It’s easy to think of artificial intelligence (AI) as something far off in the future. However, it already operates virtually everywhere in your life and is quite likely to be powering certain functions within your sales technology ecosystem.
This article provides a primer on vital AI technologies that businesses are leveraging to boost productivity, forecast more accurately, deliver superior customer experiences, and drive sales performance improvements.
We review existing use cases and shed light on potential applications for AI in sales in the future, and make suggestions on what sales leaders should be doing right now to prepare for an AI-powered future.
There are two fundamental ways to think about AI. The first is the computer-that-thinks-just-like-a-human (often turning evil) trope that film and TV producers have so expertly exploited over the past 65 years since John McCarthy, the professor of computer science at Stanford who coined the term Artificial Intelligence.
That might be the original story for AI, but it is far from the current reality, which brings us to the second, more contemporary way of seeing AI:
The deployment of one or more AI technologies in precise ways to help humans be more productive in their jobs, make fact-based decisions, and rapidly spot, predict, and act upon trends in ways that deliver high value to their enterprises.
To facilitate a discussion of how this is already in use (and will continue to evolve) in sales functions and within sales-tech ecosystems, we begin with a quick primer on five relevant AI technologies with explanations on how each works and their relevance for AI in sales.
Machine learning (ML) applications build mathematical models (algorithms) using sample or training data. Training teaches the model to make accurate predictions, decisions, or recommendations. The more a model runs — the more data one feeds into it — the more it learns, increasing its knowledge and accuracy over time.
In the early days of machine learning, training a new model or algorithm required enormous data sets plus heavy human intervention in terms of “labeling” or describing the training data. ML training then progressed to “semi-supervised,” in which ML models could learn with only partial human intervention. The next major iteration for machine learning (which started around 2006) is called neural networks or deep learning. With neural networks, the AI model learns with virtually zero human involvement.
Good examples of neural-network style ML include handwriting, speech, and face recognition applications.
In sales settings, leading providers of sales-enablement technologies are already using the technology to decipher patterns and draw useful conclusions from massive amounts of data collected through behavioral tracking of consumers — message receipts and opens, clicks, search histories, keywords, navigation pathways traveled from one web site to another, social media interactions, and so much more.
A virtual assistant is a software agent that performs tasks for a person based on commands or questions. Siri, Alexa, and Cortana are all AI-powered virtual assistants. Meanwhile, a chatbot is a software application used to conduct conversations in natural language via text or text-to-speech through messaging applications, email, websites, mobile apps, or telephone.
Virtually anyone who has interacted with customer support or a help desk in recent times has interacted with a chatbot. In practice, a chatbot will often handle the early part of a support interaction — gathering and validating account information, executing security protocols, defining problems and issues — and then hand it over to a human.
Natural languages such as Latin, Greek, Sanskrit, and English evolve spontaneously among humans versus being formally written (such as a computer programming language). Thus, NLP figures in speech recognition, language translation, extracting and structuring relevant data from free-text fields and written documents.
In a sales setting, NLP could be used, for example, to automatically digest, process, and route customer queries received via web forms.
So, where standard form fields — name, company, email address, phone number, and so forth would map 1:1 with standard contact data fields in a customer relationship management (CRM) system — NLP can be layered in to extract information from free-text fields where new prospects describe specific needs in their own words. That can assist with routing leads to sales reps with the relevant domain knowledge, or with deducing new prospects’ locations for appropriate territory assignment.
It is important to note that while we may be distinguishing AI technology subsets such as ML and NLP, the progression to deep machine learning (neural networks) has simultaneously enabled considerable advances in NLP (meaning, machine learning also helps to inform the NLP engine, making it smarter and more accurate over time). This progression has, in turn, opened new vistas for how and where business enterprises can deploy NLP to achieve enormous productivity gains in often tedious, non-value-adding business processes.
As NLP technology continues to advance, any instance of humans reading written content (or hearing spoken content), extracting critical data from that content, and then entering that data into relevant information systems becomes a candidate for NLP-powered automation. NLP not only makes such processes faster but also (at least theoretically) more consistent, complete, and error-free.
