Smart contract and AI hands

Blocked: smart contracts may have limitations in an uncertain environment

When do you need a human in the loop?  

Take a simple vending machine transaction, for example. You put money in, you receive a snack. What could go wrong? 

In fact, the machine could be faulty, the mechanism could jam, or a power cut could stop it working. And you’d be forced to call the operators to claim your money back.  

Rather than technology working like clockwork, a human would have to override it to give the disappointed punter a refund. Not all automatic transactions function free of human intervention. 

But take another example. You buy insurance for an airline ticket, and something goes wrong.  

If this transaction could be sufficiently digitised – so that flight times, delays and other contingencies could be written into a codified contract, imagine the joy at receiving automatic compensation if your flight was delayed, or your luggage went astray. Every point of the transaction would be encoded, from payment, to take-off, to refund if appropriate.  

No more time-consuming form-filling, or costly claims via a third-party agent – air traffic control data would automatically flag a mishap and the refund would appear in your digital account as if by magic. 

What is a smart contract?  

This is how ‘smart contracts’ would work thanks to blockchain technologies. We can see the appeal of these so-called smart contracts. Every possible outcome can be predicted, digitally encoded, and triggered if conditions are met.  

Unlike conventional contracts, they are automatically enforced without requiring oversight from banks, lawyers, courts or any other third party. This in theory brings speedier outcomes and cuts costs – no need for mediation and no quibbling about the terms – the result is irreversible. 

Their impact could be momentous – a secure, transparent, and efficient contract in contrast to a conventional agreement.  

It was back in 1994 that computer scientist Nicholas Szabo first introduced the concept of smart contracts, using the vending machine as an illustration.  

Smart contracts could remove the need for managerial oversight and an extra layer of humans in the loop, much like vending machines have done for snacks.  

Later with the advent of blockchain technologies, excitement grew around how these could transform and automate every transaction, both within and between companies and individuals. It’s the distributed and immutable nature of certain types of permissionless blockchains that allows the technology to support smart contracts.  

But – and this is a big but – there are limitations to how viable it is to turn every transaction into a self-executing, tamper-proof and irreversible smart contract. Where could they function best? 

Proponents of smart contracts say these would bring huge efficiencies to supply chains.  

Take an example – a food industry giant that dominates the market in banana exports. Today, a retailer might decide to buy the banana farms that supply it because this gives greater control over conditions where the fruit is grown and harvested – it’s an exceptionally fussy market. Bananas must be the right size, the right colour, the right level of ripeness and so on.  

Now the farm is owned, this eliminates uncertainty on both sides. The farmer produces and will be paid. But there is a cost to the retailer, which must invest in farms and pay farm managers to oversee banana production. 

Instead, imagine if every aspect of growing bananas destined for the consumer market could be digitised with sophisticated sensors gauging the composition and quantity of fertiliser, the level of moisture, the timing of the harvest.  

This can be encoded by a smart contract, and if farmers don’t meet certain requirements, their contract will automatically end, or they will receive a penalty. The retailer benefits from lower capital, managerial and bureaucratic costs, while the farmers can remain independent, and their payments are automatically guaranteed and possibly more lucrative. 

There are in this instance degrees of automation. Sensors could detect these conditions and a human in the loop could execute contracts if data proved that all was in order. This type of contract at least gives some flexibility for the unexpected. 

And it’s codifying the unexpected – predicting the unpredictable ­– that is the greatest barrier to the potential of smart contracts.

Are smart contracts going to deliver their promise?    

To return to bananas, what if a climate disaster damaged a crop, or political upheaval delayed delivery times, or a change in environmental regulations moved the goalposts for growers and exporters. Would it be possible to encode for this range of uncertainties? 

Could the writers of smart contracts predict another pandemic for instance, an earthquake, or forecast the outcome of Russian aggression in Ukraine? 

At best it would be extremely costly to attempt to predict and describe digitally every possible contingency and design an irreversible contract that would take these into account. In reality, this is asking the impossible – in certain situations at least, it is too complex. 

Does this mean smart contracts, eagerly heralded a few years earlier, won’t deliver?  

We may become wiser about where and when they will be applied. There are early initiatives for instance to use smart contracts for car sharing – which would see information about mileage, fuel, repair, and maintenance recorded on a permissionless blockchain. 

An insurance company has already attempted to offer insurance for flight delays based on smart contracts recorded on the Ethereum blockchain. And of course, cryptocurrencies such as Bitcoin rely on blockchain smart contracts.  

Other sectors – from supply chains, through to real estate, healthcare, legal services, voting systems, and intellectual property – have all been touted as potential users of smart contracts. Ever more available data in all sectors will allow for more contingencies to be predicted, digitised, and monitored.  

But as with any emerging technology, its potential may have been overhyped. The adoption of smart contracts will depend on many developments, from digital infrastructure to our ability (maybe aided by machine learning) to predict future outcomes – and not least whether the law on the ground recognises their legitimacy.

Further reading:

Halaburda, Hanna and Levina, Natalia and Semi, Min, Digitization of Transaction Terms as a Shift Parameter within TCE: Strong Smart Contract as a New Mode of Transaction Governance (June 9, 2023). MISQ, Forthcoming , Available at SSRN:


Natalia Levina is a Distinguished Research Environment Professor and Professor of Information Systems at New York University Stern School of Business. She teaches Digital Leadership on the Executive MBA and Executive MBA (London) and lectures on Knowledge, Work and Innovation on the MSc Management of Information Systems & Digital Innovation.

Learn how to innovate with new technology in the four-day Executive Education course Business Impacts of Artificial Intelligence.

For more articles on Digital Innovation and Entrepreneurship sign up to the Core Insights newsletter here.