Ditch the AI strategy

We’ve heard from several clients that they are afraid to miss getting on the AI train. They look at their competitors and think they are all further ahead, that they already have reaped the benefits of AI implementation projects. That's why many in the industry stress about creating a specific AI strategy. But this is putting the cart before the horse.

You should not need a specific AI strategy. AI is a tool, a solution, and unless it is core to your business offering, it is not an AI strategy you need. You need a strategy. AI can be part of it, but it should not be a strategy for how to implement AI. It should be a strategy for how to achieve business goals, meet competition or find a new product market fit.

AI can help you, sometimes. But having a separate strategy for AI will only help you find problems to a predefined solution.

Just as with any bet, implementing a new technology comes with a risk. At Kraftvaerk we conduct a discovery process to uncover the risk appetite, and how big of a bet your company is willing to make on this new technology. These should be weighed against possible gains. Starting small will give you small gains, but the risks will also be small, and you will be able to learn, and reduce the risk of future investments.

So instead of making elaborate AI strategies, what should you do? This is what we reccommend...

Just as with any piece of software, it makes a lot of sense to spend time figuring out what you want to build, and why, before you start spending time and money on an AI implementation.

Spending a few weeks discovering the needs of the organization, the constraints, the current product landscape and understanding your competition is a great investment. Not only will it help you make a good decision now, but it will also help you when the AI landscape changes (and it inevitably will).

Invest in workshops with the teams that you are going to serve. Understand their workflows and how AI could support it. Take AI out of the equation. AI implementations are solutions, and once we have a solution in mind our cognitive biases drive us towards finding problems that fit.

Instead, be curious about some of the challenges and pains people in your organization struggle with. Talk to stakeholders about their needs. Talk to legal and compliance about constraints. Talk to IT about your current partnerships. Look at prices of services and available products in the market. Document all your findings and define where to start.

A discovery process will help you understand the needs and constraints of the organization.

Our suggestion is to start small with existing services. Consider what the smallest investment you can make right now is to solve the problem for your organization. Find a solution that covers most of the needs and is within the constraints you documented during the discovery.

There is no need to start with a solution that lives up to all requirements. Instead, choose something that will get you started quick. The quicker you can get to a proof of concept the better, understanding if you are on the right path. Start by trying to find a solution that works. Only once you have proven this, it is time to care about scaling, long-term costs and partnerships. However, make sure that your experiments are purposeful. Document them and keep testing if they solve the problem. Try to align the initiatives that likely will happen bottoms up in your organization when people start signing up for AI service offerings. See if you can support the initiative by providing a compliant platform for people to use.

While doing this it is also important to avoid vendor lock-in. If you start small, with managed services, you want to have the ability to pivot your initiatives towards building your own, more custom solutions. Make sure that decision will be easy to make by avoiding very specific vendor services or locking your data, or other assets, on a single platform. 

In our experience, it is best to start by buying a fully managed, off-the-shelf solution that covers most, but not all, of what you need. Find one that lives up to the constraints you have. It is easier to add features (make an AI aware of company specific data for example) than it is to adhere to constraints (like data privacy) as the solution evolves. 

Maybe you have already experimented with some off-the-shelf solutions and are now ready to build your own solution to harvest more value – in that case we can help you get from POC to production.  
 
Or maybe you have ideas of where to apply AI but haven't gotten the overview of where to start, then we can help you get started with Data and AI blueprint service.