Sign in
Back to all companies
Fractional AI logo

Fractional AI

Acquired

Custom AI engineering for enterprises and private-equity portfolio companies.

Related collections

Fractional AI at a glance

  1. Delivery can build distribution. Work inside private-equity portfolios gave prospective buyers direct evidence of the team’s capabilities. Relationships across several companies can compound the value of a services business.
  2. Measure the error that matters. The Zapier project tracked invented endpoints separately from missing fields. Specific evaluation measures made engineering changes easier to judge.
  3. Keep safety decisions with specialists. Cando’s assistant prepared assessments and documented meetings while specialists retained safety decisions. Automating the surrounding work can save time while preserving that authority.

Overview

Fractional AI built custom AI systems for enterprises, often through private-equity portfolio companies. Chris Taylor, Eddie Siegel, and Travis May founded the San Francisco firm in 2024. Its business was the engineering work between a capable model and a dependable customer workflow.[1]

In May 2026, a new services company backed by Anthropic, Blackstone, Hellman & Friedman, and other investors acquired Fractional AI. That standalone company launched as Ode with Anthropic in July. Taylor became Ode's CEO and Siegel its CTO. This was an acquisition through the services venture, with terms undisclosed.[2][3]

Fractional’s trajectory suggests that engineering delivery became more valuable when paired with investors who could introduce many businesses needing the same help.

Founding Story

The founders met at LiveRamp. May led the company, Siegel worked in engineering leadership, and Taylor worked in sales. May later founded Datavant; Taylor and Siegel started Wove. Their careers page dates the conversation behind Fractional AI to dinner in February 2024.[4]

LiveRamp group photograph annotated with Chris, Travis, and an arrow indicating Eddie off camera
Fractional’s careers page marks Chris and Travis in this LiveRamp photograph and points toward Eddie off camera.

Source: Fractional AI

The April 2024 founding thesis argued that large companies had valuable automation opportunities but struggled to recruit engineers who could deliver uncertain AI projects. Fractional proposed a San Francisco engineering culture serving those companies through custom engagements. The founders chose services because customers' problems differed enough to resist a single standard product.[5]

Later statements show that focus continuing. In February 2026, Taylor said: “we build AI systems that integrate into how businesses actually run to drive real impact.” In a June 2025 partnership announcement, Siegel said: “Claritev is approaching AI with both strategic vision and technical rigor.” Both statements put customer systems and subject knowledge ahead of a standalone AI application.[6][7]

Timeline

  • February 2024: The founders discussed the business over dinner; the founding thesis followed in April.[4][5]
  • March 12, 2025: Fractional introduced an AI due-diligence service for investment transactions.[8]
  • June 25, 2025: It announced its Claritev collaboration.[7]
  • January 28, 2026: It launched a healthcare and life-sciences division.[9]
  • February 26, 2026: It announced its acquisition of Fabius, adding experience in AI sales and operational workflows.[6]
  • May 21, 2026: The Anthropic-backed services venture announced its purchase of Fractional AI.[1]
  • July 15, 2026: The buyer introduced the Ode name and identified Fractional's team as its operating core.[2]

What They Built

Fractional's output was customer software. For Zapier, it improved a system that converted API documentation into integration specifications. The work included manually checked reference data, separate measurements for missing and invented fields, and experiments with prompts and model choice. Fractional reports that invented endpoint paths fell from 26% to below 1% in its evaluation set. These are vendor-reported results, without a disclosed sample size.[10]

Braintrust Experiments screen with lines tracking endpoint and field evaluation scores
Braintrust tracked experiment scores for Fractional’s work on Zapier’s specification generator.

