How to approach AI SDR

We will (whether we like it or not) eventually leverage AI and data to remove people from as much of the GTM process as possible.??

This article looks at the considerations in rolling out an AI SDR strategy.? It takes a holistic approach and, rather than staking an opinion, poses seven questions to guide the conversation.??

Question #1: AI can drive efficiency - but can it improve conversion?

GTM teams are saddled with manual, repetitive work that drains their energy and focus. But these time sinks aren’t the total of their job. For instance, in the SDR function, 30-40% of time is spent on automatable tasks like list building and account research. A major part of their job, the high-leverage portion, is actually on building human relationships with prospects. The greater efficiency that comes from AI automating these tasks could clear space for SDRs to focus on higher value – and human – activities like relationship building, time on the phone, improved research, etc. This can improve conversion rates by creating deeper connections and relevance for the prospect.

When companies implement AI outbound, they hope it will drive improvements in both:?

• Efficiency: ?Eliminating headcount

• Conversion: ?Improved learning through data, testing and personalization.??

But there is still a lot of debate around this.? Does AI improve conversion in the early stages of funnel through personalization and rapid learning or is rep creativity and uniqueness still critical?

“I don’t believe that conversion will drop much at all in a world where AI is able to be hyper relevant (likely 12-24 months away from this being true)”? Austin Hughes, CEO at Unify

Question #2: Where shouldn’t you automate reps?

Similar to tech-touch CS, AI SDR models promise to automate lower ACV, earlier funnel and invest in improving ratios for Top accounts. ? However, there are definitely areas you should not automate.

“Humans will own the phones and meetings, and AI will own everything else”? Austin Hughes, CEO at Unify

“AI will be a distinguishing factor to find out the type of buyer you're dealing with (speed, consultative in order to uncover the problem, expertise, etc), and be a supportive layer throughout the entire buying cycle creating an experience where no single human being was ever involved - if that's what the buyer ultimately wants.”? Justin Tung

“Being from CS, I will tell you that this is being done well at less than 5% of companies and effectiveness is not accelerating. AI support ticket resolution is pretty darn good but CS (i.e. value creation) is not good at all.” ? Mark Kosoglow, Founder of Operator?

"People still want to buy from people. This is more important than ever." Rohan @ Nooks

There are likely (at least until AI calling can pass for humans and are allowed by the government) some areas you will always want your reps involved:

• Calls: ?You can’t automate relationship building and legally you can’t automate calls (at least in the United States)??
• Email Replies:? Prospects who reply to emails are deeply engaged.? Most companies escalate to a human as soon as they have a positive intent signal.??
• Meetings: ?Humans will always be critical here

These touchpoints are down-funnel and thus high-value.? While you may be able to automate parts of them, the cost is high for any mistakes, including:

• Lost opportunity:? If a prospect determines they are chatting with a bot?
• Conversion:? Reps can engage personally and build relationships
• Your brand: ?There are many examples of AI chatbots gone rogue

Question #3: Which approach will win - Tesla or Waymo?

There are multiple ways to attack the SDR automation opportunity.? Tesla vs. Waymo is a powerful analogy:

• Tesla is trying to incrementally take over driver functions.? They have tons of data from cars on the road but their approach necessitates lots of handovers from driver to machine.
• Waymo chose the harder short-term way to remove the driver completely and force more rapid learning of automation.??

Below are flavors / approaches different companies are taking

Question #4: Should you automate outbound without a brand?

Outbound automated emails/sequences alone doesn’t cut it.? In fact performance is declining with all the automation today.? You have to invest in all 5 components of great outbound

• Targeting: Who you target
• Brand: Your reputation
• Assets: What you give / offer
• Channel: Where you reach someone?
• Message: How you address pain points,? communicate value prop / offer

"When a prospect picks up the phone and your team says “Hi, this is X calling from Y.” if the prospect knows Y, even dimly, they are much more likely to stay on the phone for a moment longer." Rohan @ Nooks

Question #5:? Should you automate outbound without great data?

