Recover understanding
Map the relevant code, dependencies, data flows and business rules, including knowledge that has disappeared from the team.
AI-assisted legacy software rescue and modernisation
Turn the software everyone is afraid to touch into a system the business can safely change.
Important changes get stuck when old software is poorly documented, fragile or understood by too few people. I use AI-assisted engineering, experienced human review and practical testing to recover how the system works, expose the real risks and build a controlled route to the next valuable change.
Seismic Technology is my consultancy. I’m Mick Garlick, and I remain directly responsible for the investigation, engineering judgement and recommendations. Who I am and how I work.
Recognise the problem
What valuable change keeps being postponed?
It might be a checkout improvement, an integration, a recurring failure, an overdue platform upgrade or a manual process that wastes time every week. The business case may be clear, but the existing application feels too risky to change.
The first step is not a rewrite. It is to recover enough reliable understanding to decide what can be changed safely, what must be protected and what a sensible implementation will cost.
A controlled route forward
Make the next decision with evidence
From blocked change to clear decision
The question on the table
Can we make this change without putting the operation at risk?
The method
AI-assisted, human-reviewed investigation
Reduce change risk
Identify the behaviours that must be preserved, introduce proportionate tests and keep AI-assisted changes small enough for human review.
Create a route forward
Turn the investigation into a prioritised recommendation and a separate, evidence-based quote for the next change.
Decision-ready outcome
A route forward you can defend
Clear finding, priority risks and a scoped implementation quote
How I help
Focused help for software that has become difficult to trust
Blocked-change investigation
A bounded investigation into one commercially valuable change that has stalled because the application is poorly understood or difficult to trust.
Incremental legacy modernisation
Tests, documentation, safer boundaries and modern engineering practices introduced while the business continues to operate.
Inherited-system and integration recovery
Help for businesses, agencies and MSPs that have inherited an application or interface they did not build but must now support, change or connect.
A bounded first step
Start with one blocked change
The defined initial engagement is a five-working-day investigation that turns uncertainty about one valuable change into a clear finding and a practical next decision.
Relevant evidence
Commercial judgement backed by difficult legacy work
Specsavers, through BJSS
Over three and a half years, maintained and extended international Drupal and Symfony platforms, became a subject matter expert for two important subsystems, gave presentations on those areas and led high-risk work in them.
Virgin Media, through Hayes
Over two and a half years, contributed within an agile team to legacy PHP migration and extension, including an IP address-management system moved toward Symfony 4.
Retail integration
Over six years at Arcadia through Retail Business Solutions, led middleware design and live implementation, worked across supplier boundaries, and recruited, trained and built tools for the technical implementation team that delivered the first few hundred stores; during a separate fifteen-month Dunelm engagement through Paul Mason Consulting, recovered missing interface knowledge, wrote Retail-J XML middleware and supported testing.
Live mobile game
Recovered and modernised the live 130,000-line Dragon Lords PHP game back end, including missing-source reconstruction, tests and an AI-assisted PHP 5.6 to 8.2 conversion completed in 14 days.
Full case study, live mobile game
Recovering and modernising a 130,000-line PHP back end
Dragon Lords provides detailed, inspectable evidence of source recovery, an AI-assisted upgrade, human review, testing and continued modernisation.
Accountable engineering
AI-assisted does not mean unsupervised
AI helps investigate a large codebase, trace behaviour, draft documentation and accelerate controlled changes. I remain responsible for the engineering decisions, review and validation.
Work is divided into human-reviewable batches and supported by tests and other quality gates appropriate to the system.
AI tools are not given production access. Source-code use, data handling, confidentiality requirements and permitted tools are agreed for each engagement. Sensitive customer data is redacted or replaced with purpose-built fixtures where necessary.
Start the conversation
What is the change your team does not trust the old system enough to make?
Give me enough context to understand the commercial goal and the software risk. I will tell you whether a bounded investigation looks like the right next step.
I will read this personally and use it only to respond to your enquiry.