There's a common belief that AI + humans > humans alone. But that's not always true...
This week, I’ve been speaking with procurement leaders who are inundated with Procure-tech pitches. We all know the promise: AI will streamline processes, cut costs, and make teams more efficient. But the reality? Many AI tools take months (or even years) to implement, require a dedicated person to manage them and add complexity rather than reducing it, especially if suppliers or customers can’t (or won’t) use them.
I hear the same in sales:"we signed up to Salesforce and they said it would fix our leaky sales funnel but it's proved more complex than that".
More dashboards, more integrations, more time spent troubleshooting. So before you greenlight another AI solution, ask yourself—are you saving work or just shifting it?
Historically, new technology hasn’t always meant less work—it often just meant different work. I was a junior investment banker in the early 2000s - most companies didn’t have corporate websites yet - imagine that?! Gradually not just websites but online annual reports and databases of financial reports became the norm, we suddenly found ourselves doing a TON more analysis, working deep into the night… because we could.
Before that, deals had still got done. Senior bankers delivered advice. Back of an envelope analysis was a thing and (I am told) there was just one Excel-running computer on a trolley. It was wheeled around the office between teams (except in summer, when it I’m reliably informed was reserved exclusively for live cricket scores 🏏).
More data and more tools often create more tasks, but not always more productivity.
A study of radiologists found that AI assistance inconsistently affected performance—some radiologists’ results improved, while others actually got worse. AI doesn’t automatically enhance human output; its impact depends on how it’s used.
(Read more from the Harvard research here.)
This raises an important question for executives pursuing digital transformation in procurement and in B2B sales: Does adding AI help solve this problem or does it create more complexity?
AI isn’t free—at least, not yet. LLMs are currently expensive to run ($ and environmental impact).
As an aside, here at Nibble we put in place systems to make sure we don’t send the same request to a LLM multiple times over different chats to focus on efficiency. I have not seen other AI developers consider this and would love to know how much it happens.
OpenAI’s deep research costs around $200/month—potentially less than hiring an analyst, but still a meaningful cost (although I am sure it will get cheaper, not least driven by competition from other providers, including China’s low cost competitor Deep Seek). Sometimes we have found those roles are already outsourced to a country where skilled workers are cheaper than AI agents (for now) so AI doesn’t make sense.
For enterprises, this means the focus will may well be on augmenting (or outright replacing) expensive roles first which is the opposite to what we have seen in prior waves of outsourcing and business transformation.
At Nibble, we’ve seen firsthand that AI negotiation only makes sense when the process is repeatable and needed at scale, for example to execute hundreds or even thousands of similar negotiations. If you spend 15 minutes setting up an AI negotiation and only use it once, the efficiency gains disappear.
Next time, why I think “AI co-pilots” are now a little passe… 😉
You can subscribe to this newsletter by clicking here.
Interested in Nibble?