Since the emergence of ChatGPT on to the scene in late 2022, interest in the power of Generative AI has been astonishing.
As with any new technology that captures the attention and imagination of those who love technology and are adept at social media, there has been quite a lot of noise about AI which has not necessarily been accompanied by an attendant amount of light.
So, what’s all the fuss about? How much of it is hype, and how much of a real opportunity is it for organisations and their people?
First, some definitions (if you know about this stuff, skip ahead).
What is Generative AI?
Generative AI, at its most basic, is best understood in two parts. ‘Generative’ means that it can create new digital content – whether graphics, video, text, or speech. Generative AI can also write new computer code – perhaps not obvious to most of us, but potentially greatly accelerating the creation of new software or new apps.
The AI piece means – of course - Artificial Intelligence. The definition of AI has not really changed much since its invention in 1956. It was then, and is now, the simulation of human intelligence by machines – usually computers.
While AI powers most of the world’s search engines, the move from the provision of a series of links that match some or all of the search terms used to a human-like answer to a question is a quantum leap. The ability that the user now has to ask follow-up questions moves beyond the realm of ‘looking something up’ and into the realm of having a digital assistant.
Although named for its generative capability, it is the interactive component of these new technologies that is most exciting.
Discover how to seamlessly integrate AI into your sales process. Download the whitepaper.
Imagine searching for something on the web. You open a search engine, type in what you are looking for and in milliseconds you have 2.5 million links which will, ostensibly, answer your question. Many of these will have a tangential connection, at best. You will need to wade through the sponsored links which have popped up because you entered some keyword that someone somewhere thinks surfaces your interest in a holiday to Barbados, a magic cure for baldness or the chance to win a car in a raffle (delete as per your most frequent searches).
If you think the search was not very good, enter another search question and start the process all over again!
Now, imagine asking a question of a Generative AI tool. It will provide an answer. You may think the answer is terrific; you may feel it is not quite what you were looking for. You ask it another question. This time, you are not entering a new search for information, but requesting a refinement or extension of the previous answer.
You may ask for the source of the information presented. You may refine your prompt to ask for the output in a different format, or to include a specific perspective or to set some kind of boundaries around the response given. You have interacted with the tool as though it were an endlessly patient and beneficent human.
This human-like interaction with a computer is the basis of Prompt Engineering – and the dream of the early pioneers of Artificial Intelligence.
Prompt engineering is the process of structuring text or a question that can be interpreted and understood by a Generative AI model in order to produce a desired output. When anyone first tries to interact with Gen AI, inevitably, there is a process of trial and error. The first prompts may be somewhat wide of the mark. They may not quite produce what was required. Effective prompt engineering increases efficiency by reducing the number of interactions to achieve the desired output.
However, while the information presented may be produced quickly and be more accurate than simply consulting a single internet source found through a search engine, this does not make it infallible. Although different Generative AI tools work differently, they mostly draw data from the internet. The rules set by their creators determine the information sources they use and, therefore, the reliability of the output.
Some Gen AIs only source information from peer-reviewed papers. Others will extend out to other sites where there is robust fact checking – including certain media sites, government and intra-governmental research bodies, and encyclopaedias, as well as academic and educational websites. Others will simply use the internet, or a version of it, depending on how and on what content it was initially trained. This potentially presents risks related to copyright infringement, incorrect information and material posted by interest groups with – potentially – scant regard for truth, honesty or accuracy.
Machine learning is the blanket term for all the processes that help the AI to improve its ability to respond to prompts. Essentially, when you create a prompt, the information you feed in is retained by the Gen AI.
The number and content of the prompts you wrote in order to satisfy your request are recorded and the algorithms are adjusted to improve performance (i.e. fewer prompts required to get to a satisfactory response).
AI, sales and learning
If you read those definitions, you might think, “That’s interesting. We have new tools that review large amounts of data and give us an answer. What’s that got to do with sales, commercial skills or helping people to change their behaviour?”
