We are in the golden age of AI. New and novel applications of AI show up every day, transforming businesses and lives. On the other side of this AI bounty are the failed and clearly mismatched projects which make you feel like the AI hammer is being used to smash (learn to smash?) every nail.
While the scale and nature of FAANG and a few other businesses justify investing heavily in AI (or tech for that matter) for any edge, with a 0.001% improvement in any measure is tens or hundreds of millions of dollars of impact, the rest of the business world is struggling to gain from practical applications of AI.
In general, for software projects - well thought out use cases with cross-functional teams have a decent chance of success. A cohesive team comprised of business embedded SMEs, strong application developers in tune with business processes, technologists who can roll out complex, scalable systems are key to rolling out high ROI technology projects. AI projects are no exception.
Yes your data scientist is the star of the show and yes they can produce very compelling research to fund a project but it takes a team to deliver results. Beatles won’t be the legends they are if John was a solo act.
Key takeaways from our current experiences delivering AI-enabled successful projects to production:
1. Know your requirements.
2. Analyze what is needed to deliver value: Statistical models -> ML -> Deep Learning
Invest in Data - 80% success relies on having clean data (labeled correctly for supervised learning).
3. Create a cohesive team to iterate over the said requirements and deliver incrementally.
4. Create a logical elastic architecture, able to scale appropriately as your usage grows.
5. Agile Devops, auto-scaling, easy access to frameworks -> if you cannot manage this you are bound to fail.
6. Continuous business validation, know when you are in your happy place.
7. If you are stuck in research paralysis, take a step back, realign your goals to business objectives and deliver value now.
At GuardX, we continue to deliver end-user applications to our large enterprise clients on our open extraction platform. We enable our client data scientists to shine, providing data preparation, data tagging, validation, UI and business process delivery while they build the core intelligence (the sexy shiny bits!). Reach out to us at email@example.com.