Gear Up Your AI: Fine-Tuning LLMs
January 31, 2019

4 Steps You Should Take Before Investing in AI

Table of Contents:

With the advancements we’re now seeing in artificial intelligence (AI), it’s obvious why interest in the technology has skyrocketed. In fact, one recent Gartner study saw interest and investment increase by a staggering 270% since 2015. This year alone, 23% of companies are planning to invest in AI. Still, with the vast potential of this technology, where should companies even begin?

Below are the 4 key steps companies should take before investing in AI:

1. Define your business problem:

What problem do you want AI to solve? If you’re an online marketplace looking to moderate content, computer vision AI is perfect for your use-case. Want to improve your customer experience? CV AI can help you make better product recommendations for them, based on what they actually like, while chatbots let you offer round-the-clock customer service without having to hire more customer service representatives.

Still, AI cannot do everything. And even with the tasks it can do, it may not be worth your investment in it right now. As such, be sure to sit with key stakeholders, especially the people who will actually be working with AI, and figure out what or where you are trying to improve in your process. Take a look at the day to day operations of your company to see what tasks are recurring and repeatable. Defining these tasks will give you an idea of what is prime for automation.


2. Do your research:

Once you figure out your problem, it’s time to look into what kind of AI can help you fix it. For instance, I mentioned computer vision above, but this is just the umbrella term for a number of different applications allow computers to recognize text. However, the latter technology also gives computers the ability to understand it, and in turn, respond to it appropriately. This is the power behind chatbots and virtual assistants.


Identifying the kind of AI you’ll need is a critical step, but unless you are well-versed in AI, it can be challenging. Just how challenging this step is for you, however, can be a big help for the next step.


3. Decide whether to build, buy, or outsource:

If you’re unsure of what AI you’ll need, guidance from a team of experts that is solely focused on AI technologies can help you decide. Whether those experts should be in-house or third parties depends on a few factors, like the uniqueness of your use-case.


If your use-case is a critical differentiator for your business, having an in-house team of experts to not only quickly build a world-class AI platform but also maintain and continually improve it over time may be the right move for you. We’ll discuss the affordability of this in the last point, but it should be noted that long-term maintenance may be a challenge. One benefit of working with vendors is they have access to a data pool and knowledge graph, allowing their AI to improve continually. Using their platform, your AI will do the same with little to no effort on your part.

For most companies, the goal in implementing AI is to augment their employees, helping them to be more productive and efficient at what they already do. This will allow them to focus more on enhancing the differentiators your business already has.


HDC_Consumer_Features_Blog_3047_Snap_and_Search-1024x576However, implementing AI may also open up avenues for you, allowing you to create unique shopping experiences for your customers, like’s “snap and search” app and West Elm’s Pinterest Style Finder. It can also help to improve your data analyses, giving you a more comprehensive view of the available data (e.g., traffic patterns, shopping trends) and potentially identify new opportunities for your business (e.g., promotions.)

So, to conclude, unless your use-case is one that will become a key component of your business and really separate your offering from the competition, and you can afford to maintain and improve that AI continually, and pay for the necessary staff complement to do so, it's more cost-effective to buy.


4. Choose the right team or vendor:

If you decide to build, but don’t already have a team of AI experts in-house, you’ll obviously have to hire them. Unfortunately, since AI talent is in high demand, it’s relatively scarce, and you’ll be competing with tech giants and the AI startups that will soon overtake them in this arena to hire them. It’s not an impossible endeavor, of course, but as with most things, scarcity is driving up the cost.

If you decide to buy, you’ll have an array of vendors to choose from. Let’s say you were looking for a computer vision provider. While all CV AI vendors offer visual recognition, there are a few you can ask to decide which is right for you:

  • What can their technology do and not do?

What media formats can you use? How many concepts can the model/s recognize? How advanced are those concepts? Are there any domain- specialized models? Can you build your own model?

  • How accurate is it?

How were their models trained? Was the training data well-labeled and suitable for the concepts? Can you send feedback to make models more accurate? Are they continually improving their models?

  • How does the pricing add up?

Is it per-unit or per-tier? Do I even need to buy anything?

  • What kind of support, community, and integrations are available?

Who can you talk to if you need help? How easy is it you reach them?

The right AI company will answer all these questions and more.

As you look into investing in AI, remember to do your homework beforehand, to ensure you get the most bang for your buck and best product for your needs.

New call-to-action