I've Sat Through a Lot of AI Demos. Nobody Talks About This Part.

Working in technology and innovation within a mid-sized law firm means I've had a front-row seat as an extraordinary range of new and genuinely impressive AI products have started appearing in the legal technology market.

I've met some incredibly clever people building incredibly clever products. I've watched demonstrations where complex tasks are completed in seconds. I've seen information extracted, summarised and analysed in ways that would have seemed extraordinary only months ago.

The pace of development is daunting, yet exciting.

Yet one thing I have noticed is that every conversation starts with what their AI can do for the organisation. But none start with what the organisation needs to have in place for their AI to do it well.

I fear we are in danger of skipping over one of the most important parts of the AI conversation.

What exactly are we giving AI to work with?

AI doesn't arrive in an organisation with twenty years of context about how that organisation operates.

It lands exactly where it is today. It meets its data, documents, systems, permissions, processes and knowledge. It also meets all the decisions, compromises and workarounds that have accumulated over years of people simply trying to get their jobs done.

We're having increasingly sophisticated conversations about what AI can do with organisational information without necessarily having equally sophisticated conversations about the state of that information in the first place.

And in my experience, the reality inside organisations can be really messy.

Organisations don't evolve according to an architecture diagram

Over time, businesses accumulate technology. A new system is introduced to solve a problem. A spreadsheet is created to solve another. Documents end up scattered. Someone develops a workaround because an existing process doesn't quite work. That person leaves, but the workaround stays long after anyone remembers why it was created in the first place.

Then newer enterprise technology arrives. Organisations invest in increasingly sophisticated and remarkably capable platforms, but often old processes remain, old spreadsheets remain, workarounds remain - this is how organisations evolve.

AI inherits the organisation we give it

When someone demonstrates an AI product, we're usually shown an ideal scenario, information is neatly available, the AI can access it, and we're shown all the wonderful things the tool can do. But inside a real organisation, things are rarely that neat.

Before we can confidently talk about what AI will do with our information, we need to understand the environment we're asking it to work within. And that means questioning how we think the organisation operates and finding out how it actually operates.

Do we really know where our information lives? Which systems hold the authoritative record? How consistent and reliable is our data? What is sitting on legacy systems? What is still being managed through spreadsheets? Who has access to what? Where does our knowledge live and how much of this exists in people's heads?

Even our established workflows deserve another look. Are they genuinely the best way to work, or are we simply carrying forward processes shaped by systems, constraints and decisions that have accumulated over the years?

Because before we can decide what AI should do for an organisation, we need a clear picture of the organisation as it actually exists today - not what we assume it to be.

AI is giving organisations an extraordinary reason to finally examine some of the things that have been allowed to evolve quietly in the background for years. Not because everything needs to be perfect before we can use AI, but because we need to understand what we're pointing our AI tools at.

Maybe before we start looking at the tools we should ask: "What does our organisation look like, and what are the gaps for us to get real value from AI?"