DOBS AI
DOBS AI turns messy data rooms into production-ready financial workbooks. Its agents ingest ledgers, bank statements and payroll files, then run proof of cash, payroll and working capital analyses with every figure traced back to source.
What is DOBS AI?
DOBS AI is an end-to-end platform for financial due diligence. It takes a data room in whatever state the seller left it and returns a production-ready workbook, with live formulas and every figure traceable back to the document it came from. The company positions it as the only platform covering that whole path.
The workflow runs in four stages. Ingestion first: a firm uploads the data room or connects its ERP through a one-click integration, with QuickBooks, Xero, NetSuite and Sage supported alongside direct file upload in any format. Then extraction: the agent indexes every file, identifies gaps, and pulls figures out of messy PDFs, tax returns, general ledgers, bank statements and payroll files, before normalising the chart of accounts, mapping entities and flagging missing documents. Then analysis: proof of cash, payroll, customer concentration and net working capital run automatically, alongside EBITDA bridges, revenue analysis and ARR separation for SaaS targets. Finally iteration: the team reviews and comments, and the agent regenerates workbooks and reports in the firm's own format, as Excel, PDF or PowerPoint.
Two design choices underpin the accuracy claim. The architecture separates roles, since AI extracts while deterministic code calculates, and every mapping carries a confidence score. The platform is built for a scale that breaks spreadsheets: millions of ledger lines, twenty or more bank accounts, several entities per deal.
The primary audience is transaction advisory teams at CPA firms and private equity funds, but the same engine is configured for other verticals shown on the site: T12 parsing and portfolio benchmarking in real estate; lease, CAM and loan covenant monitoring in property management; change orders, closeout and quantity take-offs in construction; carrier invoice audit in transportation; supplier invoice and rebate recovery in retail.
A lighter route sits alongside the platform. The Dobs AI Auditor MCP server plugs into ChatGPT, Claude, Claude Code and Cursor: attach a spreadsheet in the chat, ask in plain English, and the analysis runs on the Dobs platform with a link back to the full workbook. It is free to start and read-only by default. The company was founded by Nikita Komarov, who spent over a decade in advisory, audit and consulting at McKinsey, EY and KPMG.
What it does
- Ingest an entire data room, or connect an ERP through a one-click integration
- Extract figures from messy PDFs, tax returns, general ledgers, bank statements and payroll files
- Normalise the chart of accounts, map entities and flag missing documents automatically
- Run proof of cash, payroll, customer concentration and net working capital analyses
- Produce a traceable Excel workbook with live formulas and a source link on every number
- Regenerate workbooks and reports in the firm's own format after review comments
- Query deal data in plain English and get answers with source citations
When to use DOBS AI / When not to
A quick filter to help you decide if DOBS AI is the right fit.
When to use DOBS AI
- Transaction advisory teams at CPA firms producing Quality of Earnings deliverables
- Private equity funds and acquirers running diligence on mid-market targets
- Corporate finance teams handling internal, vendor and SOX audits
- Small advisory teams that want to take on smaller deals profitably again
- Finance and procurement teams auditing vendor, carrier or supplier invoices at scale
When not to use DOBS AI
- Consumers and private individuals: the service is contractually restricted to business use
- Anyone under 18, who is explicitly excluded from the service
- Teams handling payment card (PCI) or health (PHI) data, which DOBS AI refuses outright
- Buyers who need published pricing and self-serve enterprise onboarding
- Anyone looking for a mobile app or a browser extension, neither of which exists
How to use DOBS AI
A typical end-to-end flow, from setup to results.
