But technology has now crossed the line into something at best closely resembling human output in the form of large language model-based AI, and the narrative of what mature legal service delivery might look like has changed overnight. We’re struggling to merge the huge potential of this new capability with yesterday’s picture — how do we integrate this magic box, loosely linked to huge sets of data and able to produce sophisticated outputs based on any variations of a prompt — in other words, something quite like us?
The answer hasn’t changed from how we approached the introduction of any new capability. For example, AI is quite good at summarising anything you throw at it; it can put together a decent set of summary bullets or paragraphs in readable prose. AI can mimic a baseline understanding of things like contractual obligations — enough to extract and summarise them, or give context around how they might be triggered. But it also makes the occasional howler and is particularly guilty of making up references. We need to check everything AI does — but the ace up the sleeve is that all this is incredibly fast, turning hours into minutes. Where would we use such capabilities?
This is an extract from an article published in Briefing magazine. Read the full article here.