Enqueued CSS Stylesheets

Enqueued JS Scripts

Recognizing Valuable Intent Signals In The Era of Super-Smart Bots

Posted September 27, 2017

Recognizing Valuable Intent Signals In The Era of Super-Smart Bots

Posted September 27, 2017

The news last month that the creators of South Park had jokingly tricked Amazon Echo devices into filling their owners’ shopping carts with needless items – including South Park DVD box sets – reminded me of a question I have been considering for quite a while now.

Essentially, you had two machines speaking to each other and conducting business without direct human interaction. Sure, this time it was a clever publicity stunt, but in the future, it could be a concerted effort to scam consumers or businesses.

Or, even more fascinating to B2B marketers, smart machines could actually be acting on the orders of executives, and spending decision-maker’s dollars, gathering information about products and services, – and creating valuable intent signals and marketing intelligence as they go about their automated tasks.

Most of the buzz around “bots,” as they are commonly called, focuses on their potential as marketing tools, to engage users at various activities in the funnel. (Martech Advisor has as good an overview piece on the hype as I have seen lately.) A relatively smart chatbot can quickly act on a whitepaper download, with a few interactive questions – the buzzword here is Conversational Marketing – to register a user for a webinar. Bots that simulate human engagement over popular messaging apps are predicted to largely cut into, if not replace, e-mail as the primary outbound marketing device.

All this sounds great for the future, and B2B marketers do need to be thinking about how to incorporate bots into their automated marketing and nurturing programs down the road.

But, as of now, we are still thinking about bots acting solely as marketers.

My question remains – when do we have to recognize that bots can act as customers? And how will we weigh in on and incorporate the data they create into the increasingly sophisticated logic that drives successful B2B marketing?

New Models of Filtering Bot Activity

At Anteriad, we’ve spent a ton of time getting to know the first-generation bots that create what we consider to be false positives for intent signals; signs that a purchase decision-maker is ready to more deeply engage with your company. (Technically speaking, a “bot” is just a piece of code that can react to various inputs as it does tasks with limited, or no, human intervention.)

Initially, the bots we hunted were site scrappers and indexers that everyone now perceives to be junk. However, the technology has become more advanced. Bots that are legitimately deployed by security providers like Baracuda to defend mailboxes can rapidly create multiple mail responses, or what appear to be valid test app downloads or form completions. No one is acting with bad intent here; there’s a legitimate reason for these bots to behave as they do. It’s just up to us, as a intent monitoring data provider, to isolate this into information that is not really useful to marketers.

Forgive a little chest-thumping – we believe our InsightBase Relevance Engine cleans up this information as well as any service in the market today. We’ve already identified 3 million or so sources of false positives which can really skew your data.

We’re confident, as a result, that we are on top of that aspect of managing bots as a problem in collecting marketing data.

Now, we have to begin thinking about how to cultivate their activity as an asset.

So, How Bot a Deal Is It?

First, we’ll have to consider how much importance to place on that fact that a company would create and send out a bot to research a topic. Obviously, writing a piece of code to execute such a task indicates some level of meaningful interest, and needs to be factored into your Account-Based Marketing heat index for that company, at the very least.

However, a big value proposition of bots is that once launched, they basically cost nothing to operate. A prospect could simply have bots out gathering and indexing content for all their purchase categories, regardless of where they are in the purchase journey. A bot downloading a ton of whitepapers to the account of a mid-level manager, and then salting them away in a directory someplace, for example, might be less valuable as an intent signal than the same activity from an actual human system administrator.

At some point, data analysts will need to evaluate volumes of conversions against some plausible standard, and begin flagging accounts accordingly. This could extend, even to webinar attendance; a bot could be programmed to attend a session and monitor for specific keywords. That’s a great intent signal, but not as rich as an actual human executive attending the live session.

How Valuable Is A Little Friction In A Conversion?

A key selling point in the bot revolution is that potential leads never have to fill out a form.

A smart chatbot will simply prompt for text or voice-based responses, which gather all data required to sign up a prospect for a webinar, without the high rate of attrition marketers see whenever users are presented with even the most simple forms. In fact, a really smart chatbot will learn how to deftly upsell your prospects right there in the same chat.

Sounds great, right? But at some point, you have to ask yourself how ready a prospect is to commit a webinar, if they just downloaded a whitepaper a few seconds ago? Remember, the key to successful automated marketing is right-time, not real-time. Careful analysis of your data may reveal that a good old-fashioned drip series is the best way to ultimately land a prospect, even if an advanced bot is better at pushing them along the funnel more quickly.

And that’s if the prospect is human. Imagine if both sides of the conversation are advanced learning bots, programmed to gather and promote as much information as possible.

Did the Bot Do What Its Masters Ordered?

As bots get smarter and begin to employ Artificial Intelligence to “learn on the job,” how will we know that a given bot is staying on task, and not just going off on a tangent, like a pair of capricious interns?

This may sound like sci-fi, but consider that Facebook recently shut down a couple of AI bots, after they invented their own shorthand language to conduct simple trading exercises.

Now picture two advanced AI bots – one designed to go out and collect information on, let’s say, Supply Chain Management, and one designed to artfully steer qualified candidates to a live webinar event. Both bots employ natural language processing (NLP), an advanced technology which enables them to understand and react to language as it is spoken or typed.
By the end of their exchange, the CFO that sent out the bot may be signed up for an online course in Girl Scout cookie delivery best practices without ever having known it.

Of course, that’s hyperbole – maybe. Ultimately, we will need to be able to discriminate between on-task bot activity and “leaning” activities which have maybe gotten a little too smart for their own good. Again, this will require an advanced analysis of interaction volume, plus some human intuition, and taxonomy tweaking, to spot those “tangents” that seem just a little too far afield from the initial offer.

Bots Pay Off Today, Hold Promise for Tomorrow

Please be aware: the scenarios of bots acting as customers like we just discussed are still largely speculative. Of course, the notion of instantaneously sending millions of electronic mails over something called “the internet” was once futurism, as well.

For the foreseeable future, bots will continue to gain traction in marketing, primarily as upsell mechanisms for high-funnel conversions, or customer order interfaces. They also work great for an instantaneous follow-up to a user-initiated search, or any other request for information on your web site. Research suggests that that your odds of ultimately landing a lead can be as much as four times higher if you trim just five minutes off your initial response time. And bots are much faster than humans.

Most current bots are based on simple retrieval of data from an FAQ or other structured data – NLP and true AI will remain the domain of high-end technology companies for the foreseeable future. And, as with all automated marketing tools, the personalization of content and messaging is essential, in both inbound and outbound applications. Integrating bot activity with your Marketing Automation System and customer database, including intent data from services like InsightBASE, will be a huge advantage in making bots seem less robotic.

If your bot is going to start a conversation with someone, at least it can pay them the courtesy of knowing a little bit about them first.