Pilot Paralysis in Generative AI?

Pilot paralysis – do you see this happening in Generative AI implementations??

Ashish Umbarkar and I were having an interesting discussion on this topic today. We both saw this phenomenon quite regularly during the early days of IoT and Mixed Reality and were wondering if we might be up for a similar challenge.?

Companies got very exciting about what could be possible with these new technologies and bought into the business value it could achieve. Understandable the time of very large implementations are rare due to previous bad experiences in system implementations. So they wanted start small, test it and then roll it out.?

However, these pilot projects or prototypes never transition to full-scale production. This phase of stagnation typically occurs after the initial testing or development stages have shown promise, but before widespread deployment has been achieved. Even circling from one pilot to the next without moving towards widespread success.

Issues we saw where

• Scaling from a few devices to hundreds or millions introduced data complexity, and network reliability?
• Integration challenges into existing system landscapes
• Security concerns due to the sensitivity of the data
• High pricing and effectiveness for MR devices
• Motion sickness due to the Augmented Reality?
• And if the devices can drive a seamless user experiences

These causes in combination of lack of clear objectives, fear of failure or inadequate results that fail to meet the high expectations are leading to this pilot paralysis. Moving beyond pilot paralysis is crucial for companies aiming to stay competitive and realize the full return on their investment in new technologies.

Here are my top 3 tips to prevent this for Generative AI:

1. Define Objectives: Clearly articulate what you hope to achieve with generative AI, whether it's improving customer or employee experience, enhancing content creation while driving productivity, or automating specific tasks.
2. Identify Use Cases: Select use cases that have a direct impact on business goals and are feasible with current AI capabilities. Start with something smaller that you can surround existing applications. We currently see great traction on topics like?

a. chat with your data in contact center settings or

b.?how can you improve knowledge management with AI.

c.??document automation for e.g. contracts, accounts or legal with AI

1. Understand the state of your data: crap in equals crap out. (Please have a read of Jennifer Stockton post:?https://www.dhirubhai.net/feed/update/urn:li:activity:7187898191508320256/)

• What is your data management strategy: Gather and prepare diverse and high-quality data sets that the AI models will use for training.
• Data Governance: Establish robust data governance practices to ensure data integrity, security, and compliance.?

We see especially on the the topic of security high concerns with our customers. Make sure you are ready to address these early on. One way to make it easier to start is instead of going the route in coding and implementing via diverse AI services yourself, you can also go and start with “Buy and Configure” (like many of the Copilots) or “Buy and Extend” (e.g. as part of LowCode platforms, e.g. Copilot Studio). These will help minimize some of the initial burden and you can learn along the way and get started.?

What is your experience in dealing with generative AI? What are your tips? What could prevent Pilot Paralysis??

------------------------

??I am Dr. Nicole Wieberneit

???I post on GTM & sales strategy, aligning customer and employee experience with AI, leadership

???Please reach out via DM or connection request