Forensic Failure Analysis
Client sign-in
Forensic failure analysis, on demand.
AswanAI gives every workshop and dealership the forensic judgement of Patrick Swan, a failure analyst with 50 years of experience, for engines, gearboxes, differentials and other mechanical components. It works the case as a team with whoever is on it. It leads the investigation the way he would: asks what he would ask, says what to photograph, look at and measure, and reasons from what comes back. It delivers a clear, defensible report the same day, marked for his review.
An example of the report AswanAI produces, based on a real turbocharger investigation, shortened, with fictitious names and numbers.
The expert behind it
AswanAI is built on the forensic practice of Patrick Swan, founder of Aswan Consulting, who has investigated failed engines, gearboxes and other machinery for trucking, transport, mining and power generation.
Fifty years in failure analysis; the last thirty documented case by case: the archive AswanAI reasons from.
- Has served as President of the South African Institute of Tribology
- Author of “Studies in plain bearing failures”, Tribology & Lubrication Technology (STLE), October 2006
- Has lectured at the University of the Witwatersrand in lubrication, tribology and the failure analysis of materials and mechanical systems
- Has served as a technical expert in legal cases
- Patrick Swan on LinkedIn
How it works
You talk to it the way you would talk to an expert, and it leads the investigation from there.
Describe the failure. Photograph the evidence.
The person on the case explains what happened in plain language, by typing or speaking. AswanAI says exactly which photographs it needs, and asks for a better one when a picture is not clear enough to rely on.
AswanAI leads the investigation, with you.
It asks the questions the expert would ask, tells you what to look at, measure and photograph, weighs every possible cause against the evidence, and checks its thinking against 30 years of the expert's past cases.
You receive the report.
A full report in the expert's own house style, with the cause, the evidence, the causes ruled out and the photographs that show it, plus a one-page summary for managers. An automatic reviewer also looks over the finished pages. See a sample report.
What sets it apart
Built so that the person holding the report can understand the failure and stand behind the finding.
A skill you suddenly have
You and AswanAI build the case together. It directs the investigation and does the forensic reasoning; you bring what only someone at the part can see, feel and measure. It works with whatever failure-analysis ability you start with and raises it, a lot: a seasoned analyst gets a sharper, more consistent investigation, and someone new to failure analysis can run a valuable investigation from day one, learning the method as they go.
Photographs checked, not guessed
AswanAI studies every photograph, and where one is not clear enough to rely on, it asks for a better one rather than guess.
A picture of the failed part
AswanAI draws a simple diagram of the failed part and pins every photograph and finding to the place it belongs, so anyone can follow the case at a glance.
Thirty years of cases behind every answer
While it works, AswanAI searches the expert's 30-year archive of investigations and training courses for similar failures, and tells you what it found and where.
Nothing quietly dropped
Before a report is made, the conclusion is tested against every fact on the case. Any fact it cannot explain must appear in the report as an open question.
Two reports, ready to send
A full report with labelled photographs for the file, and a one-page summary for managers. An automatic reviewer also looks over the finished pages (photographs, labels and layout) and points out anything that needs fixing.
Speak or type
Technicians can speak instead of typing and have answers read aloud, which is practical on the workshop floor.
Work a case together
Running a case from head office and missing a photograph? Invite a technician or workshop manager anywhere in the country. AswanAI gathers exactly what it needs from them and hands the investigation back. Every contribution is recorded against its author.
A complete record
Who opened, contributed to, requested and approved each report, and why its content changed, is kept permanently. Built to discourage outside influence and to survive scrutiny.
Your company in control
Each company has its own private instance at its own address. Your own administrator adds and removes users, cases can be kept in folders, and users can reset a forgotten password themselves.
On computer and mobile
Available on both: use it in the web browser on a computer, or on a phone, where it can sit on the home screen like any other app. There is nothing for your IT team to set up or maintain. It needs only an internet connection.
For manufacturers and dealer networks
Defensible warranty adjudication
Disputed failures decided on evidence, with reports built to be reviewed, endorsed and stood behind. That is the difference between conceding a claim and defending it.
Expert capacity, everywhere
Senior forensic capability in every dealership at once: no scheduling, no travel, no scarcity, no retirement.
Everyone on the case, uplifted
Whoever uses it is raised from where they start: a dealer principal, a workshop manager, a foreman or a technician. Each investigation is a working masterclass in failure analysis. Fewer misdiagnosis repairs, fewer comebacks, faster escalations done right.
Institutional memory that compounds
Every case recorded, searchable and consistent across the network. Expertise that stops walking out of the door.
The hard problem has never been opinions. It is reports defensible enough to stand on. That is what this was built to produce.
Where this goes In development
Failure analysis is the beginning, not the destination. These are planned, not yet available.
The workshop companion
Integrated with a manufacturer's training and maintenance documentation and model data (how each unit is assembled, its parts and fasteners, torque settings and special tools), the platform will guide a technician through disassembly without damaging parts or contaminating evidence, and through correct reassembly: the right tools, the right torques, single-use bolts flagged.
