Steak, muffins and chihuahuas: The unkept promise of ‘game-changing’ Legal A.I.
Only a few years ago every publication and conference seemed to be screaming from the rooftops about ‘robot Lawyers’ coming to take our jobs.
Experts suggested the Legal industry would soon be ‘upended’ by A.I. and that you’d better get aboard the bus otherwise it would leave without you. Ironically, very few of these experts could explain exactly what A.I. was, how it worked, or how it would benefit in-house teams.
Despite this, clients consistently told us they were ‘going to invest in A.I. and Blockchain’ without being able to tell us what problem it would solve for them. For one thing, the A.I. guys had great marketing.
It appeared to us that the Legal industry was caught in a classical A.I. Hype Cycle, and so three years ago we published ‘Legal A.I: High on artificial, low on intelligence’, where we called out the craze as a case as having “all tip but no iceberg”.
It turns out we weren’t wrong.
So what happened to Legal A.I?
To understand the answer to this question, you must know something about steaks, muffins, and chihuahuas.
All sizzle, no steak
Clients were expecting some sort of magic from A.I. and were quickly disappointed for one simple (but legitimate) reason: There was no A.I. to be seen in any of the products.
These unscrupulous vendors who were trying to sell ‘the sizzle, not the steak’ fell into their own trap, when the excitement around A.I. turned out to be somewhat of a false positive.
The problem with A.I. is not just one of technical risk (i.e. building a product that consistently solves a big problem). The real challenge is product-market risk: are you building something that people want to buy?
In a world where the average Legal function is very risk averse and has adopted no Legal technology to date; is the first thing they buy going to be a ‘black box’ they don’t understand?
To try to address this fear, vendors called their products friendly names like Kim, Josef, Watson etc.
The Legal A.I. marketing campaigns weren’t total failures either. Some GCs were interested, largely those who wanted to be perceived as ‘innovators’. They wanted their personal brand to be associated with being ‘at the bleeding edge’.
The challenge, of course, is that the bleeding edge crew only represented about 2.5% of the total addressable market (enough to fill a few conference stages but not enough to build a business around), and they always struggled to explain to their boss the exact problem the A.I. investment solved or the value it created for the business.
However, for the above reasons we don’t make big claims about it – it is a feature, not a product.
Muffins and Chihuahuas: ‘The Brave Ones’
The second answer to the question ‘what happened to A.I.’ is that the true ‘brave ones’ are still there toiling away in their garages; tweaking their tools and awaiting the breakthrough in Product Risk or Market Risk.
We very much respect these folks, however, they confront a multitude of headwinds before they reach ‘the slope of enlightenment’.
Telling the difference between a chihuahua and a muffin is relatively straightforward for three reasons:
1. There are a lot of photos of chihuahuas and muffins (i.e. a large data set);
2. Muffins and Chihuahuas actually look quite different;
3. Unless you are planning on eating a muffin as a result of the A.I. answer, and instead find yourself eating a dog, the risk of getting this wrong is relatively low.
Further, this branch of A.I. is ‘categorisation’ (which is successfully used today in discovery and due diligence applications).
To be really valuable to in-house Legal teams you really need ‘reasoning’. It’s not enough to know whether it’s a chihuahua or a muffin if there’s no actionable insight.
Further, the costs of developing the main relevant branches of Legal A.I. (areas like natural language processing & machine learning) are so huge it’s unclear whether it will mostly be solved by a vertical specialist like Legal AI or a horizontal generalist like Google, AWS or Open AI.
The Start of the Legal A.I. Winter
The truth, as this recent Economist article points out “Supervised machine learning doesn’t live up to the hype. It isn’t actual artificial intelligence akin to C-3PO [a humanoid robot from the “Star Wars” films]. It’s a sophisticated pattern-matching tool.”
The combination of technical disappointment and the market reality of a COVID World guarantees companies are only going to invest in tools that have a clear, low risk, rapid path to ROI. This means clients will continue to invest heavily in automation, workflow tools and systems of record, forgoing the bright lights of the conference circuit for a few more years.