Feels like you missed the generative AI train? 5 steps for speeding ahead in 90 days

I’ve been talking to founders across the Global South about generative AI (GAI) as often as I can since early 2023. The founders in our portfolio of 350+ companies are generative AI users, not creators. As with any other disruptive situation, these founders can be divided into three groups:

  • Ahead of the Curve: companies that have already shipped something.
  • Fast Followers: watching and prototyping but have not shipped yet.
  • Late for the Train: don’t yet know how to get on the train/don’t have any resources to apply now.

This article is for any founder who feels like they’re late for the train — or is all aboard, but not going fast enough.

Reviewing examples of all three groups will help founders know where they really stand. Those who are Ahead of the Curve had at least three things going for them: They saw the opportunity early, they had ready-made situations to which they could apply generative AI, and they had engineering talent available to get something prototyped and into production in a timely way.

One example is a farming e-commerce company that has already taken 30% out of its customer service costs by putting a farmer-lingo-capable chatbot in front of its customer service agents and expects to get savings to 50% over the next quarter or so.

A Fast Follower has prototyped means to cut costs and increase the speed of recruiting blue-collar workers by adding generative AI–driven steps to its interview and candidate engagement workflow. Because they have a complex workflow with high throughput, they must be careful about how quickly they deploy; initial testing is showing massive improvements in multiple dimensions.

Here are five clear steps to move from being late for the train to speeding ahead in much less time than you’d think.

Finally, a Late for the Train startup provides solutions for call centers and has done some initial evaluation and planning, but has not yet determined how/when to best add generative AI to its product roadmap, which is already stressed with demands from existing customers.

Here are five clear steps to move from being late for the train to speeding ahead in much less time than you’d think:

  1. Adopt a simple language so everyone can communicate clearly about this disruptive tech.
  2. Get your entire team onboard at the high level (many of them may already be there without your knowledge).
  3. Ensure that you are not letting cloud LLMs “hoover up” your data in ways that expose it to competitors or bad actors.
  4. Establish a Red Team to be disruptive internally.
  5. Measure progress on generative AI adoption and communicate it to the company on a consistent basis.

1. Type 1 and Type 2 generative AI applications

There are plenty of new technical words and concepts around AI, and many have written about them, so you don’t need more from me, except this one concept: From an adoption perspective, there are broadly two paths you can be going down, which are not in any way exclusive.

The first is using generative AI to enhance what you’re already doing by increasing productivity or quality of operations or existing customer interactions. Let’s call this a Type 1 application.

The Ahead of the Curve example cited above is Type 1: Companies using generative AI to improve sales communications or help with market research are doing Type 1 work. Type 1 projects can be implemented on an individual or departmental level. And most importantly, they are table stakes for every startup these days — must-do activities. If you want to get funded and can’t show clear adoption of Type 1 applications, you’re in trouble. But Type 1 initiatives alone will not make you an AI company from a VC perspective.

Type 2 efforts are bigger, riskier, and much more important to your survival and to your ability to attract capital. With Type 2, you are looking to create entirely new ways of approaching a vital aspect of your business, or potentially your entire business, building on generative AI.

The upside from Type 1 is a reduction in cost and increased speed/productivity — everyone is doing or will soon be doing these. The upside from Type 2 is potentially unlimited, as you are creating new ways to create and deliver value that might get you access to new customers or gain substantial competitive advantages over others who are not deeply embracing generative AI.

An example of a Type 2 innovation might be a regional B2B marketplace that currently publishes information only in English as it’s the common denominator language in the region. That marketplace now can use generative AI to cost-effectively publish information simultaneously in four local languages and enable its customers to find products/services with a conversational interface (rather than cumbersome search queries and complex filtering) using their language of choice. This Type 2 innovation opens the market to untold numbers of non–English speaking customers and also makes it faster for all customers to find what they are looking for and close the deal.

Another Type 2 example would be a gig worker aggregator who empowers their giggers with generative AI to accomplish tasks in entirely new use cases for new customers that they never could have served before. Giggers got smarter by the “augmented intelligence” of generative AI; they and their aggregator both make more money. If you are looking for funding and don’t have one or more Type 2 efforts underway, you will have a difficult time raising money. That’s as of July 2023, only seven months after ChatGPT changed the world. The expectations of investors are increasing as quickly as technology advances.