Training, delivered where the work is
The same app that runs the investigation will become the trainer: model-specific knowledge and procedure in every technician's pocket, raising workshop knowledge, productivity and efficiency across the network.
Due diligence, answered
The questions procurement, IT security and legal teams ask, answered plainly. Further documentation is available on request.
Where is our data stored, and who can see it?
Each customer runs on a dedicated, isolated instance at its own address, with its own data store and its own operating-system-level separation from every other customer. Your cases, photographs and reports live in your instance only. Access is controlled by your own administrators through per-user accounts; AswanAI personnel access systems solely to operate and support the service.
Hosting is on servers in the United Kingdom, chosen for infrastructure quality and data-protection maturity. All traffic is encrypted in transit (TLS, with strict transport security enforced).
Is our data used to train AI models?
No. Case analysis uses leading AI providers under commercial API terms that contractually exclude training on customer data. Your data is never sold, shared with other customers, or used to improve public models. The knowledge base the system reasons from is the documented case archive of Patrick Swan's practice, thirty years of investigations, not customer data.
Can we record VINs and other confidential identifiers, and where do they go?
Yes. A dedicated “Case & vehicle details” field captures the owner or customer name, VIN, registration, chassis, engine and fleet numbers. These values are held in your instance and are never sent to any AI: not the investigating model, the report writer, nor the image analysis. They are used for one purpose: to appear on your own report. The analysis itself works only from a redacted technical description of the failure, which carries no identifying details. So the real identifiers can appear on the finished document without ever leaving your instance for a third-party model.
How does the platform handle personal information (POPIA)?
By separating the identifiers from the analysis. Owner and vehicle details (VIN, registration, chassis, engine and fleet numbers) are entered in a dedicated field, kept in your instance, and never sent to any AI model; they are used only to print on your own report, and can be left out entirely. Everything the AI reasons from is a redacted technical description that carries no identifying information. Case data belongs to the customer, is processed only to provide the service, and is deleted from the instance on written request.
How is access controlled?
Per-user accounts created by your administrators via single-use invitation links, with no self-registration. Your own company administrator can invite and disable users and issue password-reset links, and users can reset a forgotten password themselves by email. Failed-login rate limiting at the application layer plus network-level intrusion banning. Session security follows current web standards (secure cookies, HTTPS-only). Single sign-on integration is on the production-rollout roadmap.
How do you know the AI's conclusions are right?
Every AI model is tested before we use it. It works real cases from the expert's archive that it has never seen, and its conclusions are scored against the expert's own verdicts. A new model is adopted only when the results support it.
In daily use the system is designed to say what it does not know: it asks for missing evidence rather than guess, every statement shows whether it was seen in a photograph, reported or reasoned, each conclusion is tested against every fact on the case before a report is made, and reports are issued pending expert review.
Can decisions made on these reports be defended?
The entire system is built for defensibility: every material claim in a report carries its evidential basis (seen in a photograph, reported by a named party, or reasoned inference); what the user reports is recorded as their report, separate from what the photographs show, so a reviewer can see exactly what rests on whose word, and the finding itself is AswanAI's own; limitations (unavailable parts, discarded evidence, unverifiable history) are stated in the document; and rejected alternative explanations are addressed, not ignored. Findings are never a bare probability score: every conclusion is a logical, explainable account of the evidence, the alternatives considered, and why each was eliminated, so the person holding the report can understand the failure and defend the finding to their own customer. Reports are issued the same day, pending expert review, and written review by Patrick Swan is available per case. Commercial and legal decisions always remain the customer's own.
Is there an audit trail?
Complete and permanent. Every case records who opened it, who contributed which evidence, who requested, staged, approved or cancelled each report, and on what grounds the report's content changed. The system is deliberately designed so that no user, including administrators, holds editorial control over a report's findings: challenges are weighed on their evidence and the record keeps the reasoning.
Can reports carry our branding and format?
Yes. Report branding and format are tailored to your standards: supply your preferred report format and branding and AswanAI applies it to your organisation's reports. The forensic content and its evidential discipline stay the same; the document looks the way your business needs it to.
What happens if we stop using the service?
Your data remains yours. On request, AswanAI provides your case data and reports and deletes them from the instance. There is no exclusivity and no lock-in in the licence.
Who is behind AswanAI, and what supports the analysis?
AswanAI is a South African company built on the 50-year forensic failure-analysis practice of Patrick Swan: five decades of investigations across trucking, transport, mining and power generation, thirty years of it captured as the documented case archive on which the system's judgment is built (see the expert behind it). The platform is developed and operated by AswanAI with a disciplined engineering process: every release passes an automated test suite of over 3,800 checks, changes are version-pinned per customer instance, and fixes during pilots are typically delivered same-day.
What does reliability and support look like?
Automated health monitoring across every instance (a daily health report, and an immediate alert to the operator if an AI service the platform relies on stops accepting requests), automated backups, and graceful recovery designed into the product (an interrupted connection never loses a case; investigations resume where they stood). Pilot customers get business-hours support with best-effort same-day response and a weekly review call; production service levels are agreed in the licence.
See it on one of your own failures
Bring a recent failure from your workshop and we will work it through with you, live.