In sales, chatbots can be used to collect basic profile information from new contacts before handing it over to appropriate reps. A few advanced CRM systems already offer AI-powered chatbots capable of self-learning.
Augmented analytics uses statistical and linguistic technologies to improve data management performance, from analysis to sharing and creating insightful business intelligence. The technology enables humans to work with vast and complex data sets, which will only continue expanding at exponential rates in the future.
Intelligent applications, meantime, inform and facilitate decision making. Examples include the recommendation engines found in Amazon and Netflix. Ideally, smart applications improve business performance outcomes by orders of magnitude compared to what humans can accomplish independently.
In sales, an AI assistant might monitor customer interactions, behaviors, word choices, and so forth, looking for patterns that typically signify advancement toward a deal close. The agent/rep can then recommend the next-best actions that are, based on real statistical analysis and probability, the most likely actions to result in a deal close.
AI visualization uses algorithms to explore relationships among entries in a graph database, generating images (visualizations) of data that make it easy for humans to understand and respond to critical (and otherwise difficult-to-spot) trends within data.
Visualizations showing how information (or disinformation) propagates through social media networks is one good example of this technology.
In B2B commerce, good examples include visualizations for relationships among companies in a supply network or potential relationships among customers within a CRM database.
It would be a mistake, for example, to quote two different prices to two members of the same family or business enterprise; a visualization app that detects such relationships could prevent this from happening.
Before discussing AI use cases in sales, it is essential to note that not everything currently being sold as AI qualifies as such. For example, plenty of applications and functions automate manual processes but do not use AI technologies (machine learning, NLP, and so forth) to do it.
For example, one might have a tool that automates the scheduling of appointments. A predefined set of rules automatically launches an email or SMS message that prompts a customer to set up a call or meeting. That type of tool qualifies as robotic process automation (RPA) but not AI, as it does not involve any kind of algorithmic learning or human-mimicking interaction with a customer. Thus, it saves time but does not introduce any new intelligence to the exchange.
An AI-powered scheduling app, by contrast, might constantly monitor a customer's brand interactions across multiple channels (website, email, phone, support tickets, and so forth). It might then determine algorithmically that the customer is getting ready to make a new or repeat purchase and then either launch an automated outreach to the customer or prompt a sales rep to do it. If it’s the latter, the bot might also provide the representative with data-driven recommendations for when and how to best contact the customer, which products they might be interested in buying, opportunities for cross- and upselling, and so forth.
Being able to distinguish between RPA and true AI will be necessary for sales leaders, moving forward. Unfortunately, they are already encountering a barrage of conversations that do not always differentiate between real and pseudo-AI.
In considering potential use cases for AI in sales, it is helpful to think of four buckets (the four Ps): productivity, prediction, personalization, and performance.
AI is most likely to be conflated with robotic process automation (RPA) when considering the potential for productivity gains. However, as demonstrated in the scheduling example above, plenty of applications can automate and accelerate business-process execution, thereby increasing productivity, without introducing new value or intelligence.
AI comes into play productivity-wise when it can:
Independently gather and process information
Present insights or recommendations that humans could not generate on their own (because the data sets are too vast)
Generate usable insights much faster than any human
Let us say, for example, that your marketing team runs a campaign that generates a set of "thin" leads, containing only first and last names and email addresses. Assigning a human to research and enrich the lead data would be hugely time-consuming and expensive. An AI-powered engine, by contrast, could go out to the Internet, search far and wide for publicly available information associated with the names and email addresses collected in the campaign. Using NLP, the agent could scrape relevant data, normalize it, and automatically enrich each new lead profile. Even where it cannot get to 100% enrichment, it would still be delivering an enormous gain in productivity.
The bot might also generate a visualization to show how various new leads are connected (by company, job role, and so forth). That would be authentic AI as the application would perform a human's work while also adding value and intelligence by conducting its search independently, more rapidly, and more thoroughly than any human could do efficiently or effectively.
