If you’re an enterprise IT leader, you’re probably feeling the crunch when it comes to artificial intelligence. You know you could be using AI to gain more insights from your data and gain a competitive advantage. Increasingly you’re hearing that you’ll be left behind if you don’t get on board the AI bandwagon.

According to a recent Gartner survey, 37% of CIOs worldwide have already implemented AI or will do so in the next year. They predict this number will grow exponentially, with over 70% of companies expected to integrate AI for productivity improvements by 2021.

There’s no question that forward-thinking leaders need to be planning for AI projects in 2019. But what approach should you take in order to move forward?

PLANNING OUT YOUR AI PROJECT – SHOULD YOU BUILD, BUY OR PARTNER?

‍Build if…

• You already have a strong team of data scientists
• Building your own AI solution is a competitive advantage for your company

Buy if…

• you can find an off-the-shelf application that can be easily customized to suit your needs

Partner if…

• you don’t have strong data science skills in-house, or you need to augment your existing team
• off-the-shelf applications aren’t appropriate for your use case or would require too much customization

WHEN TO BUILD

A company will generally build their own AI solutions in-house if the solution provides a competitive advantage for them, or if it’s important to be able to claim an in-house AI capacity for competitive reasons. For instance, a recruiting software company that uses AI to optimize hiring decisions will probably want to develop this capability in house.

One advantage of this option is that an in-house AI team will be already familiar with a company’s data and infrastructure, so there is no need to get a third-party up to speed. However, AI solutions built in-house generally take longer to build and deploy, as the company isn’t able to take advantage of the prior learning that an external vendor would have already experienced. In-house AI teams also may not have access to the third-party data sets that an external vendor has compiled.

This option is really only viable for companies that already have a strong in-house team of data scientists. In reality, this is not true for most companies. A recent Gartner survey asked enterprises how they would characterize the extent to which their organization had the qualities necessary to mine and exploit their data. Seventy percent of respondents indicated their company has limited or no capability in this area.

There’s no question it’s challenging to find, hire, and retain top AI talent, and data scientists are in high demand. Keeping an AI team on staff requires a lot of overhead, making this option a very expensive one.

WHEN TO BUY

Buying a pre-built AI solution may be an attractive – and cost effective – option if you can find software that does exactly what you need. But an AI solution is unlikely to work off-the-shelf unless it represents a very narrow use case. It’s more likely that you’ll need to modify the software to meet your needs. These customization costs could add up quickly. You may also be paying for functionality you don’t need or want, because it’s included in a pre-built package.

Because each company’s data is so different, it is very difficult to find a plug-and-play AI solution that will meet your exact requirements.

WHEN TO PARTNER

Partnering with a third-party AI company to build a customized solution is a way to get the best of both worlds. You get an AI solution that is customized to your own data and requirements, without carrying the overhead of your own in-house AI team, making this a more cost-effective option than building it yourself.

By partnering with an AI company, you’re able to take advantage of what the vendor has learned from other similar AI projects so your team doesn’t have to start from scratch. You’re also able to leverage third-party data in conjunction with your own data to better train machine learning models. You can rely on the vendor for updates and maintenance, so your team doesn’t have to take time away from other tasks.

Even companies with their own data science team may find this approach attractive for developing specific projects. For instance, if your data scientists are focused on building AI into your core product offering, they may not have time to build AI solutions to improve business processes. Working with a third-party AI company on these operational projects is a way to efficiently augment your existing team.

WHERE TO BEGIN

The first step with any project – AI or otherwise – is ensuring a good understanding of the problem you’re hoping to solve. What are your objectives in creating an AI solution? What other requirements do you have?

Then take a look at the resources you have to put towards the solution. The most obvious resources to consider are people and money – but what about your data? How much do you have and what format is it in? For most companies, up to 80% of their data is considered ‘unstructured’, i.e. not stored in a traditional database. This type of data can’t be used for machine learning models without being processed into a structured format first. This may need to be your first step towards your ultimate AI solution.

NEED SOME HELP?

If you’d like help scoping out your next AI project, the team at Primal AI would love to explore this with you! Reach out at hello@primal.com to speak with our data scientists.

Editor's Note: This blog was originally published on May 28, 2019. It was updated on September 4, 2024.