2. Getting your entire team on board at a high level

The best way to not miss the train is to get everyone on it, even if you don’t know exactly where it will take you yet. Have an all-hands meeting and announce your high-level intent, and be clear that you want everyone to be experimenting and learning (with the important caveats in #3 below), every day. Get them focused on Type 1 adoption, being clear that NOT applying Type 1 strategies everywhere possible will lead to the company losing its competitive edge. Also, tell them how you will approach Type 2 and how vital that path is to your success. You could raise awareness in all areas with department-level brainstorming sessions, starting in the middle management of your team, or by asking ad hoc teams to form and do the same. I’ve also seen some companies doing weekend “hackathons,” and others are doing team meetings with fun games to help people understand the value of GAI.

3. Don’t let cloud LLMs hoover your data

I recently wrote about the dangers of cloud GAI providers “hoovering up” your data and using it to train their AI for the benefit of all, including your competitors. In worst-case scenarios, your employees could inadvertently expose personally identifiable information of your customers to the GAI, which could end up being available to bad actors. You don’t want either of these things to happen. It’s not too hard to avoid them — please read my article and use Bing/ChatGPT4 (or maybe Google) to find other options that could have been published more recently, and act on it ASAP!

4. Establish a Red Team to be disruptive internally

I worked at Sun Microsystems in the ’80s during its first intense growth phase. One lesson I never forgot from founding CEO Scott McNealy was that it’s better to disrupt your own business than to let your competitors do it for you. A few years later, Clayton Christensen famously described this problem in “The Innovator’s Dilemma.” The book has seen some years but is plenty relevant in any time of rapid technological disruption.

In this phase of disruption, my strong recommendation to founders — which I’m pleased to say a number have taken and already shown great results from — is to form a “Red Team” charged with prototyping a new business that has the goal of taking away the majority of your customers in less than a year by using a combination of Type 1 and Type 2 generative AI strategies from the ground up to create offerings that are simply better/faster/cheaper/more profitable than what you do today.

The Red Team has the benefit of full knowledge of all the good, bad and ugly at your company, no legacy baggage of any kind, and freedom to operate enabled by the piles of VC funding that are coming into AI-forward companies every day.

I recommend three people on the Red Team: an up-and-coming young-ish developer who has shown some propensity to dive deep into generative AI already on their own; a data scientist or strong data analyst who can [safely] access all of your most valuable data; and a business development/marketing leader who can help chart the disruptive business strategy alongside the technical members. That’s it.

One of our portfolio companies formed four Red Teams, each charged with disrupting a different aspect of their business. How long should this take? The first report to your executive staff should be in less than 60 days from when you start, and you should plan to have results presented to the company no later than 30 days after then.

Google’s Sundar Pichai allegedly declared a “code red” after the OpenAI ChatGPT launch in November 2022. One might presume that if they had a few well-staffed Red Teams looking at disrupting their cash-cow search business with the very tech they themselves pioneered in 2017, they might have avoided being caught so flat-footed.

This leads to a final point. Your Red Team should help you see the opportunity to get ahead via GAI, but you will need to seriously evaluate if you can maintain strong market differentiation in this new world, and if not, how do you further modify your plans and create new innovation moats that will give you long-term room to grow ahead of fast-moving competition.

5. Measure and report progress on generative AI adoption

You should apply the principle of “if you can’t measure it, why bother” to your generative AI initiatives just like everything else. Since adoption is a journey, you want to measure both activities — for example, the percentage of the team who reports using generative AI tools to increase productivity at least three times a week, and results, such as savings in customer service expenses, increase in conversion rates, etc.

Pick a small number of KPIs in these areas (not more than three at the top corporate level), do a baseline measurement, and then report at the company level at least quarterly. At team levels, you might look for weekly reporting in the early days to ensure you are seeing trajectories heading in the right direction.

The Nozomi is Japan’s fastest bullet train, reaching speeds of 186 miles per hour (300 kph). It was disruptive when it started over 30 years ago, and it’s still among the fastest in the world.

If you are feeling late for the train, or like you are not moving fast enough, follow the five steps above. In less than 90 days, you will be transporting your team and company to a different world as your competitors watch you speed by.