Forensic technology consulting for law firms

Technical evidence,
made clear.

We examine the documents, messages, systems and money trails at the center of a dispute. Then we tell you in plain English what they prove, what they don't, and what would settle it.

Read-onlyHash-verifiedAI at scaleRepeatablePlain English
01Matters

What's in front of you?

Pick the situation closest to yours. Each one shows what we examine and what you get back.

02Capabilities

The whole technical record, under one roof.

Disputes rarely stay in one format. A signature leads to an email. The email leads to a bank record. The bank record leads to a server log. We follow the evidence across all of them.

A

Documents & signatures

  • E-signature certificates and event histories
  • How each signature was made and how each signer was verified
  • Native originals vs. re-printed copies
  • File and PDF metadata (hidden data inside a file, such as when it was made)
  • Page-by-page contract comparison, typed vs. handwritten terms
  • Delivery records checked against what the contract required
B

Messages & media

  • Screenshot and video authentication
  • OCR (software that reads text in images), then review by eye
  • Clock, battery, date and screen-size consistency checks
  • Mapping files to the devices they came from
  • Flags on sensitive data such as GPS locations and patient details
C

Email, money & people

  • Business email compromise (BEC): a criminal poses as a trusted sender to redirect a payment
  • Fake vs. real email comparison: signature blocks, sender details, missing warnings
  • Large e-mail reviews, with every quote checked against the original
  • Bank wire paperwork review
  • Digital-currency tracing on the public blockchain, wallet by wallet
  • Financial-regulator registers and company formation and dissolution records
D

Software & systems

  • Read-only preservation of live servers, including server-clock corrections
  • Compiled programs (machine code) checked for hidden logic, such as a license "kill switch"
  • Delivered software compared with its own build instructions
  • Rebuilding transactions from raw logs, including FIX trading messages
  • Matching records against outside statements
  • Source-code origin and open-source license history
E

Litigation support

  • Plain-English reads of complaints, motions and cited opinions
  • Evidence maps that rate each issue: proven, only a lead, needs a subpoena
  • Technical subpoena schedules and preservation-letter targets
  • Another firm's report checked finding by finding against the records
  • Briefing papers for status conferences
  • Methodology statements another expert can reproduce

Not on this list? Ask. Most disputes now run through software, and software leaves records.

Describe the problem
03AI at scale

Agentic AI for the records no team can read by hand.

Modern disputes come with tens of thousands of e-mails, years of logs and whole codebases. TNT builds agentic AI systems: software agents that read, sort and cross-check records in parallel. On a matter, the agents carry the volume. The findings are still checked against the original records.

01

E-mail and document review

Read and sort large mailboxes and document sets. Pull out people, dates, money and topics. Build timelines of who knew what, and when.

02

Deep data mining

Search logs, transaction records and databases for patterns, gaps and things that don't fit: the trade that doesn't reconcile, the odd login, the clock that was moved.

03

Public-records research

Search regulator registers, company formation and dissolution records and internet-number registries, then connect what they show.

Related: Case 02
04

Money tracing

Follow funds through bank records and across the public blockchain, step by step. Identify the records needed to confirm where the money ended up.

05

Testing the other side's report

Check another expert's report claim by claim against the original records. Show what the record supports, and what it doesn't.

06

Software and code review

Map an unfamiliar codebase or program. Find hidden logic, such as a license check. Trace where the code came from.

Built for the matter, not off the shelf.

When the evidence needs it, we build an agent workflow for that one evidence set. Each step is logged and described in the methodology, so another expert can follow it. This is the same agent engineering behind Inner Hive, TNT's multi-agent platform for law and finance.

innerhive.ai

Agents do the reading.
The facts get checked.

  • AI is a tool, not a source of facts.
  • Key figures are re-checked with standard tools that do not use AI.
  • Quotes are checked word for word against the original.
  • Each report lists its checks, so anyone can repeat them.
18,000+e-mails reviewed, every quote checked against the original
196findings in another firm's report, each tested against the records
200,000+trades rebuilt from raw logs and matched to broker statements

From one recent matter.

