or only 3/20 have moved into Production?
Here are a few reasons why companies are stuck in Pilots for the last 2 years
---------------------------------------------------------------------------
1. Enterprises are doing Pilots with all types of AI companies - Nvidia, Cloud Providers, Foundation models, Middleware platforms and Applications. They don't know the difference between these, and what they need for a particular business requirement
2. People alignment - PMs and Engineers, have no clue how to involve the
"GenAI" people. Lot of confusion around how much and who all in the organisation should have AI at the centre of their agenda. All VP, Sr.VP, CTO are doing their own "shopping", while developers are championing their own wrappers.
3. Prototype != Product, they are unable to differentiate between a great demo with UI/UX and a production-grade AI-centric product that can deliver on all customer scenarios.
4. No RoI - It is not that AI cannot provide substantial RoI. Its just that no one has created an RoI calculator that spans across IT + People-cost+ Development-cost. Even the Application Vendors are new to this!
5. Dirty Data - A lot, a lot of data has been resting in peace in data lakes. Leadership believes that they have lot of data, Data Engineers believe that their pipelines are perfect! Reality is these pipelines were never built for AI world, and that the data sets in the lake would require months of cleaning.
Some solutions to unlock being Stuck-in-Pilot
-----------------------------------------------
1. Start with Evals - Yes, given a use case, don't write its BRD/PRD. First write down exhaustive list of input/output scenarios, evaluation metrics, and test data sets.
2. Create an RoI calculator - Create a calculator, that transcends IT + People-cost+ Development-cost over 3year lifecycle. Make practical assumption on adoption time across departments, and technology cost doing down 50% per year (minimum)
3. Data Roadmap - Your roadmap to curating data and creating robust real-time and batch data pipelines is much more important than the product roadmap itself. This is infra! Product roadmap is a purely intellectual activity without infra.
4. People alignment - Find people, across the org, who are willing to suspend their ego/Titles/Experience and put outcomes at the centre. Lets call them your AI champions. Give them a budget and access to CEO/CTO (without the middle-men)
5. Buy vs. Build - Buy everything! period... your vendors (mostly startups) know much better than your VP, they have greater skin in the game as well. Only when you are convinced that something has significant - PnL or strategic or CX impact, you can start the "Build" project.
P.S. - Happy to chat, if your stuck in pilots while having a mandate from the Board, pressure from CEO/SVP, and are still feeling lost.. hit me up on DMs
This post was originally shared by on Linkedin.