5 Questions to Ask an AI Chatbot Provider
Despite so many benefits to deploying chatbots, many companies hesitate when the time comes to rolling one out. What holds them back? Not knowing where to start. With numerous providers on the market, picking the right one is a tough decision. To help make the process easier, we’ve put together this handy list of the top 5 questions to ask chatbot providers as you’re weighing your options.
1) How do you make your chatbot “intelligent”?
In simplest terms, chatbots behave intelligently because of how much data is inside them and how their software manipulates that data.
Here’s how chatbots get their data:
- Developers manually code or label data.
- Data is ingested without the need to manually code or label it .
Here’s how chatbots manipulate that data:
- Developers write rules-based code.
- An application program interface (API) uses complex algorithms.
Keep in mind that when a chatbot requires manually coded data and/or rules to manipulate data, it can get prohibitively expensive if you want to scale up to numerous use cases. If you’re interested in building a simple system with relatively few use cases, this might work for you.
A chatbot that uses an API that simply ingests the data can take you a lot further, a lot faster. Chatbots with natural language understanding and dialog capabilities can engage in multi-round conversations with your customers, mimicking actual human agents.
Once you understand how the chatbot provider ingests/integrates the data, you can follow that up by asking: “What data will you need to prove the efficacy of your chatbot vs. my existing customer experience?” This should give you a good idea of how quickly you’ll see results.
2) Will this chatbot get “smarter”?
Chatbots that require engineering resources to massage and code data on the backend can get smarter. But there’s a big caveat. You’ll need to pay for the man-hours to make it happen periodically, which can get quite costly. As you gather new data from your customers and want the chatbot to handle new scenarios, you’ll need to hire engineers and developers to build those use cases in on the backend by writing new rules and labeling more data.
In contrast, AI chatbots that rely solely on API-ingested data depend on sophisticated backend algorithms. With a much broader and fresher data pool at its disposal, and response-matching algorithms running behind the scenes, these chatbots can adaptively become more and more “intelligent” the longer they’re in place. Moreover, maintenance costs are kept to a minimum because all of this learning happens automatically. Deep learning is self-optimizing and automatically ingested by the AI on the backend. Once the system goes live, it continues to collect and ingest more information from conversational interactions, learning from them while growing out the dataset.
3) Is this chatbot scalable?
Chatbots that rely on coded and labeled data end up costing significant time and energy to scale up. Why? Most companies don’t want to spend the massive amount of time and money it would take to make the system function for every possible scenario. As a result, most only handle Tier 1 use cases. Keep in mind that after the chatbot is rolled out new scenarios are common, which means that new rules must be written, new data must be labeled, and old rules must be modified to deal with them. Every time this happens, engineers and developers (either yours or your vendors) need to be involved.
However, the newest high-tech chatbots, aka virtual assistants, can scale up to 80-90% of your use cases because your data is simply uploaded into the chatbot. Use cases are only limited by the amount of your organization’s data and its ability to be exposed via APIs. Once the system has been populated with your data, some rules are written to configure how the data is organized and categorized. Essentially, however, it’s just massaging the data around the edges. There’s no need to write 10,000 rules from scratch.
4) Will I have any granular control over the chatbot on the backend?
Chatbots that require engineering resources are basically an AI black box. Once all the backend coding and programming is done, customers will interact with the system and you won’t have any control or knowledge of what’s going on behind the scenes. The chatbot’s responses, as well as when it escalates to a live agent, have all been predetermined and are now locked away. It’s scary to place so much trust in this first point of contact with your customers. After all, your brand image is at stake.
In contrast, the most sophisticated chatbot on the market features a user-friendly management console. With it, CX owners can control real-time system responses according to internal best practices. It’s as easy as opening an easy-to-use interface and replacing the system’s current response with a new one. With this chatbot you can program multi-round conversations in the console, which ultimately means you can rollout more complex issue resolution around a much broader range of use cases.
5) What’s your pricing model?
Essentially, you have your choice of three pricing models for chatbots:
- Professional Services: Here you are paying a set hourly fee for the services of the vendor’s developer team as they manually code the rules and ongoing updates to the chatbot.
- SaaS subscription: This model is not tied to any qualitative metrics or system ROI. Sometimes, its pricing is based on the total chat volume.
- Enterprise license: You will negotiate an agreement ahead of time for an up-front annual licensing fee for the system. How much you pay will be based on how much the technology is worth to you, and how much the provider wants for its set-up and maintenance.
- Issue Resolution based: With this model, you’ll be paying for the actual resolved issue the chatbot resolves, tied directly to your call center analytics such as cost per contact. It is 100% ROI-based.
We’ve given you a lot to think about and ask here. For a handy guide to how we answer these questions, we’d invite you to download our solution brief below. (Or better still, schedule a demo to see why Bloomberg named us one of its 50 Most Promising Startups.)