Why You Should Aim for a 10× ROI – ‘Return on Intelligence’
As organisations start using AI properly, token usage and consumption starts to become a concern. I’m increasingly seeing this conversation play out in two – very different – ways.
Some organisations have committed significant money to AI capacity, licences or tokens, but aren’t yet using what they’ve bought. They’ve made the investment; but haven’t found enough valuable use cases to turn it into a return.
Others have the opposite problem.
Adoption is growing, consumption is rising, and people are starting to question the bill. How much are we spending? Why is it increasing? How do we get it under control?
They look like different problems, but I think they are two sides of the same issue – that we’re focusing too much on consumption and not enough on what we’re getting back.
If you’ve bought AI capacity and aren’t using it, consuming less clearly isn’t the answer. You need to find more valuable things to do with it.
Equally, if consumption is increasing, automatically trying to reduce it could be exactly the wrong response. If that additional intelligence creates significantly more value than it costs, you probably want more of it, not less.
So perhaps we need a better way to think about AI spend.
What return are we getting on the intelligence we’re buying?
That’s what I mean by the term ‘Return on Intelligence’.
And I think we should put some ambition around it.
For every £1 we spend on AI, we should be looking for £5 to £10 of value in return.
I’m not suggesting 5 to 10x is some universal industry benchmark. It isn’t. I see it as a useful commercial ambition.
If we can’t see a credible route to that kind of return, perhaps we should question the use case. But if £100,000 of AI spend can create £500,000 to £1million of value, the interesting question isn’t “How do we reduce the £100,000?”
It’s “Where else can we make that equation work?”
Some of that value is straightforward. People find information faster. Repetitive tasks disappear. Analysis that took hours happens in minutes. Customer service colleagues get the right information while they’re actually speaking to customers.
At scale, those small improvements can add up to significant value.
But I think there’s a bigger opportunity, particularly in customer experience.
For most of the history of customer service, intelligent conversations have been expensive because they’ve required people, and people have a finite amount of time.
So, we’ve rationed conversations. We’ve decided which customers we contact, which interactions justify human attention, what goes to self-service and when it makes economic sense to proactively engage someone.
Entire operating models have grown up around a simple constraint: there are only so many conversations we can afford to have.
AI starts to change that.
Imagine somebody buys a product or service from you. Today, you might send them an automated email and wait for them to contact you if something goes wrong.
Historically, it hasn’t made economic sense to proactively engage every customer and understand whether they need help.
But what if the economics change? What if AI allows you to understand the context, identify the right moment and have a genuinely useful conversation before that customer ever needs to contact you?
Perhaps you prevent a problem. Perhaps you discover they haven’t understood something. Perhaps you identify something else you can genuinely help them with.
The interesting thing is that you haven’t made an existing interaction cheaper.
You’ve created an interaction that previously wouldn’t have happened at all.
You’re finally achieving that Return on Intelligence…