio
- Drift (now SalesLoft)
- Intercom
Document-Driven AI Chatbot
The most well-rounded applicant, this is a true AI chatbot that doesn’t require prebuilt workflows, and can interpret questions and provide answers based on the data you feed it, but without any programming or “true” training.
Pro: A middle solution between a fully custom AI model and a more simple chatbot. Usually competent at finding the requested information quickly and surfacing it for users through natural conversations. Somewhat easy to control output with preset and customizable rules. Easy to scale, as the model can adapt to new documents as soon as they’re added without needing new rules.
Cons: Can have weird behaviors if not tested thoroughly. Data must be well-sanitized to avoid surfacing anything too proprietary. Can struggle with complex, multi-step processes that require combining multiple documents or layering information, like troubleshooting. Limited personalization options.
Examples:
Fully Trained Custom Chat AI
For most companies, this is complete overkill, but for some working in highly regulated fields like healthcare, financial services, or defense, this solution may be the only option that passes regulatory hurdles and is compliant with the law.
Pros: Fully customizable. Can have individual models tailored toward specific buyers or buyer types that have access to different data or different approaches. Broadest capabilities with regard to data types. A true end-to-end AI experience.
Cons: Expensive, time-consuming, and difficult to build, let alone get right. Can be unpredictable if not trained well and tested extensively, though it also offers the most options for embedded guardrails like topic restrictions and conversational pathways.
Examples:
Onboarding Your New AI Salesbot Hire
You’ve made the leap! After countless interviews, reference checks, and technical hurdles, you’ve signed the paperwork on your new AI sales assistant. Now what?
The journey to a thousand AI-powered conversations with buyers begins with a single document — or rather a single large pile of every possible kind of document you can find. Whether you’re using an off-the-shelf solution or building your own, one thing all AI tools have in common is they need lots and lots of data.
Since the goal for this implementation is to build a chat assistant that can replace a sales rep in answering buyer questions, you’ll want to feed it all the data you can gather about your products, how customers might use them, and how you stand out from competitors. Some suggestions include:
- Product spec sheets and detailed technical manuals
- Public knowledge base
- Internal knowledge base, troubleshooting walkthroughs, and support/success documentation
- Blog posts and similar resources about your product (But first go through them to make sure they’re still up to date and relevant!)
- Sales scripts (not just from reps, but also sales engineers, sales enablement, customer success, and account managers)
- Transcripts of sales conversations with buyers
- User feedback and survey (anonymized!)
- Competitor research
- Customer personas and buyer journeys (optional: good to have if they’re well-developed and accurate, but can hurt if they aren’t)
- Compliance and legal documents to provide guardrails on how the AI handles itself and requests
The more data you provide, the more comprehensive and useful the assistant will become. However, you’ll need to make sure the data provided is tagged, categorized, and contextualized for the best results — more data means more upfront work to build robust metadata.
You’ll also need a formal style guide to ensure that voice, tone, and style stays consistent and on brand. The last thing you want is for your chatbot to suddenly begin saying things your reps would never say or going off message. Having a formal style guide adds another set of guardrails to make sure your tool does more good than harm.
Finally, you’ll need time. AI is a new, complex technology. It’s not the kind of thing you set up in an afternoon and call it a day. Even simple implementations using common platforms can take weeks of building, testing, and refining to get right. But if you can devote the time and resources, the ROI makes it well worth the investment.
Embracing AI
Demands to do more with AI are unlikely to stop anytime soon. While the popularity and excitement around this technology waxes and wanes, the trend is very clear: Executives, investors, and even buyers (if reluctantly) will expect more AI and more from AI in the future.
The challenge for sales teams now is to upskill as quickly as possible, understand the landscape, and be ready with proposals for workable solutions when the call comes from the corner office. Having a plan to roll out a virtual sales rep in your back pocket can make a revenue leader look like a superhero with future-sight, and in an increasingly challenging business environment, that’s the kind of street cred no sales team can afford to pass up.
But beyond looking good for management, AI really is the future. Customers increasingly want to figure things out on their own, with little-to-no interaction with sales reps. Giving them the option to get what they want right now in a way that makes them comfortable and perfectly addresses their needs is a no-brainer. It’s the Sales 4.0 solution, and it will increasingly be the solution that brings in new business.
Looking to improve your buyer enablement and scale with AI? Check out Claraty, your ultimate demo assistant.
Learn More
