Adversarial Intelligence

Off-the-shelf AI tools make every analyst faster. They don't make the analysis right. That is what we engineer.

Most institutions have plugged generic AI tools into consequential workflows and are hoping the output holds. We take a different approach. We build purpose-built AI systems around your most critical processes, deploy them inside your infrastructure, and hand them over permanently.

Request a Private Briefing See Use Cases →
Bespoke

Built around your workflows, not a generic tool configured with prompts.

Sovereign

Deployed inside your infrastructure. Your data never leaves your environment.

Permanent

You own it outright. Not a subscription. A capability your organisation keeps.

Defensible

Every output is traceable to source. Built to satisfy institutional audit requirements.

Use Cases

Starting in institutional finance.
Where the stakes are highest.

The workflows below are where we are building first. These are workflows where AI currently falls shortest and where the cost of a missed signal is most consequential. This is not the limit of where the architecture applies.

Private Equity · Pre-Deal

Pre-Deal Due Diligence

Adversarial review, programmed to find reasons to fail the deal first, across 100% of the data room. Every contradiction surfaced before any investment case is built.

See the use case →
Fund Management · Ongoing

Portfolio Monitoring

Portfolio company performance tracked against the investment thesis. Leading-indicator alerts for EBITDA variance, covenant drift and credit stress before the reporting cycle, across the full portfolio, not just flagged positions.

See the use case →
Credit · Underwriting

Credit Underwriting

Structured credit analysis with full audit trail. Built to your credit committee format. Traceable from source to every conclusion.

See the use case →
Compliance · KYC / AML

KYC Review

Beneficial ownership traced through complex corporate structures across jurisdictions. Inconsistencies in disclosed structures surfaced automatically, with a full audit trail built to compliance requirements.

See the use case →
The Methodology in Practice
Same data room. Two documents.
Neither wrong in isolation.

Together, they described a year of full cost exposure with zero revenue.

Capital Costs Schedule
Commissioning completes
Q1 Year 3.
Internally consistent. No flags raised.
Financial Model
Revenue starts
Year 4.
Internally consistent. No flags raised.
Cross-Document Gap
71% → negative
Base case IRR → Capital at risk
Detection Time
< 1 hour
Review Type
Phase 2 diligence pass
First-pass adversarial review
Recommendation
FLAG
Do not proceed. Management team review required.
Pre-revenue GP stake · Anonymised · Phase 2 adversarial review · $5M investment decision
Request a Private Briefing →
What your LPs are already asking
"Describe your AI strategy for due diligence. Is it auditable? Independently verifiable?"

LP DDQs now routinely ask whether AI-assisted analysis can be audited, explained, and independently verified. This is not a future requirement. It is being asked on live mandates now.

Not an answer

"We use an off-the-shelf AI tool."

An answer

"We run 500+ verification checks against every data room, with a full cell-level audit trail built to IC review requirements."

The gap between capable and defensible. That is where we work.

If your current AI output isn't holding up to institutional scrutiny, we should talk.

Request a Private Briefing
What We Build

Three types of work.
Each with different demands.

The workflows that matter most in financial services each present a distinct intelligence challenge. We engineer for all three.

The Adversarial Engine

Interrogate before you trust

Adversarial means the system is programmed to find reasons to fail the deal first, before any investment case is built. Every contradiction is surfaced and must be accounted for or disclosed. No conclusion is drawn from unverified inputs.

  • Pre-deal due diligence
  • Credit underwriting
  • Completion accounts review
  • Fraud and misrepresentation detection
The Continuous Monitor

Surveil across entities and time without stopping

Ingest across entities, instruments, and time. Surface what each source reveals about the others. Continuous, not periodic. Designed for workflows that cannot wait for the next reporting cycle.

  • Portfolio and loan book monitoring
  • Post-merger integration
  • Covenant and condition tracking
  • Regulatory reconciliation
The Institutional Synthesiser

Build the output your committee can defend

Build the document your committee expects, structured to your output conventions, with a full audit trail from source to every conclusion. Nothing asserted without traceable derivation.

  • IC memo and investment note generation
  • Credit committee reports
  • Board and risk committee submissions
  • Regulatory filing documentation
Full Auditability

Every conclusion has a source.
Every source has an address.

Every finding in a Refoveo output is traceable to the exact source document, page, and cell it was derived from. Your team can navigate from any conclusion to the primary source directly.

This is not a summary with footnotes. It is a navigable decision chain built to satisfy institutional audit requirements, allowing your committee to interrogate every finding independently.

If a finding cannot be traced, it does not appear.

The Engagement Model

Prove it first.
Build it for you. Own it permanently.

Every engagement follows the same three phases, designed to demonstrate value on your real data before any commitment, then build the right system around how your organisation actually works.

Phase 01: Forward Deploy

Prove it on your data

We run the full methodology against two of your closed deals. Your team sees exactly what it surfaces before any commitment is made.

Phase 02: Co-Innovate

Build it around you

A specialist embeds with your team. The intelligence system is calibrated to your workflows, your output formats, and how your organisation actually makes decisions.

