AI strategy

AI Strategy Playbook Summary (MITTR x Boomi, 2024)

1. State of AI Deployment

• 95% of companies use AI; 99% plan future use.
• Only 5.4% of U.S. firms have achieved enterprise-wide deployment.
• 76% use AI in ≤3 functions; most are still in the pilot phase.
• Key use cases: code generation, customer support, document analysis.

2. Strategic Shifts Required

Organizational & Infrastructure Readiness

• Successful AI scale-up hinges on:
  • Modern data architecture
  • Cross-functional data governance
  • Platform and cloud migration

Vendor Strategy

• Most firms won’t build their own LLMs.
• Off-the-shelf + fine-tuning > building from scratch.
• Multi-vendor AI ecosystems recommended.
• Preference for vendors that ensure data privacy and update agility.

3. Investment Trends

AI Readiness Spending

• 2022–23: Mostly flat
• 2024: >90% of firms plan increases
• 33% plan 25–49% increases
• Spending categories:
  • Data pipelines, platform modernization
  • Strategic/cultural change, AI modeling

Cost Barriers

• Smaller firms (rev. $500M–$1B) face budget constraints.
• GPU availability and infrastructure requirements pose major hurdles.
• AI infra includes:
  • Compute (e.g., GPUs)
  • Skilled labor
  • Model lifecycle management

ROI Challenges

• Shift from cost savings to growth enablement.
• ROI needs to reflect:
  • Hard gains (automation, task speed)
  • Soft value (employee satisfaction, innovation enablement)

4. Data Core: Foundations for AI

Key Challenges

• 49% cite data quality as #1 deployment bottleneck.
• Large firms struggle more due to legacy IT systems and complex data silos.

Best Practices

• Data lineage tracking across systems and models.
• Prioritize data liquidity: real-time, cross-platform access to data.
• Leverage metadata to contextualize unstructured data.
• Replace centralization strategies with contextualized architecture.

Legacy Systems

• Inhibit scalability and data flow.
• Only 1 or 2 individuals often know how to maintain them.

5. Risk and Regulation

Technical Risks

• Generative AI risks:
  • Hallucinations, copyright infringement, systemic bias
  • Cyber threats (e.g., prompt injection, data poisoning)
• 98% of executives prefer safety over speed.

Data Governance

• Privacy and compliance = top concerns for financial services.
• Increasing use of AI for cyber defense (e.g., vulnerability scanning, threat hunting).

Regulatory Landscape

• EU’s AI Act: risk-tiered compliance obligations.
• U.S. Executive Order (Oct 2023): emphasizes safety testing, oversight, and transparency.
• ISO-based standards emerging, but fragmented.

6. Conclusions and Principles

• Year of Foundation Building: Data hygiene and platform robustness are critical.
• Custom Use Cases win: Domain-specific AI > generic GPT applications.
• AI ROI must consider long-term, organization-specific outcomes.
• Risk-aware adoption is becoming the default enterprise stance.

"You don’t have to be an AI expert to get value from AI." — Matt McLarty, Boomi