With an estimated of 64.8% typical sales rep activity being non-revenue generating, there are plenty of other examples of how one might use AI in sales to create enormous gains in productivity. Ultimately, AI in sales enables businesses to run either smaller or similarly sized teams that focus more on consultative selling and other revenue-generating work.
Other essential sales tasks and processes that stand to benefit from AI-powered productivity gains include:
AI-powered chatbots could also qualify or enrich lead profiles in CRM, interacting with customers in their preferred channels and adding information to their profiles, then handing over in-process conversations seamlessly to knowledgeable sales reps.
In addition to identifying contacts who are most likely to convert to sales, a CRM-based AI bot with visibility into detailed customer interactions and purchase histories can recommend specific actions for sales reps to take, plus optimal timings for those actions.
Say, for example, your company is launching a new product that is a significant upgrade to the existing one. Sales leaders might plan a big inside sales push to reach out to all current product owners with a pitch for the upgraded version. As no one has interacted with the new product yet, there would be little data to gauge buying intent, but there may still be sufficient data to prioritize the massive call effort. For instance, a machine learning CRM module monitoring all records of customer-brand interactions might notice that contacts who frequently ask for new or improved product features in tech support forums are among the most amenable to conversion in sales-upgrade conversations.
Where sales cycles involve completing complex customer requests for quotations or proposals (RFx), AI in sales could contribute to productivity gains by digesting and auto-completing large portions of RFx documents and repurposing responses and other inputs from past, successful RFP completions.
Forecasting is another area where AI in sales can bring game-changing results and genuine competitive advantage.
A robust CRM system with embedded artificial intelligence can generate predictions around customer behavior by deeply analyzing and learning from the aggregated data of similar past interactions with all clients. Thus, AI in sales can suggest which products to pitch, at what times, and through which channels to increase the likelihood of closing sales. Of course, the CRM would need to capture, structure effectively, and store detailed information around all customer brand interactions; that is something that only a few CRMs do well at this time.
AI might find trends in marketing engagement (opens, clicks, social media buzz) and in related sales activities (new queries, repeat calls, repeat website visits, and so forth) to spot imminent demand surges. By detecting and feeding those insights to manufacturing, procurement, supply chain, and logistics operations, the enterprise can ensure that no sales opportunities are missed due to constrained supply. Better demand forecasting virtually always leads to happier customers and sales teams, as products are more likely to be available when wanted.
Or leaders could conceivably use AI in sales to adjust for forecasting bias with specific sales reps and managers. Imagine, for example, that Kevin P is almost always too optimistic in grading deal-closing potential, while Mary G is consistently pessimistic. A well-trained ML algorithm could spot and adjust for this. That would improve job experiences for the reps, reveal opportunities for management coaching, and make a sales group's overall revenue forecasts more consistent and accurate over time (which would typically result in more precise earnings expectations and happier sales personnel).
Better prediction, in turn, enables greater personalization, which, when used judiciously, can dramatically improve customer experiences.
As with process automation, plenty of personalization happens already that is not based on AI. Populating an email with a person's name, for example, is not powered by AI.
Where AI comes into personalization would be predicting, based on one's profile or how they have acted in the past, how they are most likely to behave in new situations, and recommending the most likely actions that will convince them to buy in each unique case. The customer may never be aware that this kind of personalization is happening.
An AI module in CRM might, for example, determine that Jane Doe never answers her phone between 9 am-5 pm, picks up occasionally in the 7 pm-9 pm window, and never between 9 pm-7 am. Or it might determine that she never answers her phone but does read text messages typically in the morning, suggesting morning SMS as the best means of reaching out and winning engagement with Jane.
Another example might analyze the search terms various people type in at a business website. For example, people already familiar with a particular product line might use more specific terms than one who is entirely unfamiliar. Imagine having an AI bot that could identify who is who, recommending one script/interaction for the newbie prospect and a more sophisticated interaction for the more knowledgeable searcher.
The possibilities for micro-segmentation and personalization in sales interactions are practically endless. For example, is there a contact who never buys without some type of discount? AI could flag this for a rep, who could either increase their opening price offer as a hedge or proactively offer a small discount at an optimal time to close the deal.
