ellipse
pattern
May 15, 2018

Top Pitfalls to Avoid with Your AI Deployment

[24]7.ai

Top Pitfalls to Avoid with Your AI Deployment

Ready to use AI to bring about big changes in your company? We outline the six most common pitfalls that we’ve seen companies encounter, either in the planning or execution stage of their AI initiatives. Knowing these pitfalls in advance will help you sidestep them all, and deliver an AI implementation that goes more smoothly and paves the way for ongoing progress.

Some of these pitfalls are granular in focus while others are more broad, but each one can slow you down, undermine your success, and cause unnecessary headaches that most companies can easily avoid with a little planning.

6 Costly Pitfalls to Avoid with Your AI Deployment

1. Not Enough Data
In the digital age, organizations have access to unprecedented amounts of customer data, but many aren’t using it to its full potential. This oversight is a troubling one for organizations ready to deploy AI, because the success of their deployment will hinge largely on the quality of data.

Data is the key to really understanding your customers and delivering the best experiences. A careful analysis of data will help you discover who your customers are and what motivates them. As well, it will allow you to anticipate their needs, simplify how they like to interact, and learn from previous interactions to make future experiences better. Collecting the right data is especially important for organizations preparing to deploy an AI-powered chatbot to improve their customer service experience, because what a chatbot can and can’t do is heavily dependent on data.

If you don’t have the right data sets being pulled and fed to your chatbot, it won’t learn the correct ways to respond to customers, and will end up causing more problems for your organization than you started with.

On the other hand, if you gather and input the right data, your chatbot will be equipped with the ability to understand interactions and intent and provide accurate responses and optimal experiences. For this to happen, there are three important types of data to utilize:

  • Customer profile data. This is standard customer identity data that every company should have. It includes things like name, demographic, geographic location, preferred channel, etc.
  • Interaction data. This is the ongoing record of past interactions and conversations with
  • each customer.
  • Relationship data. This includes previous social comments, feedback given, or information such as usage, share of wallet, etc.

Combining this data allows you to deliver personalized interactions for easier, faster, more complete resolutions, as well as more timely and relevant offers for higher conversion.

It’s also important to understand data collection and optimization will be an ongoing process. A trusted vendor can guide you through the process to ensure your AI deployment sets off on the right foot.

2. Trying to Fix Everything at Once
Bearing witness to this incredible period in history, where AI capabilities are transforming from vision to reality is exciting, but can be a double-edge sword. Eagerness to usher a whole new world of benefits into the customer experience leads to over-zealousness to transform digital self-service, the IVR experience, sales, and marketing all at once. It’s far better to zero in on one area, build success, and move on. Look at your most common or most important customer journeys to identify problem spots that can be improved with AI.

For example, consider a retail telecom provider suffering a high volume of calls from customers who tried to upgrade their smartphone online but ran into difficulty. This is an example of a journey that can be fixed.

An intelligent chatbot in the digital channel can be used to help match customers to the specific upgrade options that are available according to their unique plan. Such a chatbot eliminates confusion and helps the customer complete the transaction in a self-serve manner. This might reduce the journey from seven steps and two channels, down to four steps in one channel.

This is a relatively easy win to put in place and then apply the same scrutiny to analyzing and repairing other journeys in other channels. 

3. Lack of Vision for AI
AI advancements in the customer experience are moving too fast for you to take a reactive approach. You need to build a vision for AI. Look to the automobile industry for examples of how they build vision into their culture. Take Aston Martin, as just one example. Google the “Aston Martin Vulcan” or “Aston Martin Valkyrie” and you’ll see futuristic-looking vehicles that are not production-ready. But these models are important because they represent Aston Martin’s vision of the future of the company, the industry, and the customer experience. By creating these vehicles, customers, employees, and the market know where the company is going, and can engage with this vision now.

Similarly, you need to develop a vision for AI. How do you see AI impacting customer service? What about sales? What will be the main transformative differences in your customer experience five years from now, compared to today?
Without a vision, your company is stuck in wait-and-see mode where you constantly have to play catch up to bring your customer experience up to industry standards.