According to an oft cited piece of research by Gartner, “60% of sales activities will be undertaken by AI by 2028”.
Goldman Sachs told the world in early 2023 that “300 million jobs worldwide will be affected by Generative AI”.
Those are big numbers. Is there any evidence to support these assertions?
I decided to ask a Generative AI – Claude.ai by Anthropic. This is what it told me:
"Goldman Sachs has not provided detailed analysis or methodology to back up their claim that 300 million jobs could be affected by AI. This seems to be more of a broad estimate or projection rather than a definitive research finding."
A few points about this claim:
- It does not necessarily mean 300 million jobs will be eliminated, but that they could be impacted or disrupted in some way by AI and automation.
- The types of jobs most susceptible are routine, repetitive roles that can be automated by AI systems and software – think data entry, customer service, basic office tasks, etc.
- However, AI could also create new jobs in areas like AI development, data labeling, analysis of AI outputs, and so on.
- The timeline for when these 300 million jobs may be affected is unclear. Widespread AI disruption could take years or decades to fully play out across industries.
- Major analysts and think tanks like McKinsey and the OECD have made similar estimates about the future impact of automation on the global workforce being in the hundreds of millions of jobs.
In summary, it's an eye-catching projected number meant to illustrate AI's potential to reshape the job market, but the specifics behind Goldman's 300 million estimate are not clearly laid out. The overall direction of AI disrupting certain job categories does seem directionally valid based on most expert analysis.
So: possible, but it is an estimate.
How speculative or well researched this best-guess is isn’t clear, but the final bullet point seems to suggest Goldman Sachs are not alone in their thinking.
What about Gartner’s claim that 60% of sales activities will be undertaken by AI in just 5 years from the date of the projection being made? I asked Claude again. The answer was pretty lengthy, so I’ll pull out a few key statements:
"There does not appear to be detailed published research or methodology behind Gartner's specific claim that 60% of sales activities will be undertaken by AI by 2028.
AI is certainly being applied to many sales and marketing activities, like lead generation, data analysis, forecasting, customer segmentation, etc. However, claiming 60% of all sales activities will be fully automated is quite an aggressive projection.
Many sales tasks, especially those involving human relationship building, negotiation, and closing deals, are difficult to fully automate with current AI capabilities. Augmenting human sales staff with AI seems more realistic in the near-term.
Gartner's specific 60% projection for 2028 does not appear to be convincingly backed up by publicly available analysis or data. Reasonable experts may disagree on the precise extent and timeline for AI replacing human sales roles."
Uncover how to create a seamless integration of AI into your sales operations. Download the whitepaper.
Let’s pick up on two of those points:
1. AI is being applied to many sales and marketing activities
The range of sales and marketing activities imagined here go way beyond the traditional role of the Business-to-Business field sales professional. As well as there being no clear distinction between business-to-consumer and B2B, the automation using AI extends to inside sales, lead generation, sales forecasting and revenue operations.
Some of this automation is about the customer:supplier interaction. Chatbots may have been around for a while, but we have all had experience of frustrating interactions with badly programmed avatars on websites. The power of Generative AI to more effectively understand the customer questions and route the customer towards a transaction clearly replicates and potentially replaces some inside sales activity – especially for in-bound agents.
Furthermore, Generative Pre-trained Transformers (or GPTs) can be embedded in customer-facing websites supporting online purchasing – providing much richer and more effective interactions than a few FAQs with pre-programmed answers. For organisations who have a business based on repeat orders, these models are likely to be both effective and a requirement of procurement teams seeking to reduce the cost of just-in-time supply lines and stock replenishment.
One of the most regularly seen applications of Generative AI in sales activity at the moment is in the creation of emails going out to potential prospects. Many of you will have noticed an increase in the number of direct emails in your inboxes since the emergence of ChatGPT. A seller can feed in a name and job title in order to generate a more personalised email that the standard mail out and, of course, by working from a database of contacts, the whole process of sending out direct messages via email, LinkedIn or other platforms is relatively simple to automate – even down to including the opportunity to book a follow up meeting with the designated seller via an online calendar tool. Significantly more messages can be generated with fewer resources and those messages appear more personal.
