Building a Custom AI Solution? Choose the Partner as Carefully as You Would Choose an Off-The-Shelf Software
We have a story for you…
We know a company who bought an AI solution for their incoming order triaging. It was supposed to sit between Zendesk and their ERP and actually place the “easy” orders, follow up to get specific questions answered for the ones with moderate complexity, and escalate the complex ones and/or anything that could not be specified in a timely manner with the AI back-and-forth.
The website for this solution touts that it will reduce manual entry for orders by over 85% and it will more than 3x increase order processing speeds.
This sounded GREAT to the company we talked to – and they enthusiastically signed up for the service.
Only to learn, months later, that their people were still manually handling 89% of their orders, and average order processing times had actually INCREASED twofold.
So what happened?
The company brought up these issues to their sales rep at the AI solution company. Apparently there was a service “upgrade” that the AI solution company “forgot” to provide to this company. They were now several iterations behind the latest and greatest version of their algorithms. And when they asked for the upgrade, they were told that they needed to pay for the changes and new training because it was so different than what they bought in the first place. WHAT?
Needless to say, the company is now talking to US about discerning a NEW vendor to do what the original vendor promised.
The Promise of AI
The excitement around AI is understandable.
Companies can now automate decisions, connect disconnected information, reduce manual work, and build tools around processes that were previously too specialized for traditional software. In some cases, a custom-built AI solution may solve a problem more effectively than an off-the-shelf system ever could.
But the technology’s potential does not make the partner decision less important.
It makes it more important.
Companies often spend months carefully evaluating an ERP or other off-the-shelf system like a payroll application, Customer Relationship Management too (CRM) or Human Relations Information System (HRIS). They document requirements, compare vendors, review demonstrations, call references, examine implementation approaches, and consider the long-term cost of ownership.
A custom AI solution deserves the same level of discernment.
Possibly more.
Custom Does Not Automatically Mean Better
A custom solution can be designed around your business rather than forcing your business into a predefined system. That can be an enormous advantage.
It can also create an expensive, highly customized application that only one development company understands.
The wrong partner may build exactly what you asked for—even when what you asked for was not what the business actually needed.
Or worse! They may PROMISE to build something, but they actually don’t know your industry or your needs, and it ends up not doing what is needed at all.
Or you are their guinea pig and they improve in the future, but don’t upgrade you along the way (as happened to the company we shared about earlier in this article).
For these reasons it is important to be discerning about your AI Solution, custom or not, just like you would when you evaluate any off-the-shelf system.
A partner that is AGNOSTIC to the technical solution – yes, like us at AJC – can help you determine:
Whether the problem is worth solving
Whether AI is truly necessary
Whether the process should be improved before it is automated
Whether an existing platform could solve most of the need
Whether a smaller pilot could validate the opportunity before a major investment
Whether any particular developer or software is QUALIFIED to do the work in the first place.
The best developer is not the one most eager to build.
It is the one willing to be candid about whether they can meet your needs or not.
