Snowstone

More than 10,000 agents, by Snowstone's count, read a company's email, chat, documents and system exports in parallel, match the same customers, vendors and people across systems, and rebuild each process from its transactions, approvals and queues. Take the models away and what remains is a set of read only connections to data nobody has read.

Every connection is read only, and the service cannot create, change or delete anything in a connected system or send from it. Each citation and calculation is checked against its source, a reviewer examines anything still open, and staff review the final report by hand before delivery. Partners and management decide which findings to fund and who owns them, and an optional decision log records their reasons and the later results beside each finding.

Every number and statement links to the file version and period it came from, interview notes and professional judgments are labeled apart from system data, and expected sources that are missing or restricted are listed. When two sources disagree, both are kept with the evidence for each, and a finding the data cannot settle is marked open with the source that would settle it. Estimates stay as ranges where the data will not support a single figure, and no accuracy rate for findings or estimates is published.

Snowstone reports more than $2 billion of transaction value reviewed and more than $50 million of value creation opportunities found. Its examples carry dollar ranges, such as 1,400 trade confirmations a day faxed and retyped by hand at an estimated $216,000 to $303,000 a year in staff time, but no client, fund or person is named, so every figure is self reported.

Each engagement runs in its own single tenant environment, data from different clients is never combined, and client data trains no model, Snowstone's or anyone else's. The environment, copied source data, derived data and credentials are deleted within 30 days of an engagement ending unless the client instructs otherwise in writing, security logs are kept for up to 12 months, and only anonymized service metrics such as processing times and error rates outlast the work. Comparisons across portfolio companies and process maps that carry over to the next acquisition depend on material kept between reviews, and what form that material takes is not described.

For data in connected systems Snowstone acts as a processor under each client's written agreement, and the client carries the lawful basis for processing and the notice owed to employees, customers and vendors whose details sit in those systems. Scope is set by system, entity and date range and recorded, automatic filters can strip sensitive files and email, and requests from individuals are forwarded to the client. Transfers out of Europe, the United Kingdom and Switzerland rely on Standard Contractual Clauses and the UK International Data Transfer Addendum.

Both an ISO 27001 certification and a SOC 2 Type 2 report are listed as in progress, so Snowstone holds neither today. Controls listed for its Vanta trust center include intrusion detection, central log management, annual risk assessments and encryption in transit, with encryption at rest added in its privacy policy, and an NDA and data processing agreement are signed before any data is shared.

Analyzing how a portfolio company operates calls for no license, and Snowstone holds none. Its privacy policy commits that QuickBooks Online data serves only the client that authorized the connection and that Google data follows the Google API Services User Data Policy, including its Limited Use requirements, which are platform commitments rather than supervision.

Findings describe processes, roles, incentives and systems, and Snowstone commits to never assessing the performance of named individuals, the line that matters most for a tool that reads a whole company's email. Its deliverables still point at people through workload and dependency, since the organization map shows who work actually routes through and key person findings show what stalls when one employee is away. Bias and accuracy testing of its own models is unpublished.

The published terms cover the website alone, cap liability at $100 and call worked examples illustrative rather than a warranty of results. Engagement terms sit in a separate written agreement that is not public, so the remedy when a sized synergy or an EBITDA estimate proves wrong is unknown. Insurance against data leakage is offered where applicable and covers exposure of data, not analytical error.

Anthropic and OpenAI supply the language models, OpenAI also provides text embeddings, both process in the United States, and Microsoft Azure hosts the engagement environments. Both model providers work under zero data retention terms that bar training on client data, Snowstone commits to updating its provider list before adding a new one, and clients can run local models on their own infrastructure instead. Model versions are not named.

Read only connections cover email, chat, file storage, ERP, HR, accounting and industry software, with QuickBooks Online, Microsoft 365 and Google Workspace named and sign in through Microsoft Entra ID. The same customer under different names in accounting, sales and contracts is matched across systems, and each proposed match is shown with its confidence and source. Output arrives as workbooks and board decks in the fund's template, and no API or developer documentation is published.

Clients choose between a dedicated Snowstone cloud machine, their own cloud account, or their own servers for the strictest rules. Engagement environments run on Microsoft Azure in the United States or another region agreed with the client, and the standard setup needs no network or firewall changes and one to four hours of the client's IT time.

Pricing sits behind a sales call booked through a Calendly form. Reviews come in three sizes, from a read of the data room and core software to full read access across the company, each scoped to one portfolio company, and none carries a published price, fee basis or engagement length beyond the one week target.

Private equity operating partners, value creation teams and portfolio operations leads are the main buyers. Portfolio company chief executives, finance and operations chiefs and technology leaders get their own material, including executives new to the role, and operations, turnaround and value creation consultants can run Snowstone behind their own engagements in their firm's template. Buy and build platforms get a dedicated integration workflow, while fund size, sector and strategy are not separated.