Has AI been over-hyped in business?

(TLDR at bottom)

Maybe it has, but I wonder if much of this perception is a lack of understanding about exactly how much additional work is involved to deliver a project : I suspect a large number of projects are falling over because of a lack of forward planning and engagement.

From experience I'd say that if the answers to the below questions on how business is supporting data science are "Yes" then the project might have a fighting chance:

• Are there people who can help bring the right data together and help structure it (and continue to restructure it) in a way that can be utilised by a data scientist?
• Do those people understand the data and its lineage? And are they linked in with infrastructure, software and IT teams so that they can plan ahead and influence for changes and updates?
• Are there people lined up to fix historic & patched problems in the data feeds so that the models are working off trusted data without workarounds as much as possible?
• Does the senior leadership understand that this kind of project needs significant investment in systems, processes and people and that it likely won't fit into traditional project frameworks?
• Is there a specific project with measurable outcomes that the model will support, preferably running in parallel with an un-modelled version (control)?
• Is there someone in the business who is prepared to sponsor and implement the model?
• Are project sponsors prepared to hear the phrase "it didn't work, but we've learned something useful"? Are they prepared to start again. And again?
• Is there a mechanism for regularly and reliably feeding modelled data into your live systems so that it can be acted upon? When there are issues how would it handle them?
• Are the owners and handlers of the live system on board and have they been given opportunity to plan in any changes that might be required?
• Is there a reporting mechanism that can pick up model performance and flag any issues automatically and are there people who can interpret the significance of these flags on an ongoing basis?
• Are there people who can translate complex mathematical workings into digestible business language (and vice versa) and who will be able to spot comms issues, keeping the project on track?
• Is the business prepared for a long term investment of constantly and consistently monitoring and measuring performance and ready to invest to revisit that model when it no longer performs?

TLDR:

An AI project will need at least one really good data engineer, engaged and educated leadership and sponsors, a well chosen project, mechanisms for delivery, monitoring and error-handling, strong comms and a long-term strategy.

...and I haven't touched on what skills or resources your data scientist(s) would need to actually build the model...