AI That Works: Lessons from MIT and BlackRock

MIT's latest research reveals a stark reality: 95% of generative AI pilots fail to achieve meaningful results. The culprit isn't faulty technology but rather that most organizations approach AI as experimental projects rather than foundational infrastructure.

The study's most compelling insight lies in identifying what actually works. Companies that invested in purpose-built internal tools achieved significantly higher success rates than those attempting to build solutions from scratch. This finding aligns perfectly with what we observe in finance: while ad-hoc pilots inevitably get shelved, tools seamlessly integrated into existing workflows drive genuine adoption and measurable ROI.

BlackRock exemplifies this infrastructure-first approach. Engineering directors recently detailed how the firm developed an "App Factory" that empowers domain experts to design extraction workflows within a controlled sandbox environment, then deploy them cloud-native within days rather than months. The system is systematic, measurable, and built with compliance requirements from day one.

The bottom line: Organizations treating AI as core infrastructure rather than experimental curiosities are the ones capturing real value. For research and investment teams, this translates to a clear directive: invest in pre-built platforms with robust governance and transparency capabilities already established. Everything else remains trapped in proof-of-concept purgatory, never graduating beyond the pilot phase.

~Ryan Erickson, Founding Executive

TLDR: What you need to know about AI in Finance

MIT: 95% of GenAI pilots fail
Success rates are far higher when firms buy purpose-built tools vs building from scratch.

BlackRock’s “App Factory”
BlackRock engineering leads demonstrate how they reduce AI development cycles from months to days.

Accenture & BBVA prove ROI
AI workflows boosted engagement and cut costs, with ROI topping 20%.

Deloitte’s playbook
Eight principles for using GenAI to boost efficiency in finance operations.

Basis Prompt of the Week: Citations done right!

💡 We had a client ask us to try this prompt in a live demo, and the result was impressive. Basis pulled together internal research notes to generate a concise, accurate thesis that gave the team instant alignment on their own collective view.

BASIS FEATURE OF THE WEEK:

Earlier this year, Zakey Faieq wrote our most-read blog on creating rich citation experiences with LLMs. Since then, he has evolved that work into a core Basis feature.

Now, every AI-generated summary or Analyst answer comes with inline citations. In the example below, you can see three citations to internal research notes. When you click on a citation, it takes you directly to the underlying source.

It is a simple but important shift that makes it easy to verify insights before trusting and sharing them across the firm.

About Shift
Shift builds AI-powered research tools that help investment teams surface insights and drive better decisions. We're a team of builders and finance experts based in Charlottesville, VA. Learn more at shifthq.ai.