Source: Fractional AI

For Cando Rail & Terminals, it built Peter the Safety Agent. The system drafts risk assessments from company materials and helps document expert meetings. Human specialists retain the safety decisions. Fractional reports more than 20 hours saved per assessment; that claim does not establish independently measured safety improvement.[11]

Siegel's AI Engineer Summit talk documents another project: a voice interviewer for consulting research. It shows how controlling question changes and measuring conversational behavior required application logic around the speech model.[12]

Fractional AI CTO Eddie Siegel explains the engineering behind a consulting research voice agent at AI Engineer Summit 2025

Market Position

Target Customers

Fractional served enterprises and private-equity firms evaluating or improving portfolio companies. Its healthcare division named Claritev, Iodine Software, Definitive Healthcare, Beghou Consulting, and Atropos Health as clients.[8][9]

Market Size

A reliable standalone market estimate for Fractional's addressable engagements is unavailable. Its opportunity depended on the budgets and workflows it could reach.

Competition

Anthropic identifies Accenture, Deloitte, and PwC as partners serving large enterprises. Its new venture targeted mid-sized businesses needing hands-on implementation. That distinction explains a proposed customer segment; it does not establish an uncontested market.[13]

Business Model

The founding thesis describes consulting combined with hands-on software development, tailored to each client. It also proposes sharing project risk and upside. Public materials do not disclose a standard fee schedule, the share of contracts using that structure, revenue, margins, or customer concentration.[5]

The due-diligence service extended the relationship earlier in an investor's decision process. Fractional could assess a target's AI exposure before an acquisition, then help build systems afterward. This is a plausible commercial connection. Customer conversion between the services is undisclosed.[8]

Traction

At its January 2026 healthcare launch, Fractional reported experience across more than 50 production deployments. The release includes named customer comments about architecture planning and engineering collaboration. That supports delivery activity, while leaving contract value, retention, and profitability unknown.[9]

Post-Mortem

Delivery relationships shaped the acquisition

The acquisition announcement says Blackstone already knew Fractional through work across its portfolio. Hellman & Friedman also pointed to experience in its portfolio. Those relationships gave prospective buyers evidence of delivery before they bought the firm. The announcement describes Fractional's team and delivery capabilities as the new venture's operating foundation.[1]

Portfolio investors can introduce a services team to multiple businesses. The team can then learn implementation patterns across those engagements. Combining that access with Anthropic engineers could make customers easier to reach and technical problems easier to resolve. This is an interpretation of the structure, not a measured post-acquisition result. Anthropic itself says adoption needs both hands-on engineering and knowledge of each business.[13]

The continuing business changes the lesson

In July, Taylor described the customer problem: “the challenge is making it real.” Ode retained the founders in operating leadership. Fractional's current consultation page directs inquiries to Ode. Its delivery work continues inside a larger services firm.[2][14]

The evidence cannot show whether this structure improves margins or whether its tie to a model provider limits customer choice. Before the acquisition, Fractional publicly described itself as an OpenAI partner and documented work using several model providers. The later Claude-centered venture makes provider choice a material question for its next phase.[6][10][13]

Key Lessons

  • Customer relationships can become acquisition evidence. Fractional's work for Blackstone and Hellman & Friedman portfolio companies preceded their venture's purchase. Delivery gave the buyers a basis for evaluating the team.
  • Measure the specific failure. The Zapier work distinguished invented endpoints from missing fields. That made changes testable instead of relying on a better-looking demo.
  • Keep human authority explicit. Cando's assistant handles preparation and documentation while specialists retain safety decisions. Automating preparation can leave safety decisions with specialists.

Sources

  1. Fractional AI acquisition announcement
  2. Ode launch and operating leadership
  3. Buyer's counsel confirms transaction
  4. Founders' history and recruiting approach
  5. April 2024 founding thesis
  6. Fractional's Fabius acquisition
  7. Claritev collaboration
  8. AI due-diligence service
  9. Healthcare division and deployment claims
  10. Zapier evaluation case study
  11. Cando safety assistant case study
  12. Siegel's AI Engineer Summit talk
  13. Anthropic explains the services venture
  14. Fractional's current inquiry page

Pressure-test this opportunity

Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.