But let’s remember that spam is still spam and the ability to annoy more people more quickly at a cheaper cost seems a limited benefit of applying this breakthrough technology.
2. Some sales tasks are difficult to automate with current AI capabilities.
This speaks to the uniquely human skills of sellers and the role that those sellers play in the buyer’s journey from contact to contract.
Over five decades, Huthwaite International has been researching the behaviours that successful sellers use differently – and more effectively – than their less successful peers. Some of these behaviours are “uniquely human” in that they cannot be easily replicated by artificial intelligence (at least at the moment).
The first of these is asking high-quality questions that uncover customer needs and requirements. We know from our research and the research of other academics that asking questions builds rapport. Rapport between the seller – and by extension the seller’s organisation – and the buyer and their organisation. Asking questions which, on face value, may feel intrusive or which involve the buyer in revealing information they may be reluctant to share, relies on some kind of rapport. That rapport may be built on the basis of professional respect, acknowledgement of expertise or personal connection (and maybe all three). Whatever the basis for that rapport and connection, it is essential if the seller is to gain high-quality answers to high-quality questions.
The great thing is that asking those high-quality questions contributes significantly to the creation of rapport and to the building of trust. In using AI to sell, most buyers are unlikely or unwilling to forget that they are talking to a machine – however smart or however much it mimics human interaction.
Having uncovered problems, good sellers recognise that customers live with problems. Until the problem they are facing is also matched with a vision of a future when the problem is resolved, many buyers will endure the pain they are experiencing, rather than make a potentially expensive or risky decision to embrace change. Again, the presence of a trusted human with clear expertise can support the customer’s progression from despair to hope much more effectively than any form of automation.
Finally, influencing someone towards a particular solution and demonstrating organisational capability and the value of working together gets harder the greater the potential consequences of making a wrong purchasing decision.
Example: Choosing a paper supplier is relatively easy – we can revert back to our old supplier if the promises made do not materialise. It’s much more difficult to reverse the decision to install different printing equipment into which the paper will be loaded.
People are influenced by people – not by bots. Mitigating risk, feeling positive about change and willingly stepping into the unknown, are best undertaken with the support of a skilled and trusted human.
Whilst those examples are “uniquely human” skills and behaviours, they can be supported – and super-powered – by better access to data, which is summarised and presented to the salesperson in ways that are quickly digestible and able to be put to use.
AI can help here. Those organisations who use their own data to help generate summaries and answers to seller questions can enhance the ability of skilled sellers to undertake those sales tasks.
However, the companies who succeed will have in place clear governance regarding their use of AI. They will have mitigated or removed any potential risks of sharing their data with a piece of software that lives on the internet. Let’s not forget that these Generative AI tools improve using Machine Learning – any interactions with them generates data which informs future interactions. While it is not a given that this involves you surrendering any of your confidential data or data about your customers and prospects, it is also not a given that it doesn’t.
Successful users of AI to enhance seller capabilities will also build on a foundation of having trained their staff to spot anomalies, mistakes and biases which have the potential to produce misleading or plain wrong responses to questions asked of AI. Generative AI is a co-pilot that assists people in decision making, planning and taking action. To succeed, smart technology needs smart people.
Key takeaways
It would be easy to say that sellers’ roles could be replaced by AI. Smart organisations recognise that customer acceptance will be key to the role that AI plays.
AI could handle most aspects of remote selling – but business-to-business selling has always required more than order taking. Smart salespeople, supported by better – and more accessible – data than ever before could be released from low value and repetitive tasks. This will enable them to play a crucial role in enabling buyers to make informed choices to meet their needs.
Uncover more AI & Sales insights
Discover more about how AI will affect salespeople in the coming years and learn what you can do to stay ahead of the curve.
