Zero project failures since 2007 — the engineering rules behind it
I founded Software Tailor in 2006. Nineteen years on, the company has shipped through four distinct economic cycles, six Fortune Global 500 customers across pharma, finance, government, legal, defence, and energy, and zero project failures since 2007. That last number is the one investors and procurement reviewers ask about most often, and it's the only one that needs explaining. Here are the four engineering rules that hold the record up.
Rule 1: One delivery model, no service-vs-product blur
Every Software Tailor engagement runs on the same delivery model — a tightly-scoped pilot phase that ships working software, followed by an expansion the customer chooses to fund (or not) on the basis of what shipped. We don't run a separate "services" practice that ships against requirements docs and a separate "product" practice that ships against roadmaps. The two would conflict; one would always be cross-subsidising the other.
The single model means every engineer knows what "done" looks like before they start. That's the precondition for not failing.
Rule 2: Customer cohort discipline — only customers we can ship for
Our customer cohort is selected — not opportunistically. Six Fortune Global 500 customers in 19 years isn't a slow sales pipeline; it's a deliberate selection bar. Every customer we take on has to pass three tests before kickoff: their problem is one we've solved before, their internal sponsor is the person who will use the software (not a layer of intermediaries), and their hardware/data environment lets us ship without a six-month procurement detour.
A customer that fails any of those tests becomes a referral to someone else, not a project. This is the rule that costs us revenue most often. It's also the one that has zero exceptions across 19 years.
Rule 3: Engineering discipline mapped to a recognised framework
Long before AI risk management had its own framework, the engineering rules I codified for Software Tailor pattern-matched the same shape: govern the work, map the risks, measure the outcomes, manage the changes. NIST AI RMF 1.0 [1] formalised that pattern for AI in 2023, and the April 2026 Critical Infrastructure profile [1] extended it to regulated industries. Reading those documents, our internal discipline maps cleanly onto the four functions.
What that gives us: every project is auditable end-to-end against an external framework, not just against our own habits. When a procurement team asks how the pilot-phase scope gets decided, the answer is the same answer NIST would give for risk management — and the customer's own compliance team has already internalised that vocabulary.
Rule 4: Recordable, recoverable, recreatable
The fourth rule predates the audit-trail vocabulary we now use for AI Suite. Every Software Tailor engagement runs in a state where any decision, any artefact, any version of the code can be recreated from version-controlled inputs. That's not unusual for software engineering generally; what's unusual is that we apply it to the project, not just the codebase. Sprint decisions, scope changes, customer sign-offs — all recorded, all recoverable.
The discipline is why our content-free audit logs approach for AI Suite shipped the way it did. The audit-row pattern is the same pattern we already applied to projects. The vocabulary is borrowed from compliance frameworks; the practice is borrowed from how we ship.
What carried through to AI Suite
The Local AI Suite + AI Admin Console line — see Why we ship AI as installable binaries, not cloud SaaS — is the same engineering discipline applied to a product line instead of a custom engagement. Same single delivery model (desktop installers with a free 1-week pilot), same customer cohort discipline (regulated industries with clear data-residency constraints), same framework mapping (NIST AI RMF + EU AI Act deployer obligations), same recordable-recoverable-recreatable audit posture.
Zero project failures since 2007 isn't a slogan. It's the result of a small number of rules applied without exception. The same rules now govern how AI Suite is built.
References
- NIST. "AI Risk Management Framework (AI RMF 1.0)." nist.gov/itl/ai-risk-management-framework. Accessed 2026-07-15.
Related articles
Test the same discipline on a real engagement.
A free 1-week pilot is how we onboard every Fortune Global 500 customer. Same delivery model, same recordable / recoverable / recreatable posture, same engineering rules.