Avoiding AI lock-in
One of the big problems with deploying a software as a service (SaaS) product across an entire organisation is that once it is widely used, swapping it out becomes an extremely costly exercise.
In the 1990s people used to talk a lot about middleware, a chunk of software that sat between a relational database system and an enterprise application. The idea was that if access to the database was programmatically only possible through the middleware, it would be possible to swap it out without the need to make major changes to the application. There were clearly plenty of compromises, in that the functionality surfaced through the middleware was of the lowest common denominator, in terms of the subset of features that were common across all the relational databases it supported.
It was a great idea until you realise that when you do have a genuine choice and are not locked in by existing enterprise licensing agreements and corporate-wide policies, the reason people choose one relational database system over another is exactly because it offered something they acknowledge would be extremely useful.
Enterprise SaaS
The world today is dominated by a handful of major enterprise software providers offering SaaS to customers who need to use these products to help their organisations not only run business processes, but also differentiate themselves sufficiently from similar organisations.
A few weeks ago IBM rolled out a new generative AI-based chatbot across the business for human resources, to help IBMers find answers to HR questions more easily. Undoubtedly, this is a great use case for GenAI and reduces the burden of having HR staff handle every employee query directly.
But, the approach it has taken is using a chatbot that sits on top of its existing WorkDay HR platform and a new Successfactors platform. In other words, the chatbot is rather like middleware. Even though both WorkDay and Successfactors offer AI-capabilities themselves, IBM’s approach means that it can train and optimise its own AI chatbot, which integrates through application programming interfaces (APIs) with both WorkDay and Successfactors.
From an IT decision-maker’s perspective, this means that the underlying SaaS product can be separated from end users and client applications that require access to its underlying functionality. Instead of relying only on the user interface the SaaS provider offers, end users can interact with an AI system that has been specifically optimised to take into account the uniqueness of the organisation they work for. This is great from an employee experience perspective. And just like middleware for relational databases, the IBM chatbot shows that greater flexibility is possible if you avoid the AI capabilities built into SaaS products.