AI as a source of insight or as a route to summarising, and making sense of large data sets, is one perspective on how sellers’ capability may be enhanced by these new tools.
But learning at work is far more than knowing things. AI is great at what sellers should do, but less effective in how they perform their tasks. This is especially true for those skills and behaviours that go beyond following a process and focus on the crucial communications skills all sellers need.
Enter Conversation Intelligence.
From sales enablement platforms to CRM systems to standalone apps, it is now possible to record and use AI to analyse what sellers say and how they behave in their conversations with customers.
Or is it?
It would be glib to describe these tools as neither conversational nor intelligent. Unfortunately, for many of them, that description is only too true. But that doesn’t mean that AI cannot provide analysis and insight about real, live customer interactions. It can, but the current crop of tools have been rapidly rushed into the market, and – as so many technologies have done in that past – many of them over-promise and under-deliver.
Make informed choices about conversational intelligence
Let’s look at the issues (and this list of considerations and questions may also provide an assessment guide if you have an opportunity to introduce one of these tools to help you build your team’s sales capability).
1. What has the tool been trained on?
You know when you contact an organisation – or when an organisation contacts you – you may hear an automated voice tell you that ‘this call will be recorded for training and quality purposes’? Well, one of those training purposes is now to train an AI tool.
These announcements are usually associated with contact centres or inside sales teams. From a sales perspective, many of these calls are transactional, hoping for a quick decision to purchase a relatively low value item or to book an appointment with ‘one of our consultants’. Where the focus is on selling a product or service, many inside sellers use some kind of script or a series of talking points. These conversations are extremely numerous. Recording them is relatively easy. As a result, the bulk of existing Conversational Intelligence tools for sales have been trained on these conversations.
What the AI ‘learns’ is whether the seller followed the script, addressed any objections raised by the customer or managed any conversation about a competitor product or service. They also gather frequently made comments from prospects or customers and use this data to refine the scripts and call guides. The individual agent’s performance against these metrics is presented to them in a dashboard or chart with potential guidance about how to improve – either delivered by the AI, or by a coach or manager reviewing calls with their team.
The problem for business-to-business sellers is that such insights are not entirely relevant to many of the more complex communication behaviours required in a consultative sales call. The idea of perfecting one’s use of a script – especially one based on the features of the product or service under discussion – would be of limited value. Huthwaite International’s research into effective seller behaviours has repeatedly found that describing a product’s features is not persuasive, and often leads to customer objections (and that – contrary to some sales training mantras – a customer objection is not a buying signal).
It follows, therefore, that perfecting one’s ability to parrot the product description would be counterproductive at best and could be actively harmful to your prospects of success.
Discover how to seamlessly integrate AI into your sales process. Download the whitepaper.
2. Training and standards
This inevitably brings us to a discussion of what standards are applied by the AI. To provide feedback on performance, there needs to be some kind of exemplar or definition of what good looks like. This may be organisation specific.
Most conversational intelligence tools are built on the infrastructure of one of the commonly available Generative AI tools – such as Chat GPT, Bard or Gemini. These Large Language Models are general. In AI, a Large Language Model means humungous!
The processing overhead – and therefore cost – of using a model that may be able to analyse your sellers’ capability, at the same time as being able to provide you with a detailed history of the capital of Mongolia, is significant. A general tool may not be quite as focused as you will need, and you may be paying for features you neither need nor will improve seller efficiency.
What’s more, to gain differentiation from other organisations, you will need to train these tools to focus on your own standards.
- What does a good conversation sound like and look like for you and your organisation?
- Is there only one approved approach, or will it be different depending on where the customer is in their journey?
- Can your in-house experts anticipate the different questions, concerns or challenges your seller’s may face?
And what about scale? If you have had the good sense to realise that you would like your sellers to have effective business conversations that are valued by prospects and customers – then you have a lot of training to do.
Neural Networks underpin most Generative AI tools. They are the mechanisms which underpin machine learning. They attempt to replicate how the human brain works to characterise data, weigh options and draw conclusions (Source IBM). To adapt these neural networks to learn what good looks like for you will likely need many thousands of data points within examples given to the AI.
To start that process, you will need to agree, adopt or reinforce a common language and agreed set of conversational behaviours across your entire sales team. Your sellers will need to know what these behaviours are, why they are important and how they enable them to meet their objectives for each call.
Most importantly, your sellers will need to know how to communicate with their customers, not just what to say about your company, its products and services.
3. Implementation mechanism
Once you have a tool that can ‘listen in’ on a conversation between your seller and a customer and provide feedback, how will that work?
Some Conversational Intelligence tools record what the seller says in live conversations but do not record the responses from the customer. While this may work from a data privacy perspective, it is – by definition – not conversational.
What about recording live conversations? Will your customer consent? What are the ethics and data protection considerations? What are the technical considerations if the conversation is face to face? Can the system easily differentiate between different voices? If so, how many? (And if you can’t record live conversations, how will the AI pick up on – and learn – the behaviours used in successful sales meetings and how they differ from those conversations which are less successful?)
If live conversations are problematic – or the use of AI analysis is for training purposes ahead of live customer conversations – how is the feedback provided? Are seller’s left to interpret the AI analysis, insights and feedback or are they supported by a trainer or coach? How easy is it for sellers to interpret the dashboards? Is it clear to either seller or coach what the analysis is telling them and therefore what improvements could be worked on?
While it may seem like conversational analysis is a powerful tool for sellers (and I genuinely think it has the potential to be so) there are challenges. Coaching is perceived to be difficult and time consuming. Many managers – even in companies with a so-called coaching culture – are reluctant to coach and may consider it not part of their role. Many sales leaders have been promoted, not because they are good at supporting their teams, but because they have a good track record as a seller.
For some, this will be because they use positive behaviours and can share their experience. For others, it will have been more luck than capability. For a third group, they may be great but don’t necessarily have relevant insights with which to help their colleagues – or the skills or desire to share them.
The best seller may not always turn out to be the best coach.
Many sellers are inexpert at reflection and – again – may feel that reflecting on analysis and data is something that takes time they don’t have or for which they are psychologically ill-suited or ill-prepared. Many may even feel resentful at being provided with support via a computer, rather than from a real human colleague or a trainer/coach they respect.
What you analyse is only part of the problem, how you use that analysis to improve performance will be a significant part of the success (or otherwise) of any deployment of a conversational intelligence tool.
Key takeaways
SPIN® Selling is based on observation and analysis of what successful sellers do differently from their less successful colleagues. It will come as no surprise that Huthwaite International – the originators of SPIN® Selling and the organisation that carried out tens of thousands of observations of live sales conversations – believe that recording conversations and analysing behaviours is a good thing. Is AI yet able to provide this analysis? Is the analysis that it can provide valid and built on good practice? Is what is observed and recommended likely to lead to performance improvement?
The answer is possibly – but more work needs to be done. We’re working on it, but we’d rather get it right than rush to market with a product that creates more questions than it provides answers.
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.