- Book a walkthrough on the site: demos are run on your own engagement data, not on sample files
- For the enterprise platform, sign the written agreement (MSA, SOW and DPA) that governs access
- Upload your data room, or connect QuickBooks, Xero, NetSuite or Sage through the one-click ERP integration
- Let the agent index every file, classify the documents and flag what is missing
- Review the normalised chart of accounts and entity mapping, using the confidence score shown on each mapping
- Run the standard analyses: proof of cash, payroll, customer concentration and net working capital
- Check the generated workbook, where every number links back to its source document
- Send feedback and adjustments; the agent regenerates the workbook and reports in your firm's format
- For the lighter route, create a free account on app.dobs.ai and generate an API key under Settings, then API Keys
- Add the MCP server URL to Claude, Claude Code, Cursor or ChatGPT, attach a spreadsheet and ask your question
Pros & Cons
Pros
- Every figure links back to its source document, and the workbook keeps live formulas
- Dual-path verification separates AI extraction from deterministic calculation
- Built for genuinely messy data: millions of ledger lines, twenty or more bank accounts, multiple entities
- Serious security posture: private VPC, SOC 2 Type I, AES-256 at rest, TLS 1.2 or above in transit
- No training on customer data, with pre-trained models isolated from public AI models
- A real free entry point through the MCP connector, with no credit card required
- Broad integrations across finance, property and construction systems, with no rip-and-replace
Cons
- No published pricing at all for the enterprise platform, and no pricing page on the site
- Platform access requires a negotiated written agreement, with three to twelve weeks of implementation
- SOC 2 Type I only, which attests controls at a point in time rather than over a period
- The site contradicts itself: three vertical pages claim SOC 2 Type II where the security page says Type I
- Terms and privacy policy still describe an earlier accounts payable recovery business, not due diligence
- No support email is published, and the footer's Careers and Contacts links are inactive
- Fully JavaScript-rendered site, with no about, contact, pricing or features page and no mobile app
Pricing & Plans
DOBS AI publishes no prices. A permanent free tier does exist: an account can be created on app.dobs.ai and the first analyses run through the MCP connector at no cost and without a credit card. Paid plans are described only as taking over once volume grows, with no amount, currency or plan name disclosed. The enterprise platform is not sold self-serve at all, since access is granted exclusively under a written agreement, so its pricing is negotiated case by case.
- create an account on app.dobs.ai
- connect the MCP server and run the first analyses at no cost
- unnamed and unpriced
- described only as taking over when volume grows
- contractual only
- governed by a Master Services Agreement
- a Statement of Work and a Data Processing Addendum
- Quality of Earnings delivery through partner CPA firms
- powered by DOBS technology
- Internal
- vendor and SOX audit configuration for corporate finance teams
- three deployment packages
- each including private deployment
- security setup and dedicated support
- none priced
Data, GDPR & hosting
A consolidated view of how DOBS AI handles your data.
GDPR overview
The company is based in the United States and addresses the GDPR explicitly, without ever claiming certification in so many words. The privacy policy sets out European rights for EEA, UK and Swiss residents, covering access, rectification, erasure, restriction, objection and complaint to a supervisory authority, states the legal bases relied upon, and commits to a 30-day response for GDPR requests against 45 days for CCPA. International transfers are covered by Standard Contractual Clauses where required. A Data Processing Addendum forms part of the enterprise contract, and privacy@dobs.ai is the single point of contact. No Article 27 EU representative is named anywhere on the site. People whose data is processed on behalf of an enterprise customer are redirected to that customer. The policy is version 2.0, effective 27 April 2025.
Who owns the data?
DOBS AI works under two distinct regimes. For its public website it acts as a controller. For everything processed on behalf of enterprise customers it acts as a processor, and that processing is governed by the signed Master Services Agreement and Data Processing Addendum rather than by the website privacy policy. Customers keep their material: files analysed through the connector stay in the customer's own Dobs workspace, every workspace is isolated per tenant, and the customer can delete content at any time. A private VPC deployment is offered so that data never leaves the customer's own perimeter. The website, its content and the software itself remain the property of Dobs.
Reuse rights
DOBS AI states that it does not train its AI models on customer-specific data, and that it runs pre-trained models only, fully isolated from public AI models. Service improvements may draw on de-identified, aggregated data that cannot be linked back to any customer, and only within the limits of the signed agreements. Website data serves a narrow purpose: answering business enquiries, sending information to contacts who asked for it, first-party analytics and security. The site uses essential cookies and Google Analytics only, with no third-party advertising cookies and no behavioural advertising, and states that it has neither sold nor shared personal information in the preceding twelve months. On the MCP connector, tools are read-only by default and anything that creates or changes data is separated and opt-in.