Up- and cross-selling are other forms of personalization that AI can support. Place something in your cart on Amazon, and you will immediately see other products people have purchased along with the product you have just saved. The same technology that drives such an engine could also inform the sales team of cross- and upselling opportunities. What products do customers typically bundle? What products do they usually come back to buy within a short time of purchasing an original product?
The ultimate for AI in sales is to more rapidly and consistently close more, high-profitability deals.
AI in sales can predict which prospects to focus on and can also constantly realign the sales pipeline, helping a sales team concentrate its energies on deals with the highest ROI potential.
If you have one potential deal valued at $10,000 with a 50% chance of closing and ten possible contracts valued at $1,000 with a 75% probability of closing, AI in sales would advise the team to prioritize the ten seemingly low-value deals versus the one more significant value deal.
In the same vein, AI in sales might also help one understand which customers need to be prioritized. Again, the key here is having a CRM capable of looking across the entire customer lifecycle. For example, customers who are heavy users of tech support, who have high rates of product returns or discount requests might be automatically deprioritized in a sales queue because their post-sales behavior erodes any profitability associated with their account.
Likewise, a machine learning model might determine customers' sensitivities to price changes or propensities to negotiate for discounts. It would be beneficial, for example, to understand which size discounts typically satisfy which customers; a rep could then be sure to offer no more (and no less), pleasing the customer while also moving more quickly to a profitable deal close.
Conversation intelligence — that is, real-time monitoring and transcription of sales discussions — is yet another fast-expanding and exciting field of AI-powered enablement for sales that has big potential for driving sales performance improvement. Using a combination of NLP with machine learning, such applications track conversations in real-time, analyzing word choices, emotions, tones, and other parameters that signify a prospect’s sentiments or buying intentions as they shift and evolve naturally in conversation. Now, imagine conducting a sales call with a steady stream of real-time sentiment analysis and recommended actions popping up: “This prospect is ready to buy; time to ask for the close!” or “This person remains a bit skeptical; here is a white paper you can offer to send their way that might help.”
Sales training and asset management are two other areas where AI in sales has great potential to drive performance improvement. Imagine, for example, an intelligent agent capable of analyzing all the interactions associated with deals that fail to close within certain timeframes. Based on what went wrong, the agent might then suggest a certain digital training module for a rep to complete. Alternatively, it could serve up a series of marketing assets with specific suggestions for how the rep should use them to reheat a cold lead.
Leading sales technology companies already offer some of the AI capabilities described above, and there is no question that more AI is coming to the sales function soon. Sales leaders need to be deciding right now where they plan to be on the adoption curve. First movers and close followers will be more likely to experience benefits such as enhanced competitiveness, while lagging companies stand to benefit from the lessons learned as AI in sales apps move increasingly from theory to practice.
To prepare your organization for AI in sales, here are a few big questions to consider:
Where does our customer data reside? Is it all in one CRM that is integrated for marketing and sales? Or fragmented across multiple systems? The more fragmented the data, the more difficult it will be to train ML algorithms effectively.
What kinds of customer data do we have? How might that inform (or fail to inform) the questions we would want a machine learning model to answer for us? For example, do we capture all customer-brand interactions into a single, unified system? If key data is lacking, now is the time to start rectifying that issue.
Is our current CRM provider investing in AI-powered functions? What is their history with AI? And are they offering authentic or pseudo-AI? This should, for certain, be a major point for discussion in any renewal or upgrade conversations.
What kind of relationships do we have with our sales tech providers? Can we help to influence or drive their forays into developing AI-powered functionality?
How are our competitors using AI? Do we have opportunities for competitive advantage by becoming a first mover or close follower in this area?
What is the digital maturity of our sales teams? Is there advance training and development we might do to ensure that AI-powered technology is taken up quickly and used effectively when we start making it available? Can we begin planning now for the change management that would be required should we decide to incorporate more AI in daily sales workflows?
Sorry, our deep-dive didn’t help. Please try a different search term.