Like many companies that cater to small business, Fundbox Inc.’s website is a primary source of sales leads. The company, which makes short-term loans available to business owners who are having temporary cash flow problems, makes it easy for visitors to create an account with just an email address, phone number and password. But the downside of such simple sign-up is getting those visitors ever to come back.

Fundbox has found a solution in an artificial intelligence-based email assistant that reaches out to people whose activity has lapsed to invite them back. In most cases, those sales prospects, or leads, are never even aware that they’re dealing with machine.

Lead nurturing is one of the most frustrating challenges of business-to-business marketers. The web is full of tire kickers who leave traces of themselves behind when they download a document or listen to a webcast but never speak to the company again. A 2015 survey of B2B marketers by BrightTalk Inc. reported that more than two-thirds said finding high-quality leads is their top priority while 59 percent said lead quality is their biggest challenge. MarketingSherpa LLC estimates that nearly three-quarters of sales leads are unlikely every to make a purchase. That adds up to a lot of wasted time for sales representatives. One study found that bad leads cost about $20,000 per sales representative.

Stuck in the pipeline

That kind of waste wasn’t acceptable to a five-year-old venture-backed company. “We’re growing very quickly, and it doesn’t make sense for us to use people to reach everyone who gets stuck in the pipeline,” said Basile Senesi, head of Sales at Fundbox. “The question was how to communicate as quickly as possible in the way people want to be contacted?”

The Turing test is a benchmark of artificial intelligence that challenges a machine to engage in natural language conversations that, to an uninformed observer, appear to be human. Conversica Inc. is applying that concept to email, and so far the machine is holding its own.

Conversica’s automated assistant shoulders the burden of sending follow-up messages to people who have expressed interest but never followed through. At Fundbox, they call the software Rachel. Using a combination of data about the prospective customer and knowledge of past interactions, Rachel constructs customized messages that appear to come from a human, or at least people think they do.

In contrast to the impersonal, canned messages that are typical of marketing communications, Rachel’s messages are in plain text, written in the first person and use an informal, conversational tone, such as “Hi, Jim, I saw that you registered on our site two weeks ago and haven’t come back. Is there any more information I can get you? Please just let me know and I’ll set up a call. Take care and have a great afternoon!”

Conversica says its customers have sent and received about 200 million messages using its system, giving the software a substantial training base with which to create messages that are both personal and persuasive. The software has delivered an immediate lift in lead response for Fundbox.

“It’s so easy to believe you’re having a conversation with a real person that people will often ask about Rachel when they call,” Senesi said. Automating routine follow-ups has enabled the company’s team of about 20 inside sales representatives to increase their account quotas by up to 30 percent and has delivered about 60 new paying customers in the past two months, he said. “The several thousand leads we’ve generated have more than paid for the service,” he added.

Charming, and with metrics, too

Integration with Fundbox’s Salesforce.com Inc. customer relationship management system enables Rachel to draw on individual prospect information such as account history, company name and location to sculpt a follow-up. It also ties into Fundbox’s marketing automation system to measure results, such as how quickly emails are opened, which is an important indicator of lead interest. Fundbox can track results on a dashboard that shows which leads have been referred by the intelligent agent.

“It takes a lot of the guesswork out of our salespeople’s hands by telling them which customers deserve more immediate attention,” Senesi said. “Our marketing department of four people does the work of 10.”

Conversica wasn’t productive out-of-the-box. It took about two days of training with sales operations specialists to teach the software how to speak to customers and to define the internal service level agreements that were expected. The engine can be continually tuned to conduct A/B testing of different subject lines or salutations, for example, so it gets better over time.

But applying AI to human interactions also requires a delicate touch. Fundbox has been sensitive to the risk of making customers feel that they been fooled, and so only uses Rachel for the initial follow-up. “We use it delicately because we know people have an aversion to dealing with robots,” he said. Occasionally, customers ask if Rachel is human, and staffers are coached to always be transparent. “We don’t shy away from owning up, and nine times in 10 the customer response is positive,” he said.

Fundbox’s Rachel may not meet the Turing test standard on all fronts, but it’s making a difference where it counts.

Image: Pixabay