Industry

A 7-Step guide to implementing AI & Automation

No one wants to be in this situation, but we’ve all been there.

Have you ever tried to contact customer support to solve an issue, but you were pushed into using a decision tree - even though you knew that your problem was complex and could only be solved by speaking to a human.

Well, that’s what happened to me last month. Not only did I have to go through the process several times, but as I expected, it also never solved my problem. 

As a customer, it left me disappointed. I wasted time and my problem was still not solved. I was stuck at square one. Only now, from a customer service perspective, I was in a worse place. 

“For today’s consumers, who want instant gratification and prefer self-service, AI has lost its stigma. People are more comfortable than ever dealing with chatbots, but an ineffective experience does more harm than good”.

ADA 2019 CX Trends and Predictions

AI and automation are fantastic when used properly. But, if you implement customer-facing automation, you risk that these customers aren’t looked after as you would like, and they end up giving up before contacting support - and how will you know that they’re unhappy?

No matter whether you’re setting up customer-facing or agent-facing automation, you need to get it right for your customers.

Since everyone’s setup is different, you want to know what will work for you. So here’s a brief introduction to implementing AI and automation to ensure customer experience remains at the heart of your decision:

  1. Are we ready?

  2. Does it answer to a specific business goal?

  3. What ROI should I expect? 

  4. Research

  5. Get your team on side & walk before you can run

  6. Track everything & don’t give up 

  7. Don’t fool your customers

1) Are we ready? 

Some AI/Automation requires a certain level of customer service ‘maturity’. We don’t mean to be condescending, or refer to any sort of intelligence or professionalism! We’re actually referring to the amount of data your team has to work with. 

Simple automation can just be a few lines of code that you’d be able to implement yourself if you have the resources. More advanced automation, such as technologies that recommend responses to your agents or automatically categorize tickets, will need a solid set of historical  data

Using this example for automatic categorization, you would need at least 1,000 tickets with a high past 'categorization accuracy' to develop and train accurate models. 

The more complex your customers' requests are, the more data you need. So, the more complex your aspirations, the more accurate your data has to be! 

Plan Sticky Notes (Blog Image)

If you think you aren’t ready, consider the effect this will have on your customers.  

2) Does it answer to a specific business goal?

Any strategy that you implement is certain to be linked to KPIs. It’s no different here. When you’re considering automation, think about what is the central focus of your project. 

  • Lowering handling times? 

  • Increasing % self serve? 

  • Improving Customer Satisfaction?

There should be a direct link between the project and the business goal - and a clear strategy for measuring it. 

For example, if you are looking to lower handling times, you could look to use automation to supercharge your agents

If you want to increase deflection, you might consider implementing some sort of knowledge base automation.  

3) What ROI should I expect? 

Then, you need to think about ROI (Return on Investment). Before jumping to any decisions, the best thing you can do is try to test the automation/AI manually. By this we mean that you should replicate the behaviour of the potential project using your own resources.

Ben McCormack, who has worked for Trello and FullStory, writes: 

“For a project that requires a lot of time writing code, you might want to test it manually before investing the time to automate it. When doing the process manually, are you getting the value of it that you expected? If not, it doesn’t make sense to automate it”.

Secondly, he considers the trade-off between time and money: 

“If it costs $75/day to do it manually and $750 to automate it (so the subsequent process costs virtually $0 per day), you’ll get a return on your investment after 10 days and save around $20,000 per year going forward. If that’s more valuable than other potential ways to spend your time, it’s time to start coding”.

Read the full article here

Coding Blog Image

Going one step further, you should also consider the cost of upkeep. What resources will you need long-term to fix anything when it goes wrong or to update it as your process evolves?  

If you’re considering engaging with a vendor, you should consider the resources that are saved on your side by handing this responsibility over to them.

4) Research

Even if you are considering implementing these strategies yourself, attend as many demos and speak to as many vendors as you can. They will each offer unique insight into a specific area of expertise. 

They will ask you questions about your setup, the goal behind your project and about your expectations. They might ask you questions you hadn’t thought of before. This will really help you narrow your focus and have a clearer understanding of your strategy, as well as giving you the inspiration to enhance your current project or even fuel future strategies.

Use communities to inform your decision-making and hear from other people who have implemented similar processes. Find out first hand what the dos and don’ts are. The Support Driven Slack Community is an excellent example.

5) Get your team on side & walk before you can run

The final hurdle to implementation is getting your team on side. Communicate why this project is going ahead, what it will look like for your team and address any reservations that they have. For example, if it's a change that will impact your agents on a daily basis, you should hear them out. Understand what they’re worried about and reassure them. 

Once you get them excited about the project - and when they’re aware of timelines - it will make the transition and implementation much smoother. 

When you’ve made the decision to go ahead with the project, plan clear milestones for its implementation and set goals for the teams involved. If you are working with a vendor, take full advantage of all the touch points they offer you.

Manage your expectations. Early results may not be exactly what you had anticipated but there’s no reason to lose faith in this project so early on. Always look at what you can do to make the implementation process smoother or make time savings. 

6) Track results 

This is an obvious one, I’m sure it comes as no surprise to you!

It’s important that you measure your core KPIs from day one, so it’s well worth setting up a new dashboard or search for this project to monitor its progress too. This will give you a reference point of performance & enable you to gauge whether the improving results are due to this project or another factor. 

Check on this report/dashboard more regularly than you would on any other, at least for the first few weeks. Keeping a close eye on progress will allow you to pivot your strategy in the early days.

Share these results with your team and, if they aren’t quite what you expected, discuss internally (or with your partner) ways to build on the process. Don’t give up; seek out guidance from the community if things aren’t going completely to plan. 

7) Don’t fool your customers

If you’re considering implementing customer-facing automation, don’t try and trick your customers into thinking they’re speaking to a human. They’ll either

a) know that they’re speaking to a bot or,

b) they’ll find out a bit later (which is considerably worse).

Using this example, don’t try and dress up a chatbot as a human by giving it a real name.

When your customers are warned that they’re interacting with a bot and not a human, they’ll be more patient. How much more patient they’ll be, we can’t say. Even though people are getting more used to dealing with AI, it’s again worth considering that an ineffective interaction with a could well do "more harm than good".

Turn your customer service data into better customer experiences