By
Jarrod Davis
|
Date Published: April 12, 2024 - Last Updated August 09, 2024
|
Comments
As technology marches forward, existing solutions are often caught off-guard by new developments that are outside their core competency and begin to cobble together a bunch of minimum viable products and acquisitions out of a belief they need to stay competitive across the board, versus focus on what they’re best at.
Yet, that leaves customers with a solution that does one thing quite well, and the rest often not. The vendor impulse is understandable and lord knows marketing materials make them sound like full-fledged AI solutions. But do you think a CCaaS or CRM vendor really spent the last 5 or 8 years dedicating their full attention and resources to AI?
While platform add-ons can be initially appealing, their chance of success is about the same as a long-time professional basketball player suddenly deciding to take up baseball. Here’s why bolt-on solutions will ultimately cost you more in terms of dollars and CSAT and what you should keep in mind when evaluating platforms for enterprise use cases.
What Specifically does “AI” Mean?
There’s a reason Artificial Intelligence is an umbrella term which requires a massive Wikipedia article: it’s not all the same. The same AI which powers bots in online games is not the same as natural language processing or Netflix’s recommendation engine. So, when we’re speaking of AI within the context of customer experience, we’re generally talking about one thing:
Conversational AI.
Conversational AI consists of two main components. Firstly, it’s powered by natural language processing and understanding. Simply put, that means receiving human language and understanding the intent behind it (e.g. “I want to return an item” or “I need to change my reservation). That is the AI part of it (NLU). The second component is process orchestration and automation based on what it understood (Natural Language Processing, NLP), i.e. the ability to get stuff done based on that user’s intent.
Imagine sitting at a table in your favorite restaurant. A waiter comes and listens to your order. That’s Natural Language Understanding. They understand English and what you are communicating (e.g. “I want X, Y and Z but no pickles”). But that alone won’t get you any food.
The waiter still needs to turn that understanding into action, i.e. organize and write all of it down, take it to the kitchen (which you can think of as an integrated enterprise system!) and then additional subprocesses must be initiated to prepare all the individual dishes, coordinating the timing of all of them, and notifying the waiter when finished.
And of course, the waiter needs to deliver the final product. That’s Conversational AI and that’s what’s needed by most CX professionals when they think of AI. Generative AI does play a role, but it’s not as big as you may think.
The Uncomfortable Questions or Beating a Dead Horse
I’m not here to tell you that some of the quick acquisitions or bolt-on AI features are all bad. For small and even some medium-sized companies, they may be perfectly adequate and the right purchase in the same way going to a big box store is a reasonable decision to furnish a student dorm of first apartment. The cost is right relative to your needs. But when you’re looking to take on thousands or tens of thousands of inquiries daily, sometimes even within an hour or two, using the word “AI” on your website doesn’t cut it for table stakes. Consider
- How many developers are dedicated for their AI solution?
- How long has their solution existed?
- How often do they ship new releases and how substantial are they?
- How big is their support and CSM team for AI?
- How many enterprise customers and case studies exist for their AI solution?
- How have analysts ranked or reviewed their AI capabilities specifically.
- Are they perhaps a Magic Quadrant Leader for CCaaS and not even on the map for AI?
- How much flexibility do they offer to mix and match 3rd party services?
- How many communication channels are supported and enterprise integrations?
The intent here is not to put anyone specific down, but to make clear, the devil is in the details when it comes to enterprise level AI solutions so when you’re an airline and need to handle say thousands of sessions a minute due to a snow storm, such as the one in Munich recently. As an example, Google Dialogflow can only handle 1200 requests per minute via chat whereas airlines will get several times that during a storm, strike or other major disruption.
At minimum viable product or even a basic option from a hyperscaler is probably not going to cut it. But again, this is why it pays to really understand your current and future requirements.
Finding the “Switzerland” of Solutions
As platforms such as CCaaS, CRMs, Digital Experience, Customer Engagement and others all scramble to offer AI solutions, the risk of market consolidation increases due to acquisitions, increased competition and a flood of funding.
Moreover, the same reasons also raise the risk of rapid changes meaning that many of the lesser established players, as well as specific technologies such as STT, TTS and even LLMs are changing so fast that within just several months, your previous best choice may quickly be overtaken. Hence, the importance of neutrality by design, a la Switzerland.
In order to mitigate the risks above, it is critical to look for vendors with open solution architectures and platform agnostic options. This is exactly the opposite of classic platform vendor lock-in.
Example features to consider are:
- How many LLM integrations are available (both vendors and specific models from each)
- Can you use your own LLM?
- Can you switch NLU models or are you stuck with one?
How many STT and TTS options are available? Can you choose a different vendor for each? What about different ones for each language?
Many platforms do offer a marketplace with integrations for multiple CRMs for example, but we’re talking about the nuts and bolts of the technology here. The leading TTS vendor for English may be abysmal at Spanish and French for example. So what happens when you are forced to use a single vendor overall, and the same for every language? Tough luck everyone else?
Some of these features likely won’t concern you during implementation and initial rollout, but when that day does come, you don’t want to look around and find yourself in platform jail .
Summing Up
As with so many things in life, choose what’s right for you. The problem is often figuring out the difference between what people are trying to sell you and what they’re actually selling you. Don’t settle for “let’s just take our existing platform’s solution which they’ll throw in for cheap” but put out an RFP and don’t be afraid to ask some hard questions.
The answers, or lack thereof will be telling. Today’s AI can really work wonders for your contact center and it’s must-have infrastructure today. So, all the more reason to do it right the first time.