On-demand: Why do so many AI transformation programs stall out?
Published:
Updated:
2 minute read
Box x Slalom: Closing the gap between AI pilots and real enterprise impact
AI pilots are proliferating and, in many cases, failing to scale.
- 90% of companies are using or piloting AI
- 60% fail to enter production
- Of those pilots that go into production, only 5% show ROIs
Why? Teams layer generative AI on top of broken processes, declare a proof of concept a success…and then wonder why nothing changes at the organizational level. The technology works. The transformation doesn’t.
This is exactly the problem Box and Slalom have come together to solve, with our partnership extending into deep practices across SAP and Salesforce (including Agentforce Foundations Accelerator).
If your organization is sitting on successful pilots but struggling to cross the gap to enterprise impact, this conversation is worth your time:
Key takeaways
- AI can’t be rolled out like traditional software. Unlike a standard tech deployment, becoming AI-first requires real change management, business fluency, and a full transformation mindset.
- Data silos kill AI. Without sound data governance and consistent metadata, even well-built AI systems produce inconsistent and unreliable results. The data foundation has to be right before the AI can deliver.
- Zero legacy beats rip-and-replace. Slalom’s approach, focusing on small, quantifiable modernization steps tied to real business outcomes, means organizations can move forward without expensive overhauls that ultimately preserve the same old problems in a new coat of paint.
- Governance can’t be an afterthought. Organizations that skip governance planning before going to production may see early wins, but the cracks appear fast. Clear accountability structures matter from day one.
- Tie every use case to “dollars and sense.” AI projects that earn executive buy-in are the ones that directly, explicitly connect to cost avoidance, revenue lift, or measurable ROI. Name and keep measuring your outcomes.
Show notes
The Intelligent Content Accelerator, built in collaboration with Google Cloud
Highlighted at Google Cloud Next 2025, this accelerator combines Box’s ICM platform, Google Cloud’s AI infrastructure, and Slalom’s consulting expertise into a solution that lets knowledge workers automate routine deliverables at scale. Weekly leadership updates. Quarterly roadmap decks. Sales RFPs. All of it driven by Box AI’s extract agent, Slalom-built templates, and Google Cloud’s document generation, in a workflow that can run at the click of a button.
Metadata extraction to unlock unstructured data with Box Extract
Enterprises are sitting on billions of files. Contracts. Invoices. Clinical studies. Customer documents. Most of that content is unstructured and unsearchable, which means it can’t power the AI workflows organizations are investing in. Box Extract changes that. Using large language models (LLMs), it turns plain-language instructions into document extraction rules that surface structured metadata from any file type, then sends that data exactly where it needs to go:
- Into Salesforce for real-time contract analysis and CRM enrichment
- Into SAP and other ERP systems for streamlined invoice processing and vendor lifecycle management
- Into Snowflake, ServiceNow, Workday, and beyond for cross-system automation
How it all gets integrated with Salesforce, SAP, and more
Slalom’s deep practice across SAP and Salesforce, including Agentforce Foundations Accelerator, means clients can connect Box’s content intelligence to the business systems where decisions actually get made. When metadata extracted from a contract in Box flows into Salesforce Agentforce, sales teams get real-time renewal insights without switching tools. When invoice data flows into SAP, finance teams can automate processing workflows that once required hours of manual effort.
The integration layer is where AI pilots scale, shift the operating model at scale, and change the business outcomes.