Phase 03: Institutionalise

Own it permanently

Deployed inside your infrastructure, operated by your team, owned outright. You own the capability permanently, while we keep the underlying AI current as the technology advances, so every improvement is a free upgrade to what you own.

Request a Free Backtest
Security and Sovereignty

Your data never leaves.

Refoveo is deployed inside your Virtual Private Cloud. Your data, your workflow logic, and your proprietary calibration never touch a public network at any stage of the engagement.

VPC DeploymentRuns inside your infrastructure. Not a shared cloud.
Zero RetentionNothing stored or processed outside your VPC boundary.
No Model TrainingYour data does not train any public model. Ever.
Regulatory AlignmentDesigned against FCA and SEC AI governance frameworks.
Full Audit TrailCell-level decision chain built for institutional compliance review.
Private Equity · Pre-Deal

Due Diligence

The greatest risks in a data room are not in any single document. They are in the contradictions between documents. A construction schedule and a revenue model. A management account and a disclosure letter. Standard review processes are not designed to catch what is mutually contradictory across hundreds of files.

We run an adversarial pass across 100% of the data room before any synthesis begins. Every contradiction is surfaced and must be accounted for. The IC first draft is generated in the same session, structured to your format, traceable to source.

What you receive
  • Full-coverage contradiction analysis across the data room
  • Prioritised risk register with source citations
  • IC memo first draft, structured to your committee format
  • Cell-level audit trail from finding to source document
Request a Free Backtest →
Live Finding
71% → 0
Projected IRR, destroyed
£60–120M
Dead carrying costs
9 months
Hidden exposure window

A 9-month gap between construction completion and revenue commencement: invisible to standard review, visible to adversarial analysis. Confirmed before capital was committed.

Portfolio Management · Ongoing

Portfolio Monitoring

Quarterly reports tell you what happened. By the time covenant drift or credit stress appears in a report, the window to act has often closed. You need to know what is developing, across the full portfolio, not just the positions your team has flagged.

We build continuous surveillance across your portfolio, ingesting management accounts, covenant schedules, and operational data as they arrive, surfacing leading indicators before they become problems.

What you receive
  • Continuous alerts on covenant drift and credit stress signals
  • Cross-portfolio view updated as new data arrives
  • Structured monitoring reports for IC and board review
  • Full audit trail of every alert and its data source
Request a Private Briefing →
Banking & Credit · Underwriting

Credit Underwriting

Credit decisions require structured analysis that a credit committee can interrogate and a regulator can audit. Generic AI tools produce confident-sounding summaries that cannot be traced to source, creating institutional risk rather than reducing it.

We build credit analysis systems calibrated to your underwriting standards, structured to your credit committee format, with every assertion traceable to the underlying data.

What you receive
  • Structured credit analysis built to your committee format
  • Risk identification with source citations across all input documents
  • Covenant and condition summary with cross-document checks
  • Full audit trail for regulatory review
Request a Private Briefing →
Public Case: NMC Health
$4bn+
Undisclosed debt, missed by standard review
FTSE 100
Audited without detection for years

Supplier financing structured as trade payables. Related-party transactions arranged to avoid balance sheet recognition. Inconsistencies between facilities disclosed to lenders and those disclosed to auditors. The mechanisms were visible to an engine designed to look for them.

Compliance · KYC / AML

KYC Review

KYC reviews require tracing beneficial ownership through complex corporate structures, spanning jurisdictions, entities, and document types that do not always tell the same story. Standard processes rely on analysts manually cross-referencing documents that were never designed to be compared against one another.

We apply the same adversarial cross-document methodology to ownership and compliance review, surfacing inconsistencies in disclosed structures, flagging documents that contradict each other, and producing a fully traceable output built to satisfy compliance requirements.

What you receive
  • Beneficial ownership mapping with full source trail
  • Cross-document inconsistency flags across disclosed structures
  • Structured output built to your compliance team's format
  • Full audit trail for regulatory review
Request a Private Briefing →

These are the workflows where we are building first, not the boundaries of what we can build. If your workflow is not listed, that is the conversation we want to have.

Request a Private Briefing
The Founders
Ash Mondal
Co-Founder

14+ years in institutional finance, capital markets and private markets. The methodology is built directly from that experience.

Marek Piskorz
Co-Founder

Risk and change executive with over 30 years of leadership experience in investment and international banks. NED, board advisor, and active investor.

Why institutional partners work with us

Verified proof points

Every figure we publish is sourced to the actual data it came from. The real finding is more compelling than an inflated one, because it is true and verifiable.

Sovereign architecture

Your data stays inside your environment at every stage. No public network exposure. No shared model training. No exceptions.

Capability, not dependency

Every engagement is designed to build permanent capability inside your organisation. We are explicit about what we commit to, and we hand it over.

All submissions treated as confidential.

Not ready for a call?

Subscribe to The Verification Layer: thinking on intelligence engineering and institutional AI, published when it is ready.


Direct contact

ashutosh.mondal@refoveo.ai
marek.piskorz@refoveo.ai