Blog Viewer

Can Companies Implement AI In-House? An Expert Weighs In


Taking AI in-house is an option many businesses are exploring with the hopes to maintain full control over their technological infrastructure. But is it smart to implement homegrown generative AI solutions without the ongoing assistance and knowledge of seasoned experts?

The decision to build out homegrown AI is a lofty one — bringing with it enormous expenses, technical implementation challenges, and employee frustrations. To find out the true cost of creating in-house AI infrastructure, we tapped Gladly AI Training Manager Spencer Parsons.

Why In-House AI Is on the Rise

AI tools are becoming an integral element of many businesses’ daily processes. In-house AI allows businesses to tailor the technology to their unique needs and experiences, while also not having to involve an external partner for licensing and support.

However, in practice, in-house AI lacks an expert touch, and not having a seasoned partner involved can be costly in the long term. While in-house teams can manage small-scale AI, they often fall short of the expertise and capabilities provided by specialized third-party solutions for full-scale AI infrastructure.

A Q&A with Gladly AI Training Manager, Spencer Parsons

Attempting in-house AI is a major undertaking for IT teams and C-suites — and requires an enormous amount of technical nuance to consider. To delve into the intricacies of this practice, we consulted our in-house expert, AI Training Manager Spencer Parsons.

1. How Successful Are Businesses That Rely Solely on In-House AI, Particularly for Complex Processes Like CX?

SP: To date, I’ve not heard of an e-commerce company that has created a successful major in-house AI solution. A company may be able to implement small workarounds or processes using more accessible AI technology, such as ChatGPT, but creating and managing a large-scale in-house AI solution is a full-time job for a full team of people. In my estimation, most small companies don’t have the resources for it and most big companies won’t like the cost and time involved.

A company like Thankful was founded on that very basis — AI is hard, and people will need real help if they want to implement solutions that will improve agent output, reduce hiring tension, and save them money right from the get go.

The Gladly Takeaway: While small applications are easier to manage, a full-blown in-house solution is out of reach for the vast majority of businesses — and altogether unnecessary with options for external solutions currently on the market.

Read More: Gladly’s Acquisition of Thankful and New Platform Launch

2. How Complex Is the Process of Implementing AI In-House?

SP: Understanding your needs in an AI solution is probably the hardest part. This isn’t a situation where you can just throw things against the wall to see if it sticks. You have to be aiming at a clear target. You’ll need to consult with employees and customers to understand your goal. These processes take time and patience.

The tech resources that support modern AI technology are particular, complex, and nuanced to each use case. Unless you have that clear goal, the layers of a project like this can become overly complicated quickly. AI is in an ever-changing landscape, and keeping up with the latest technologies can be a full-time job.

The Gladly Takeaway: Those with a vision for their infrastructure have better odds at successfully implementing their own AI, but the majority of businesses still wouldn’t have the ability to manage the complexities of the process.

3. What Do a Lot of Businesses Get Wrong About the Scope of This Endeavor? What Is the Biggest Risk Involved?

SP: Nearly every aspect of an AI implementation is time-consuming and costly. It will take years to be able to truly integrate an AI solution. Either way, you’re going to need to hire experts to help you with your AI solution and in-house development can be expensive. From a very high-level perspective:

  • Year 1 – R&D – $250,000 minimum
  • Year 1 to 2 – Build MVP – $250,000 minimum
  • Year 2 to 3 – Train AI and QA Testing – $250,000 minimum
  • Year 2 to 3 – Full version built and implementation – $250,000 minimum
  • Ongoing annual cost for maintenance plus future upgrades – $250,000 minimum

You’ll likely spend around $1 million (or more) in the first two or three years of getting the solution to the point of true implementation. That’s banking on nearly everything going right, which we all know is rarely ever the case in tech. If it’s the solution you’d hoped for, your company and customers appreciate it, and they’ve bought in as hoped, you’re then looking at no less than $250,000 annually to keep it trained, maintained, and up-to-date with current tech.

The thing about an in-house AI solution is it’s not something you can ever really take your hands off of. Particularly now while it’s still a highly evolving technology. If you aren’t willing to do the ongoing work to keep your system in top shape, you will fall behind.

The Gladly Takeaway: In-house AI will drain your resources before it provides any ROI. This can undercut your growth, whereas external solutions are designed to scale affordably and sustainably with your business.

4. How Many More Technological Resources Do Businesses Need to Be Prepared for These Integrations?

SP: This depends very much on what kind of business you are and what you end up implementing, but with any major technology transition within a company comes new resources and new costs. The resources required to run new AI technology are in high demand, which means prices undoubtedly will be high as well. Between servers, storage, and a platform to build and run your AI on, you’re looking at implementing no less than three to four potentially very costly new tech resources.

The Gladly Takeaway: The development and implementation process of new AI is a massive undertaking — instead, look for an external solution that’s proven to provide easy implementation and use for IT and agents alike.

5. Is There a Balance of External and In-House AI That’s the Best Bet for Businesses?

SP: I don’t think so. I would tell any business owner, if there’s already an AI solution out there for what your company needs, you’re much better off getting in early with that company and riding the AI wave with them. The earlier you can get in, the more access you’ll have to provide input on how the solution evolves and works for you. In my opinion, partnership has a much better chance of a high ROI.

The Gladly Takeaway: Unless AI is already your bread and butter, there’s going to be a partner-provided solution out there that will accomplish more, and much faster, for your brand.

6. Will In-House AI Ever Surpass the Expertise and Hands-on Support of an External Solution and Partner?

SP: Definitely not. The cost and complexity of in-house are too great, and the AI market is continuously expanding to fit more and more use cases. Unless you want to be in the AI business, investing in an in-house solution just doesn’t make sense.

The Gladly Takeaway: Not only is external AI more strongly advised, but the possibilities are only growing stronger. In CX, this means agents get the benefit of using a handy AI assistant that provides suggested responses based on customer data or even takes notes and creates conversation summaries in a flash.

The Best Approach for an AI-Powered Future

As Parsons has made abundantly clear, attempting a full in-house AI solution is daunting and unproductive, to say the least.

Luckily, the AI solutions to transform your CX infrastructure don’t require a heavy in-house lift. Instead, powerful solutions like Hero AI and Sidekick from Gladly provide tandem solutions for agents to serve customers better, and for those customers to solve issues more easily on their own. Take a closer look at what makes these white-glove external options the best choice for your business solutions.