04Cases

Two recent matters.

Anonymized on purpose. These matters are confidential, and some are still pending. No outcome is claimed.

Case 01Matter pending

E-Signature & Screenshot Forensics

Was the signature real?

A law firm needed to know what the records behind a disputed e-signature really showed. In one business day, we examined the signing records and 71 screenshots and videos. We delivered two written reports that counsel could act on.

  • Litigation support
  • Document forensics
  • Metadata
turnaround
1business day
images and videos reviewed
71
files hashed before and after — 0 changed
76
Case 02Matter pending · work ongoing

Wire Fraud at a Home Closing

What can the evidence prove, and what needs a subpoena?

A family lost a six-figure home down payment to a fake "closing instructions" email. We turned a scattered case file into a clear evidence map, a subpoena plan and a plain-English brief that counsel could act on.

  • Business Email Compromise
  • Bank Wire Fraud
  • Litigation Support
  • Evidence Mapping
page motion to dismiss and 14 cited opinions, cut to a short brief
180
forensic work plan, ranked by priority
8parts
plain-English work-product reports
2
05Method

Four steps, every matter.

The method is written down before the work starts, so another expert can check it — and get the same numbers.

01

Preserve

Read-only collection. Every file gets a SHA-256 fingerprint before and after. Tools are checked against the publisher's checksum. Every step is logged — the chain of custody.

02

Examine

Line-by-line review of records, logs, images and code. Findings are checked against outside records, such as broker statements and public registries.

03

Verify

Every finding is run again from sealed copies of the evidence. AI speeds up reading. It never stands in for a fact: key figures are re-checked with standard tools that do not use AI.

04

Explain

Plain English, written for a lawyer. What the records prove, what they don't, and what would settle it — with a methodology statement to reproduce every finding.

Standards the method is written against
NIST SP 800-86U.S. federal guide to using forensic techniques on digital evidence.
ISO/IEC 27037International standard for finding, collecting and preserving digital evidence.
SWGDEBest-practice guidance from the Scientific Working Group on Digital Evidence.
FRE 702 / 901Federal Rules of Evidence on expert methods and on authenticating evidence.

ExifTool·Python·SHA-256·Tesseract OCR·ffprobe·PyMuPDF·pikepdf·ARIN lookups·FIX log analysis·Linux server forensics

When a question needs a paper original or an outside examiner, the report says so — and recommends one. Independence keeps the findings defensible.

06Who

A career building the systems that now end up in evidence.

James Tillinghast
Founder & Principal
James Tillinghast

James leads every engagement. Three decades shipping software, the last fourteen inside the AI stack — Microsoft, Microsoft Research, and Meta.

E-signature platforms, messaging apps, cloud mailboxes, trading software: the systems at the center of modern disputes are the kind of systems James spent a career building. That is the advantage. The record gets read the way the people who built it would read it.

Full background at tillinghast.tech
1995 – 2013
The Web YearsInteractive media, advertising, retail, e-commerce, mobile
2013 – 2016
MicrosoftAzure enterprise data platform
2016 – 2020
Microsoft ResearchComputer vision, Cognitive Services, applied ML
2020 – 2025
Meta PlatformsGenerative AI for the advertiser stack
2024 – now
TNTAI systems, software, data — and forensic work for counsel
07Engage

Three ways to start.

Rapid read

A fast first look

What the technical record shows, what it doesn't, and what to get next. One recent matter: two written reports in one business day.

Consulting expert

Ongoing support for counsel

Work under counsel's direction as attorney work product: evidence handling, analysis, discovery planning and pleading review.

Full reconstruction

Large-scale rebuilds and reviews

Logs, transactions, timelines and large e-mail sets, with AI agents carrying the volume. You get data files and a methodology another expert can reproduce.

Starting is simple: send the party names for a conflict check. Please hold privileged material until we're engaged.

08 / Next step

Tell us what's in front of you.

A short call is enough to scope it: what you have, what you need to prove, and how fast you need it.

Email to discuss a matter