Data matters more than ever when you are trying to automate critical touches with customers.??

• AI is only as good as the data underlying it
• Scoring and personalizing won’t work if you don’t have great data
• It's important for next gen tools to eliminate some of the dependency on data engineers. They are extremely scarce and good ones are even more rare
• When AI can use data to determine which contacts to engage with you will need unified systems, clear rules of engagement and aligned management
• See this article on how PLG changes RevOps.? The changes with AI are likely to be similar

“A big reason why we're seeing autopilot backlash is bad data. The performance of an agent/autopilot system is only as good as the data that underlies it. A lot of companies unfortunately have very messy data so unless you fix that, humans need to be in the loop to avoid some potentially brand killing oopsies”? Sandy Mangat @ Pocus

Great data can often be misunderstood as only third-party, publicly available data. But much of the best data comes from phone conversations, meeting notes, email interactions, and other first party data. Mining and reusing this information is a critical way that outbound communications can be radically more personalized. Not only for improving automated communications, but for improving SDRs’ call scripts and ongoing sales training. Rohan @ Nooks

Question #6:? What use cases should you prioritize?

There are lots of different functions SDRs serve.? It may be possible to automate all of them, but where should you start???

• Inbound demo requests:? These typically are already largely automated (with tools like Chili Piper / Lean Data / Qualified).? Prospects who have specifically asked to talk to sales can book directly on their calendar.? Given these leads convert at a high rate, it typically won’t make sense to further automate these but instead focus on getting them to speak to the the right rep.?
• Inbound scored leads:? Scored leads have a much lower conversion rate than demo requests and the quality can vary widely.? Using an automated scoring tool like MadKudu can help you determine which to send to reps vs. nurture with marketing.? AI SDR may be strong here at improving conversion over marketing nurture.??
• Outbound on Inbound: ?When a non-core persona from a company comes inbound you can explore multithreading these companies with smart automated outbound.
• True Outbound:? This is what we often think of as SDR.? It is also likely the hardest to automate for many of the reasons cited above.
• Expansion and retention: ?Combine product and intent data to target the right companies with the right products and engage them to book conversations.

As you explore these use cases, you should nuance your approach by segment.? This is likely to work better for low ACV transactions.

Question #7:? Once you have signal, what do you send?

There are many examples of automated outbound burning through TAM.??

“Triggers become less useful the more people that know about them and use them. Triggers will lose effectiveness as more companies automate the best plays.? A trigger is NOT a reason to reach out. Yet, a huge percentage of companies treat them that way.”? Mark Kosoglow, Founder of Operator

To have a chance to succeed, you need to know "what" to say when there's a signal of intent, not just “when” to say it. ? This is what a new breed of companies like Operator are working on!

“The AI has to apply the best reps' business acumen to the data to find what's compelling and tie that to their value prop. That's how you attract people into conversations vs load up a list, spam them, and hope that the timing gets you a high enough reply rate to build the req'd pipeline (which really is how this all has been working in the past).” ? Mark Kosoglow, Founder of Operator

This is much harder because classic “personalization” isn’t really personalized at all.? This automated personalization is just AI sending what an algorithm thinks is the next best word.? But to do great outbound you need creativity and pattern breaks that will engage prospects.??

Conclusion

Having spent 3 years stopping billions of unwanted emails with Gated I’ve studied this problem deeply.??

People ask me if I have a spam problem today and I truly don’t because I mark every single unsolicited email as spam UNLESS they read my LinkedIn profile.??

There I literally hand SDRs a roadmap for how to reach me.? Surprisingly less than ? of 1% of reps do this level of research. Those who take the time to write personalized emails linking my business problems to Keanu’s great movie career definitely get my attention.

I truly hope the world of outbound doesn’t lose the personal touch and creativity of reps. .??

But, If you are going to automate outbound, the system needs to be able to:

• Learn from rep feedback: ?Let reps share feedback on why they don’t think a signal is valuable
• Learn from prospect feedback: ?Pull back when it isn’t seeing engagement.??

Don’t become a spam cannon!? Please!