You don’t need your full vision in place today, but you do need to make it a priority. If you are short on vision due to lack of experience, insight, or expertise, then begin now to shore up these deficiencies with learning from your AI technology vendor partner. The right partner will help you build both a vision for the future and a strategy to get there.
 
4. Failing to Communicate on the Blending of AI and Agents
When AI is used to automate self-service interactions, agents are freed of the burden of these menial, repetitive interactions. This of course positively impacts the customer experience, your cost structure, and agent morale. However, even positive change can have negative repercussions if those impacted by the changes are not adequately prepared.
Be clear on how you see bots and agents working together in your company, and communicate this ahead of time to your agent population. When bots handle more simple interactions, what will your agents do? Do you have a plan to show agents how they’ll soon be delivering higher value to the company and also enjoy more engaging, fulfilling work?
Your agents will need to jettison old behaviors and adjust to new ways of serving customers, with the aid of AI. This includes learning how to engage at different points in the customer journey, handling more complex interactions, gathering rich, personalized journey data from their screen instead of from customers, and interpreting and acting on prescriptive guidance on resolution next steps. Proper communication on the big-picture of how and why you’re blending AI and agents will go a long way to get agents excited and eager about the changes to their role, and will also quash any morale-deflating assumptions that AI solutions are there to render agents obsolete. With a little forethought and planning along these lines, you can help agents transition into their new blended environment smoothly.

5. Failing to Sustain Internal Support
AI improvements are ongoing, which means it’s important to sustain support and momentum for your vision. The world of AI moves fast, and as you plan for the next enhancement and deployment, you don’t want to get six months down the road and find that enthusiasm for your ideas have waned. Go in knowing what improvements you expect. What predictable improvement will AI bring you? What metrics will you be measuring for success? AHT? NPS? Average Order Value? Cart abandonment rates? Have a plan in place for communicating this success. What is the process for socializing metrics internally? Does someone own this process? There’s nothing worse than getting some initial wins, then finding out that you need to go out and shore up support for AI all over again. Groom your internal stakeholders by building embers of support on the heels of your early results.

Your plan to solve big problems early will help you here. Recalling the retail telecom example earlier, the improvement to the customer’s phone purchase journey will improve satisfaction metrics and reduce escalations. This gives you measurable operational efficiency and cost savings on top of the improvement to the customer experience.
There’s no one who is not going to like these kind of holistic improvements brought about by your AI initiative, but results need to be formally communicated in order to sustain support.

6. Underestimating Vendor Roles
The pace of technology advancement has put a new kind of importance on vendor partnerships. Unfortunately, it’s common for companies that deploy a promising new technology to experience a drop in the excitement after results fail to live up to expectations. When this happens, it is often because an eager and supportive vendor disappeared after the sale. This speaks to the necessity for strong vendor relationships and commitment beyond the initial sale. Due diligence into vendor relationships, not just technological capabilities, is required and more important than ever. As part of your due diligence, ask vendors to clearly demonstrate how their accumulated in-the field experience and ongoing investment in AI technology drives benefit back to companies served. One example is the sharing of plug-and-play industry data to help get new AI programs up and running.

Also ask how they demonstrate commitment to results after implementation? Gain a deep understanding of post-implementation service, and explore outcome-based pricing models as ways to guarantee closely integrated support.

Feeling Unprepared? We Can Help!
If you’re feeling uncertain about your preparedness in any of these five areas, [24]7.ai is the right partner to help. Rely on our expertise in planning and execution of digital transformation programs to avoid the pitfalls that can derail your AI deployment. Contact us today to learn more.
 

Related Posts

Personality is Not Optional in Chatbot Design
Nov 26, 2020 Personality is Not Optional in Chatbot Design

“No response is a response. And it’…

Self-Serve Customer SatisFAQtion Starts with Natural Language Understanding