I went through the stages for AI transformation covered in the course, and based on my empirical experience and research studies across multiple companies, I see the stages needed in a different order.
The main reason why AI transformation fails in companies seems to be because of the lack of buy-in from Management, due mostly to not understanding the field and the value it provides to any business.
So even though going into stage 1 of “executing pilot projects” is relatively easy, setting stage 2 “building an in-house AI team” is really missing the training and communications stage you need to go through before being able to jump into an in-house AI team. In the course training and comms are in the last stages though.
Have you had any similar experiences or do you fully agree with the order in the course?
1 Like
Hi @Rocio_Bachmaier ,
Welcome to the DeepLearning.AI community.
Building an in-house AI team will require buy-in from management. In addition to the success of the pilot projects, providing broad AI training to the management will help in setting more realistic expectations about AI and securing that buy-in.
I agree with the ordering of the stages in the AI transformation playbook. However, I think one should adapt the tasks within the stages according to the situation at their company while being conscious about minimizing wasted effort.
3 Likes
Hi Rocio,
Is there anywhere you documented these use cases, I’d be super curious to read more about it. For context, my background is sales/go to market in various tech companies. Please let me know if I misunderstood, it sounds like this is on the broader topic of securing the leadership buy-in, resourcing and prioritization to adapt the organizational structure, to hire and enable an AI team or support an AI project build/rollout - did I get this right? In this topic, here are my observations having seen behind the scenes of 100s of companies (you learn a lot about your customers in tech sales):
(1) For most companies, the pre-requisite of quality, unified data is missing. Even the leading, most well-funded, digital-first companies I’ve worked with are generally using fragmented, poorly instrumented data that is such huge undertaking of resources to fix, ML and data science teams are hired just to tape it together, and usually are the only ones who can pull useful insights (slowly) from it. Even implementing basic tools like Amplitude, Salesforce, Looker etc on this data is very difficult, nevermind having it be usable in AI applications
(2) Executives/budget holders rarely understand or, if they do, miss the incentive to fix this problem. This is due to quarterly demands from investors, the market and so on (growth at all costs etc). What I’ve noticed is that these companies will indeed release an “AI” product for customers, while BTS the situation is as above
(3) Most organizations are still very early to being data-first more broadly, which I would argue is a pre-requisite to them being in a position to build/support AI-first teams/projects.
This is not a mutually exclusive or exhaustive list, would love your feedback on if this is relevant and happy to chat further
1 Like