What do you think of connectors, or integration solutions?
It’s true, the majority of those connectors still require a human to build them. We have been following a pattern of building these connectors and now we are creating our own solution that can help people connect to multiple machines. But it's still going to be driven by humans. It can't be automated; that's not possible. APIs also change over a period of time, so you end up reworking.
It's a difficult problem to solve because there are all these competing softwares that are also trying to build their own services layer. So you're never going to get a clean API between solutions. When I look at valuations for data fabric, I always think about how hard it will be for that market to exist. You will have software to manage those APIs, but it's not real. You'll still have to have a whole team of developers, with one speaking finance, one speaking ERP, one speaking warehouse, etc.
This can be addressed via modern autonomous data models and a data collaboration platform that obsoletes data integrations through connected data. This type of solution address the huge integration debt and spend and improves IT team agility, reduces challenges driven by data silos and the copies created as a result.
Similar setup in another pharma company where different systems needed to talk to each other and were configured using a complex ESB with 100+ connectors. The cost of bad data could be a rejected batch of medicines; we had 3 people making sure that the mesh did what it was expected to.
Today the expectation is that systems integrate even when there is dynamism in the data. There is enough tech out there to provide for integration; there will always be exceptions and that needs humans to manage it. Will it ever be fully machine run? Probably not, but I am happy to be proven wrong on this.