What should I do about AI?

There’s a lot of interest, and hype, about using AI in support. I presented a webinar with TSANet on 3/26/2024 on this topic and you can watch the recording here. This post suggests a structure for organizing the vast world of AI projects and tools in the support arena, and a roadmap for using AI in support organizations.

Four Areas to Explore for AI in Support

• Prevention. AI can help with automated diagnostics and self-healing, traditionally driven by triggers and rules that require a lot of manual updates. The idea is that AI can detect deeper patterns than humans, and also allow instant updates to the patterns.
• Self-service. Search and chatbots have long been automated, but rather badly with keywords, rules, and triggers. The newer commercial AI tools are flexible and much more accurate.
• Copilots for support agents and support managers. This include communication assistants (writing or editing customer communication), automated case assignments, case summaries, sentiment and escalation detection–all functions that are traditionally done manually but turn out to be much easier with a little help from AI. Commercial tools are starting to deliver robust functionality in this area.
• Strategy. This is the most exciting area for me, although the least developed, at least for commercial AI tools. The idea here is that we can use the vast amount of support data (self-service and assisted, and perhaps data gathered by CSMs, too) to analyze how we can improve products and processes to make a meaningful difference in the customer experience.

A Roadmap to Implementation

1. Don’t reinvent the wheel. If you can use a commercial tool, do it! There are many vendors who offer working solutions and it makes little sense to go homegrown unless you have to. If you are looking for search engines, chatbots, or sentiment analysis, to cite just a few examples, you will likely find what you need on the market. Commercial solutions are improving very fast so don’t be worried about missing out if you decide against DIY.
2. Start with internal projects. If something goes wrong, it’s much easier to contain the damage out of sight of customers. AI projects can and do fail: fail in private!
3. Anticipate staffing issues. Support organizations are reporting that skilled resources capable of running AI projects are in high demand and may well depart once they acquire relevant experience, so think about how you will staff your projects in the long run.
4. Coordinate with other departments. Avoid overlaps and political battles by liaising with IT and Engineering as you get started. This will avoid conflicts down the line and boost available resources.
5. Protect data privacy. If you are considering moving customer data to third-party repositories, consult your legal team so you don’t run afoul of privacy regulations.

And I’ll add one more: if you think that you’d rather work with a partner, you can find one. The FT Works team can help, among others.

What are you doing with AI tools at this point? Please share in a comment.