the wrong way to use AI in your sales messaging

The Wrong Way to Use AI in Your Sales Messaging – and What to Do Instead

AI is finding its way into ever more sales teams’ workflows. If you’ve ever thought that there’s a lot more talk about AI than there is actual implementation, you are on the money. There’s an industry-wide mismatch between our stated expectations of AI and our involvement with it.

According to Demandbase, just under 80% of sales and marketing professionals expect a significant amount of the applications they use will be AI-powered by 2020.

Yet the same survey found that while 36% of respondents felt they were ready to move onto AI, 64% were not – and 5% (one in 20 of those questioned) didn’t even know what it is.

current knowledge and awareness of ai applications

Source

What’s causing this disconnect between AI as the subject of a million emails and blog posts, not to say derivative stock graphics…

AI nueral network cyborg brain

Source

…and people’s actual experience of using AI in sales?

Some of it is hype, of course. And some of it is that when AI does show up in our working days and the tools we use every day, it doesn’t look a whole lot like cheesy ‘galaxy brain’ illustrations. It’s usually just whatever we’d normally do, made simpler.

But a lot of the reason is that AI is still in the very early stages of use.

Many companies that are using AI successfully are using it for quite low-level applications. They’re using it to power CRM search, enrich contacts, and to facilitate team communications. All valid use cases, all adding value – but none is truly transformative. So far the AI revolution looks like business as usual to a lot of salespeople.

And a lot of companies are slightly wary of getting deeper into an unproven technology, so they keep AI out of core functions.

We’re big advocates of handing off to AI all the work we can, so long that the work happens to be a good fit for what AI’s good at. That means anything that involves a lot of data, a lot of calculations, or both. Insights, ideas, and human connections all have to come from people.

When businesses try to apply AI to their sales messaging and they get it the wrong way around, that’s when the problems begin.

Here’s how we see businesses shoehorning AI in where it doesn’t belong – and where it should be employed to best effect.

Prospecting and messaging: quality or numbers?

The biggest issue in most sales processes arises, unsurprisingly, in the part of the pipeline that gets the least amount of effort and attention per potential customer.

Before anyone picks up a phone or sends a single email, the process of gathering and validating data on prospects is among the biggest deciding factors that can predict how well the rest of the pipeline will perform.

However, we’re all so used to working this as a numbers game where we basically try to cram as many people or accounts into the hopper as possible, that this often doesn’t surface in planning.

More messages =/= better messaging

When many sales managers look at potential use cases for AI in prospecting, they look at ways to produce more messages. More outbound phone calls and emails, and a step closer to the automated drip flows that marketing people have been using for a decade.

The ‘chatbot revolution’ hasn’t made as big of a dent in outbound sales as in inbound, and far less than in marketing, though automated voicemails have streamlined sales processes.

AI is often used to automate workflows, bringing tasks to the fore automatically for SDRs and accelerating the existing process.

Alternatively, it’s used to build support structures – autologging, and other forms of busywork automation, help SDRs be more productive, and we think they’re great – as far as they go.

But this approach brings AI to bear on your sales messages – not your sales messaging.

The alternative is to switch attention from the volume of prospects  – which AI can certainly increase – to the fit, and refocus on using AI to develop an airtight seal between the right prospects and the right messaging.

This actually requires using more AI, as well as being far more strategic about how it’s used. And in a way, it involves rethinking the whole way we approach prospect selection and research.

(So I guess we found the AI revolution after all.)

We use our own AI at this stage to sift through multiple large databases for the best matches to ideal customer profiles (ICPs), using far more data volume and a far wider variety of data points than would be possible with humans.

This is an approach to prospecting that would be impossible without artificial intelligence unless volume was super low and deal size insanely high. We can take a process that used to be available only for million dollar deals and make it extremely effective for $10,000 deals.

On the phone: script optimization or call management?

There’s no place for AI on the phone. From the moment someone picks up, every second is far too valuable to not allocate human effort to it.

For one thing, when you hear a robot voice on the phone you hang up. It reminds you of being told which number caller you are and that your call is important. Not the look you’re going for, right?

