Mira Murati's Startup Has No Product, But 1 Chart Proves It Could Beat OpenAI in the Enterprise

2026-07-02 · Source: High ROI AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Corporate Strategy & Leadership, Capital Markets & Investment Management · Depth: Advanced, quick

Summary

Mira Murati's startup, Thinking Machines, is developing customized small language models (SLMs) for complex financial tasks, as detailed in a recent paper. Their "Outcomes Engineering Approach" focuses on achieving a specific reliability metric, such as 80% accuracy, to ensure investor trust and workflow adoption. The methodology outlines six critical tasks, including Financial Article Recovery, Central Bank Document Relevancy, and Document Truncation, emphasizing an ROI-first strategy from inception. This framework positions Thinking Machines to potentially dominate the enterprise AI market by creating a continuous value "Flywheel," offering a superior alternative to traditional forward-deployed engineer models and potentially surpassing competitors like OpenAI and Anthropic.

Key takeaway

For AI Product Managers or Directors of AI/ML evaluating enterprise AI initiatives, Thinking Machines' "Outcomes Engineering" approach provides a robust framework. You should prioritize defining clear outcome reliability metrics, such as an 80% accuracy threshold, and map AI development directly to specific business workflows. This ROI-first strategy, exemplified by their financial task model, can significantly improve initiative success rates and foster continuous value creation, potentially outperforming traditional FDE models.

Key insights

Thinking Machines' approach prioritizes an "Outcomes Engineering" framework, linking AI development directly to business value and specific reliability metrics.

Principles

• AI initiatives must align with clear ROI from the start.
• Define outcome reliability metrics (e.g., 80% accuracy) upfront.
• Workflow definition is critical for effective model and agent evaluation.

Method

Thinking Machines defines an outcome reliability metric (e.g., 80% accuracy for financial tasks), then outlines 6 specific workflow tasks with supporting evaluation metrics to ensure value creation and user adoption.

In practice

• Set a minimum 80% accuracy threshold for user trust.
• Define 6 tasks: document recovery, relevancy, labeling, and truncation.

Topics

• Outcomes Engineering
• Enterprise AI Strategy
• Small Language Models
• Financial AI
• AI Product Management
• Workflow Automation

Best for: Director of AI/ML, AI Product Manager, Consultant

Related on AIssential

• The TechBeat: The AI Pilot Succeeded. The Economics Did Not. (6/26/2026) — AI pilot success requires measuring business outcomes, not just technical metrics.
• Building Knowledge Graphs To Support Agentic Workflows — Knowledge graphs must drive measurable outcomes and actions, not just represent information, to be successful.
• The Real Reason Most AI Projects Fail in Enterprises - [The Strategy Failure] — Operationalizing AI requires a dual focus on advanced development and dynamic, expert-in-loop operations for continuous improvement.
• The Outcome Density Scorecard: Measuring AI Value Beyond Hours Saved — AI value is best measured by "outcome density," delivering more valuable results with less organizational friction.
• What Happens When AI Strategy Replaces Product Strategy — AI product roadmaps driven by fear of missing out (FOMO) often fail to deliver user value.
• Combining Information & Mechanics To Build Agents That Don’t Get Laid Off — Enterprise success for AI and people hinges on demonstrating causal links between work, capabilities, and tangible business outcomes, not just possessing knowledge.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by High ROI AI.