How we MVP our AI startup to success - or The Dead of the Sales Man

Its been a long time since our last post. After a 15 months period filled with extensive user research and interviews, various smoke tests, projects with startups and SMEs, interruptions such as COVID-19 we are finally on our road to product-market fit. We continued with our low-fi tech approach to rapidly test various hypotheses and assumptions. And I want to thank my mentor John Sechrest who kept pushing me to run weekly experiments to find my product-market fit before we did the crazy stuff - building the "platform".

Our original hypothesis was that businesses need a solution to rapidly test and validate new business ideas to get from customer 1 to 1000 in less than six months.

Hypothesis 1: Businesses need a flexible dashboard to get instant and actionable insights on their growth initiative.

WRONG: There are already many flexible KPI dashboard solutions on the market but they are not helping businesses to run customer experiments.

Hypothesis 2: Businesses need help navigating through various customer experiments.

RIGHT, BUT: We built a triggered email process, that failed because email is an asynchronous communication process and emails ended up in a full inbox among many other emails. We built a Slack chatbot that engaged in one-to-one conversations which were more interactive. Our clients started learning but still did not run customer experiments.

Hypothesis 3: Businesses need a toolset to run customer experiments and growth marketing strategies.

RIGHT, BUT: We collected more than 30 different customer experiments from The Real Startup Book and Strategyzer's Testing Business Ideas, turned some of these into easy to swipe templates but even then, only a handful of companies took the initiative and ran their experiments. After a few user interviews, it turned out that most of them were looking for an automation that would relieve them of the manual work involved. Learning from our own experiments - you cannot automate customer interviews, you can only automate the process towards the interviews.

Hypothesis 4: Businesses need an easy to deploy a solution to automate their customer experiments and growth marketing strategies.

RIGHT, BUT: We figured that automating the customer experimentation process is very similar to automating a lead generation process. You just work towards a different set of KPI's. You need to find early adopters or customer subsegments to test your business idea or sell your product. One of the biggest challenges at the time was we needed to fill our own lead pipeline in order to get our own MRR rate up. So, we modeled and tested a lead gen process that would help us get a constant flow of leads through the door. The process worked to some extent but it was not perfect. What we figured out was that the biggest challenges in lead and sales automation is message personalization beyond the typical #firstname, #lastname and #companyname. The second challenge is pain level - when is a “prospect” in a pain stage and aware that he has a problem and looking for a solution.

The way we solved personalization is that we manually injected personalized content into outbound messages in order not to sound like a robot or an automated message. We see this kind of message every day and honestly, I don't pay attention to these anymore because there are just too many hitting my LinkedIn and email inbox. So, why should our own customers pay attention to "our" automated messages?

The pain level was easier to figure out because usually people self identify themselves if they are looking for a solution for their problem. So, you gotta solve the first problem in order to get to the second one. In our case, our pain level was the unpredictable sales pipeline.

Hypothesis 5: Businesses need an easy to deploy solution to automate their lead generation.

RIGHT, BUT: At some point, we turned on our human-assisted lead gen automation and it worked quite well. It worked so well, that our new clients asked us to implement our process for their businesses. The challenge here: our process worked for us because we built it around our needs and we had no idea how our process would work for our new clients such as DocuSign, T-Mobile or Charles Schwab. Lead generation and sales processes work completely differently in an enterprise or even an SME environment. But since we are rapid learners we came up with various solutions that helped our clients to solve their lead gen challenges (and our learnings will be part for a different post).

Hypotheses 6: Businesses need a lead gen automation and customer subsegments insights to focus on the right customer and speed up sales.

RIGHT, BUT: This is what we are working on now. We figured out how to automate lead gen processes and help companies to turn on their lead gen intake to increase their predictable sales. Using our Innovare AI reinforcement learning algorithm we can identify patterns in your customer subsegments and identify those with a certain pain level aka readiness to buy.

The biggest challenge that we are trying to solve now is personalization at scale. AI and automation can, of course, send rule-based canned messages but in order to find customers that are in the market looking for a solution you need to become an expert in their market, speak to them like a human and personalize messaging content along the way.

Again, it is not about the typical #firstname, #lastname and #companyname thingy - you need to take a look at your prospects’ public data and pull the important information together. This is where our human intelligence comes in. We hired a VA who completes data points of any prospect that we are reaching out to. The difference is an uptick in initial conversion rate from cold to warm prospect from 13% to 41%.

Unless we will have an AI solution available to collect that data and turn it into meaningful sentences we won't be able to fully automate lead gen processes - unless you are willing to sacrifice some of your high-ticket sales. If lead gen agencies and consulting firms are telling you otherwise they are lying to you.

PM me, if you want to learn more. I will send you all the information and blueprints you need to build your own human-assisted automation.