Even if there was an AI that could get on the phone and convincingly imitate a human being, we still wouldn’t recommend it; genuine relationships can’t be built that way, and at the point where you’re actually speaking to someone, that’s what you’re doing.

So the majority of applications for low-level, high-volume AI come in call queueing and cadence management. Simple machine learning tools, often presenting as a chatbot inside CRM or chat apps, deliver the right call at the right time, slashing reps’ reliance on their own memories and calendaring abilities.

That’s a good thing as far as it goes, but again – this is clerical work that’s being automated, not sales, and it’s not being allowed to touch messaging, only arrange messages.

However, behind the scenes of a phone call there is enormous scope for leveraging the power of artificial intelligence to adapt messaging. Phone call scripts are usually created by sales managers and optimized on the fly based on what seems to work, by harassed, overworked SDRs.

In the very best sales organizations, great SDRs have the time and freedom to test their own sales call scripts, and they do their best to be scientific about it. Any effort to be more data-driven is laudable and this approach does produce results.

But these SDRs and sales managers are working on very small amounts of data; even in teams where SDRs make a hundred calls a day, they don’t have access to the data from across the team, and they don’t have the calculating power of even a simple machine-learning tool.

These are things that humans inherently are not good at, and from the ruler through the abacus to a server rack full of big-data-crunching AI, we’ve always gotten further by outsourcing it to nonhuman tools.

So rather than asking AI to spit out a readymade call script, the best thing to do is utilize its inherent strengths by using it to constantly test call script elements against each other and optimize them. We’ve seen dramatic results by doing this, even though it’s resulted in some very counterintuitive scripts, so we know it works.

What about emails? Surely we can AI the heck out of emails, right?

AI seems like it’s built for emails. So much so that Google has built an AI that auto-generates replies from your Gmail inbox based on your previous replies, and rolled it out to all G Suite customers.

And the focus has again been mostly on more messages, or on simplifying the process of creating messages. Rather than rebuild sales processes from the ground up to take account of the power of artificial intelligence, most sales teams have shoehorned it in at a low level to accelerate the sales process they’re already using.

The result is to create sales messages that are no better aligned with what prospects want to hear than before, just in larger volumes and more carefully aligned cadences.

While that does generate incremental improvements in sales, it isn’t capable of creating really significant, exponential increases; it can add five or 10 percent to your sales, done right, but it will never double them.

Instead of putting AI where there’s space for it, it’s better to look at what it’s extremely good at and then see how to integrate that functionality into your sales process. Doing that means fundamentally redesigning your sales process, but when it’s done right, it’s worth it.

For emails that means taking the whole message apart and analyzing it part by part; everything from the subject line to the opening line to the sign-off can be optimized, through a mixture of careful matching with the prospect, and constant high-volume testing.

There are more variables in an email than in a phone call that can be tested and measured, so this channel is actually even better suited to using AI to radically increase value.

When we’ve carried this approach to its logical conclusion, the results have been counterintuitive. Clients have told us that they were so surprised by the emails we delivered, they weren’t sure if they should send them.

outboundworks testimonial

Source

When they did, they saw the number of meetings they were able to book skyrocket, and they’re now getting six or eight qualified meetings a week.

Once those meetings are booked, or once reps are talking with prospects that they have established a relationship with, trying to use AI to optimize the content of those conversations delivers catastrophically diminishing returns because the number of variables becomes too great. The kind of subtlety, skill and rapport that great sales conversations depend on is far beyond what AI can do.

But analyzing relatively simple messaging, comparing it to ICP and enriched contact data and constantly testing? Machine learning, AI, call it what you will – it’s tailor-made for that.

Conclusion

The right place to put AI is behind the scenes, and in the design and layout of your messaging. Anything that can be A/B tested to a demonstrable improvement in results is a perfect fit for AI.

But anything that requires ideas or human contact is a terrible fit for AI.

Bottom line: AI should be used so salespeople are saying the right thing to the right person, not to replace salespeople’s own skills and talent.

Have you experienced implementing AI into your own sales messaging design or delivery? Got a story to share? Let us know!