Data retention & training
Hosting summary
DOBS AI is based in the United States and states that information may be transferred to and processed there, or in other countries where its providers operate, with Standard Contractual Clauses applied where required. The privacy policy names its subprocessors: Microsoft Azure, including Azure OpenAI and Form Recognizer, for AI inference; Amazon Web Services for hosting, storage and compute; and AWS RDS running PostgreSQL for the database. Data is encrypted with AES-256 at rest and TLS 1.2 or above in transit, held in data centres with controlled physical access, and every access is logged. For customers who need tighter control, the platform can run inside their own private VPC so that data never leaves their perimeter, and the SMB offering is deployed on the customer's own infrastructure. One point deserves care: the public website resolves to an AWS node in Singapore, which reflects where the marketing site is served and says nothing about where customer data is stored. No hosting country beyond the United States is named.
Things to keep in mind
Risks and trade-offs to weigh before adopting DOBS AI.
- Automated workbooks can breed misplaced confidence: the vendor itself insists that CPAs remain responsible for judgement, and recommends human review on construction take-offs
- Performance claims such as 99%+ accuracy, 40% more deals and twice the speed are vendor assertions, not audited figures
- Client testimonials are anonymised by role and firm type, so none of them can be independently checked
- Junior analysts learn the trade on the reconciliation work this tool removes, a displacement the site itself raises as a discussion topic
- Data ownership sits in the contract rather than in the public privacy policy: what happens to your files depends on an MSA and a DPA you have to read
- The only postal address published is a Delaware registered agent, which says nothing about where the company actually operates
- Payment card and health data are refused, so check compatibility before pushing sensitive engagements through the platform
Setup & Integrations
Technical difficulty
Two very different levels. The MCP connector is trivial: create a free account, generate an API key, paste one server URL into Claude, Claude Code, Cursor or ChatGPT, and you are running in about two minutes with nothing to install. The enterprise platform is a project rather than a setup: a written agreement first, then a one-click ERP connection or bulk upload. The vendor quotes three to four weeks to first recoveries on invoice audit work, four to six weeks for bank reconciliation and eight to twelve weeks for full transaction advisory automation. No existing system needs replacing.
Deployment
Integrations
Supported languages
Behind DOBS AI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does DOBS AI actually do?
Who is it built for?
How much does it cost?
Is my data used to train the models?
What security guarantees are offered?
Which systems does it connect to?
How do I use it from ChatGPT or Claude?
How long is my data kept, and what is refused?
Should you pick DOBS AI?
DOBS AI is a deliberately narrow tool, and that is its strength. It does not try to be a general assistant: it takes the most mechanical part of financial due diligence, ingesting and reconciling a data room, and returns a workbook a partner can actually review. Two things separate it from a generic document extractor: the traceability of every figure back to its source, and the split between AI extraction and deterministic calculation. In a discipline where a wrong number becomes a seven-figure mispricing, both matter.
The security posture is serious for a company this young: private VPC deployment, SOC 2 Type I, AES-256 at rest, named subprocessors, and a clear statement that customer data never trains the models.
The reservations are real. Pricing is entirely opaque on the enterprise side, where access requires a negotiated agreement and implementation runs from three to twelve weeks. SOC 2 Type I attests controls at a point in time, not over a period, and the site contradicts itself: three vertical pages claim Type II where the official announcement and the security page say Type I. The terms of service still describe an earlier accounts payable recovery business, which suggests the legal documents have not caught up with the product. Performance figures are vendor claims, testimonials are anonymised by role, and the only published address is a Delaware registered agent.
The verdict: a credible, well-built product for CPA transaction advisory teams and private equity buyers who already know they have a data-cleaning problem. Try the free MCP connector first, since it costs nothing and shows what the engine does on a real spreadsheet, before entering a sales conversation you cannot budget for in advance.
- Choosing a selection results in a full page refresh.
- Opens in a new window.