"Let a thousand flowers bloom": unleashing the AI productivity boom within

What do organisations need to do when the pain of rolling the AI rock up the hill turns to the chaos of it sliding down, seemingly out of control? Embrace it.

Until late last year making AI work was frankly a pain – especially from the perspective of a CEO. Data had to be wrangled, experts hired at escalating cost and middle managers persuaded to buy in. It was typically a top-down process requiring ambitious leadership and significant resources. Making it even more challenging was that proving a return on the investment (ROI) was hard: one BCG study suggested that only 10% of companies surveyed saw financial benefit from their AI investment.?

Boards struggled to understand the specific opportunity of AI. Nowadays it is explained to them by their children. Literally.

The launch of ChatGPT and its derivatives has turned AI in to a consumer product. Whether writing love poetry, fashioning party speeches or helping with homework, AI is now the go-to tool for a host of social and personal occasions. It is not by chance that the first industry about to be revolutionised by Generative AI is Education.

At Best Practice AI we specialise in AI conversations and can see that leaders now understand that the pattern of corporate AI take-up is already very different.

“Let a Thousand Flowers bloom,” is how one media executive described it. For employees – of all ages – experimenting with improving their jobs just got easy. From the marketing executive speeding up personalised email creation to the software engineer coding 25% faster to the middle manager checking how better to search for missing emails the pace of experimentation is picking up.

A study this year by BCG of its consultants using the GPT 4 version of ChatGPT to do many of these tasks showed that, on average, they completed 12.2% more tasks, 25.1% more quickly and up to 40% higher quality. Weak performers caught up more than strong ones - so it is not just elite consultants who benefit.

Corporates whose initial reaction was to ban the use of ChatGPT at work, citing data privacy concerns and the risk of AI “hallucination” are taking a cautious – but wrong – approach. Because the opportunity from “bottom-up” innovation is huge. But to make this happen leadership needs to get some basic rules right.

Firstly, incentives are often not aligned between management and employees.

If one can do the task in half the time then sharing the secret with the boss may lead to a doubling of targeted delivery expectations. Work from home exacerbates this – it is potentially far easier for remote workforces to capture the surplus of AI-powered task productivity, even as the social capital within organisations gets harder to maintain at distance.

In some circumstances using ChatGPT and its ilk is seen as "cheating" - doing the working with less effort. If that is your perspective we have news for you - everyone is at it.

So sort out those incentives – and treat your employees such that they want to share their ideas. We have seen companies offering cash prizes for the best use of tools and incentivising teams through hackathons and special coaching (we offer this service).

The challenge management should be focusing on is those employees resisting change, not the ones luxuriating in it .

But management has to behave as though they can be trusted with the new productivity opportunities. Every time a CEO stands up in front of the stock market and boasts about ?“reaping future efficiencies” from AI they are setting their own cause back.

Because having staff alongside for the AI journey is key. ?

AI currently works best as a co-pilot – a tool to hugely increase worker capacity but still needing strict oversight. Having a human in the loop - necessary to set the task and to respond to the output is critical. Indeed, it is one of the key guardrails that we recommend companies implement as a foundational rule. Given AI's tendency to "hallucinate" humans need to be responsible for output. If those humans think that they are training co-pilots so that management can ultimately run things on autopilot without them that process will stumble.

Meanwhile, the underlying challenges of delivering AI that pre-date ChatGPT will not go away. Indeed they may be exacerbated by it. Complicated, non-aligned data infrastructure and architecture, terminology and definitions in data sets, different processes across the organisation or industry and differing levels of digitisation and software maturity are all huge issues for machine learning deployment at scale. A proliferation of new approaches from Generative AI – and external data links – will not make the task easier; nor will data and information pollution from “hallucinatory” AI output.

This sets a triple challenge for leaders championing AI.

1.????They will need to support and channel the flow of new ideas – sorting out the good from the mad, even bad. To do this they will need to set out rules for the game, to decide where and when to intervene or go all-in on the ideas bubbling up from within (or spotted outside).

2.????Meanwhile, the old (hard) tasks of ensuring that the right data and digital building blocks are in place (AKA getting the plumbing right) needs to continue. At least, the difference now is that the Board will be better able to understand the potential ROI on this work,

3.????Finally - AI at scale will radically change the value chain in most industries. What it means when old bottlenecks disappear and new ones emerge will be profound. ?

Contact us at Best Practice AI if this is a conversation that you need to have and want seasoned AI leaders to have it with.

?

NB: It is worth noting that the thousand flowers phase did not end well for the young Chinese encouraged to experiment with new ideas by Chairman Mao. Those who did so too vociferously were targeted by the Communist Party’s Anti-Rightist Movement and ended up in prison. This did not end well for China overall, acting as a prelude to the tragedy of the Cultural Revolution. A lesson perhaps for businesses that over-react to internal creativity?

?