I previously posted about

• here ... how buy-now/pay-later provider Klarna did serious damage to its business with an AI rollout, by too-rapid cuts to its service staff
• here ... how the resulting hit to service quality has a delayed impact on business growth through customer losses and reputation damage

So let's bring those two insights together to plan a safe AI rollout.

A growing business needing good customer support

 Here's a simplified case, something like the Klarna company situation: 

• we have a strongly growing business, needing good service to sustain its rising customer base
• we plan to roll out an AI-based service to both improve customer support and to radically cut the staff we need
• customers make support calls across a wide range of topics, each needing certain tasks to be done by the team
  • AI can automate the handling of customer calls but ...
• while the most common, simple calls can be mostly automated, the less common, trickier calls are both harder to automate and also still require human input 

It would take a lot of words to explain how the elements of this episode interact, and the explanation would still be tough to follow. So here's a graphic showing how that episode should play out week by week if we do it well ... 

So the key to getting this AI solution rolled out safely is to:

1. carefully assess how much staff time will still be needed for each task to which AI support is added
2. what that implies for the amount of human support effort will still be needed, over time, as the AI implementation progresses
3. be careful to down-size the support team slowly enough to ensure that continuing human-support effort is provided

Plan ahead with a digital-twin business model

Now this is a simplified illustration, so the practical realities will raise further complications. But however simple or complex the case, any AI-support rollout is a venture into an unknown future. 

So it is surely advisable to simulate that future, and explore the assumptions and uncertainties on which success will depend. Of course, reality will play out differently, but:

• we will understand those outcomes much better from having rehearsed them first
• we can modify the model as events unfold and adapt the remaining weeks of the project.

Here's a model playing out a less-than-ideal strategy for the service-AI rollout in this case (OK maybe rather pessimistic, but you get the idea?) ,,,

Wider implications ... 

Not only is this a simplified illustration, but it is also highly localised, just to that balance between customer calls and the support team.

I checked the realism of this model with the Chief Product Officer for a SaaS business, and we worked through the dependencies between AI's impact across multiple functions.  

• Lead customers demand platform enhancements (which AI enables faster)
• ... which adds to the scope and complexity of the platform
• ... which [1] makes customer-on-boarding more complicated (though AI may alleviate some of this complexity)
• ... and [2] increases both the range of issues on which customers need support, and the frequency of their calls
• ... which magnifies the challenge set out in the illustration above, raising the risk of those bad outcomes

So ... not only must we understand the impact of AI on multiple functions, but also understand the interactions between all of those initiatives. ... Good luck trying to do that with a spreadsheet!

_______________

To get these insights on strategy and business models directly,
and get offers on our courses,
subscribe to my insights emails here.