4 Steps to Pilot AI within Enterprise

Companies large and small are scrambling to integrate artificial intelligence. Many are doing so in a vacuum or reactively after they’ve been approached by an AI startup or service provider. Both methods usually result in costly lessons, and in the worst cases, companies decide that AI isn’t right for them.

If you work for a company that generates more than $50mm in annual revenue or more, you’ve probably been prospected by companies large and small as a potential customer for their “AI solution.” Usually the larger companies (H2O.ai, Hyperscience, Clarifai) in this space that are solving broad problems already have integration experience and can guide you through working with them. You can make reference calls and see results of previous work, but you have no such luck with startups. That shouldn’t deter you. Startups are solving more specific problems that companies in your industry face, and passing on an opportunity to work with a leading startup could come at a high opportunity cost if they’re also pursuing your competitors.

Instead, take the following four steps to ensure that you have a successful pilot that translates to a mutually beneficial relationship.

1. Identify the goal. Consider how it translates to revenue.

Far too often a startup will prospect your company before you have the opportunity to develop an AI strategy. This isn’t unusual. There are many B2B AI startups that bought prospect lists with your company’s name on it. It’s not too late to develop an AI strategy then, but companies need to be proactive and consider how they can use AI and why.

I’ll have a separate post about a developing an AI strategy, but today let’s consider a single variable for such a strategy: how it translates to revenue. Assuming that you have enough data when developing the AI strategy, decide on what will be monetized and how. AI products are far and wide. In many cases, they can lead to great outcomes, but it’s meaningless if it doesn’t deliver short- or long-term value. If you’re in retail, how do you sell more clothes? The answer could be using a company like FindMine to cross-sell other items as part of the startup’s “complete the look.” If you’re an endoscopy company with a vault full of ear scans, Clarifai can help you spot problems using trained models with 99% accuracy, a clear competitive advantage.

The goal is to make sure that whatever you are building internally or with a pilot, it should generate revenue, have a savings greater than its cost, or create a competitive advantage. If it doesn’t, it will become the digital equivalent of a paperweight.

2. Is the tech real, and is it translating to economic benefit?

We’re reminded on a regular basis of the delta between real companies and those that are riding the AI hype wave. To ensure we don’t fall into the AI hype trap, we have a great technical due diligence team made up of NYU researchers, Future Labs portfolio companies, and mentors that conduct technical due diligence on every team we accept. While tedious, employing a tech due diligence strategy will help a company make sure that what founder’s say the tech does, it actually does in a real world environment, especially in the context of achieving your goal. AI solutions can sound really good in theory, but it has to translate into sales or savings.

3. Build an internal approval and implementation process

Let’s start with the end and go backwards: Any AI strategy needs to fall under the CIO or whoever is the equivalent in your company. They need to know how the tech fits in with concurrent projects, what data can be used, and how the feedback loop is closed.

Your CIO needs to understand and be up to date with artificial intelligence. I can’t stress this enough. A CIO without a basic understanding of AI is like a salesperson without sales experience.

The implementation process is paramount. Paid solutions that aren’t deployed waste money, time, and, most importantly, are a huge opportunity cost, but you’d be surprised by how many companies pay for AI solutions that sit in limbo between signed contract and deployment.

The internal approval process for AI is just a little bit different in one way. The CIO should, in most cases, make the final decision, as long as she understands the AI being addressed. Again, with that as the endpoint, develop a funnel with the various stakeholders necessary for moving the process forward. Enterprise sales is a months-long process because of inefficiencies, so set a process and stick to it.

4. Set a range of expectations and debrief about the effort’s success or failure.

2017 is the year of high expectations for AI. Unimaginable progress will be made across a broad spectrum of industries, but many expectations won’t be met. Take that same mentality with any AI strategy you craft. It’s early days yet, and we’re still learning, so an unsuccessful pilot doesn’t mean that your company shouldn’t have an AI strategy. You many need to rethink the problem, scope, intended outcome, or timing. Debrief after every pilot or project hits a milestone and adjust expectations accordingly.

Now learn how to bring an AI venture from pilot to scale on May 31 when we host our next Focus | AI event. Industry leaders and investors Alex Poon, Michael Bishop, and Paul Strachman join us. Use this link for a free ticket.

We hope to see you there!

Steve Kuyan
Managing Director, Future Labs at NYU Tandon