Crawl, Walk, Run: How to Successfully Implement Generative AI

Crawl. Walk. Run.

It's typically the best way to implement big changes in an organization.

I've talked with many people who have said to me (in so many words), "There's no way in the world I'd let an AI interact in real-time with my customers."

I completely understand the sentiment. You're not alone.

Consider the crawl-walk-run story of autonomous vehicles. First, manual cruise control was invented. Then, adaptive cruise control was introduced, and then adaptive cruise control with lane assist. And now, Tesla (almost) has FSD and self driving taxis were on SF streets (albeit for a brief period).

Had you told me to jump into a self-driving car 5 years ago, I would have said no thank you.

For customer service teams, generative AI doesn't need to be an all-or-nothing overnight proposition. Any generative AI implementation plan needs to include incremental phases to build trust in the system.

CS/CX leaders should consider including these phases in a roll-out:

Sandbox: Test, test, test. Make sure you and your agents trust the AI's answers.

Knowledge Base Search: Use generative AI to enable agents to search thru your support knowledge base and find answers to customer questions faster.

Optional Agent Assist: Give agents the option to, with the click of a button, generate an answer to a question.

Automatic Agent Assist: As soon as a customer email or live chat arrives, have an AI automatically draft a response. The agent can edit and send the final response, and the AI should learn from the edited response.

And the final phase(s):

Automatic AI Responses to 1% of users

Automatic AI Responses to 5% of users

Automatic AI Responses to 50% of users

Automatic AI Responses to 100% of users

With the right level of trust at each phase, you can successfully implement generative AI in your org.

And it definitely won't take as much time as FSD.