Why tech developers and lawyers must identify the exact problem before adopting AI tools

As a Data/Tech lawyer and software developer, I love innovation. But I am also cautious about using solutions without first knowing which problem precisely they will solve – or using technology without first thinking about issues that may arise from their use.

I have been helping companies on the use of artificialintelligence tools for several years (advice, contract negotiation/review, etc), and recently more on aigovernance – a popular topic since the launch of ChatGPT over a year ago. (Tech-savvy lawyers are more useful in that respect than you might think, as proper AI governance requires carefully thinking about the legal framework, risks and opportunities re the development, use or commercialisation of AI tools)

One thing I tell clients – with any tech solution, not just AI – is to be sure that the solution chosen is indeed the right one for the problem identified. Start by identifying your problems/needs, then find the right solution.

“GenAI for everything” has been a frequent mantra of late (just think of the many guides on how to boost your productivity by 1000%, follow these prompting techniques to earn 100x more, etc), and some companies (including law firms) have been integrating this tech at an unprecedented pace.

But in most cases, it seems that this is a “solution before problem” approach. With the idea “this will save us time”, companies use generativeAI without understanding what LLMs actually do or their limitations.

“We’ll generate contracts in minutes”, I see some law firms and legaltech companies saying. Great, but you’ll still have to check nearly every word to ensure it’s actually what you intended, so maybe a good contract automation package might be better for you (at least, there, you know in advance what the output will be). (And if your bespoke AI solution is intended to regurgitate your templates, maybe you’re better off with a more “traditional” predictive / discriminative AI system than a generative one)

Same with marketing e-mails. Your generator might create good e-mails in seconds, but will you then have to retrain it anyway when you launch a new product or service? If so, have you really chosen the best solution with not only the least upfront cost but also the least maintenance cost?

GenAI tools help with plenty of really good use cases, but be sure that they are right for *your* use cases. Otherwise, they may lead to productivity loss (rather than gains) – as the very nuanced article below in the Harvard Business Review shows based on more rigorous studies.

And as always, if you need someone to lend a hand in identifying the legal risks or finding legal solutions, do reach out.

https://lnkd.in/eXAktE3Y – “Is GenAI�s Impact on Productivity Overblown?