Human + AI, in that order
AI processes volume; the human team interprets. Without interpretation, an automated report is just prioritized noise.
Service · GeneXus audit
A KB audit with a human team and AI-powered analysis.
I run KBDeepdive.AI on your KB, but I don't hand you a PDF and disappear. I interpret the findings with your business context, prioritize by real impact and leave you with a plan your team can execute on Monday.
The context
A 10-year-old KB isn't read: it's archaeologized.
Procedures nobody remembers why exist. Conventions that changed three times. Dependencies tangled into a knot. Automated tooling gives you data, but someone has to read it and decide what matters. My audit covers both halves: deep AI-powered analysis and human judgment about what to do with the findings.
It's not an automated report. It's an informed technical opinion.
Business context first
Before I look at code, I talk to your team. Which modules are critical, what's about to be migrated, what hurts today.
Prioritized by impact
I don't dump 800 findings on you. I give you the 20 that move the needle — and I explain why that order.
Actionable plan
Each recommendation comes with effort estimates, affected objects and how to verify the fix landed.
Real confidentiality
NDA up front, analysis on agreed infrastructure, code that never leaves without your sign-off. I do this with banks.
Comparative over time
On follow-up audits I measure evolution — where you improved, where new debt was added.
The concrete deliverables of the audit.
Navigable HTML report
Drill-down by module, object type and severity. Shareable with your tech team without going through me.
Prioritized refactor ranking
Findings sorted by impact/effort. The first item is always the one with the highest return.
Phased action plan
What to do in phases — with clear milestones and success criteria.
Walkthrough session
Live session with your tech team: I walk through each finding, answer questions, and we adjust priorities.
Executive summary
A one-pager with KB health, risks and estimated costs. Built for the steering committee, not the DBA.
Metrics history
Snapshot of key metrics so you can compare against the next audit.
From the NDA signature to the plan in your hand.
- 01
Kickoff and NDA
Scoping call with your tech team and a business stakeholder. We sign the NDA before reading a single line of code.
- 02
Secure KB ingestion
I receive the KB on agreed infrastructure — yours, mine, or an isolated environment. No loose copies.
- 03
Deep AI-powered analysis
I run KBDeepdive.AI on the KB. Automatic detection of duplicates, debt, vulnerabilities, circular dependencies, broken conventions.
- 04
Human interpretation
The human team reviews the findings, cross-references them with your business context and discards noise. What's left goes into the report.
- 05
Walkthrough session
Session with your team. I walk through each finding, we prioritize together and adjust the plan with their input.
- 06
Formal handoff
Report, plan, executive summary and metrics snapshot. I stay available for implementation questions afterward.
This audit is built for a specific case.
I'd rather say no than oversell. If something here doesn't add up for you, we'll talk it through on the first call.
It makes sense if…
- Your KB has 100k+ lines and five or more years on it.
- You're about to migrate to a new GeneXus version and want to know what you'll be dragging along.
- Team velocity dropped and you want to understand how much is tech debt and how much is something else.
- You're onboarding new devs and need evidence-based onboarding, not folklore.
- An external audit is part of due diligence (M&A, vendor change, certification).
It's not for you if…
- You have a small KB you know by heart — KBDeepdive.AI self-service is enough.
- You're after the analysis tool, not the service. It exists — KBDeepdive.AI, used directly.
- You need a formal sign-off without intent to act. I don't sign rubber-stamp audits.
Why this service exists
KBDeepdive.AI was born as a self-service product: you run the tool, you get the report, you act. It works well when you have a senior tech team with time to read and interpret.
But many clients asked me the same thing: “give me the report already processed — I don’t have time to do archaeology on an 800-line HTML”. This service is exactly that: the same tool, but with a human team that hands you a plan ready to execute.
What “deep analysis” means
It’s not just metrics. The analysis combines:
- Structural metrics: complexity, coverage, dependencies, duplication.
- AI-powered semantic detection: problematic code patterns that traditional rules don’t catch — code that “compiles but shouldn’t exist that way”.
- Cross-reference with your business: modules handling 80% of transactions don’t weigh the same as one used once a month. I know that before I prioritize.
What you won’t find here
It’s not a quality stamp. It’s not a generic PDF to print for a certification. It’s not a pretext to sell you the refactor.
It’s an informed technical opinion about your KB — with data behind it, human judgment on top, and a plan you can execute with or without me.
What people usually ask me before signing.
How is this audit different from the KBDeepdive.AI product?
KBDeepdive.AI is the tool — self-service, you run the analysis on your KB yourself. The audit is the service: I run the tool, interpret the findings with your context, prioritize by business impact and hand you an action plan. Two distinct products with the same engine.
How long does it take?
Depends on KB size and agreed scope. The automated analysis is the fast part; what takes time is the human interpretation and the team session. We close the calendar at kickoff.
Which GeneXus versions do you cover?
I work with KBs from GeneXus X Evolution 3 onwards. For older versions, let's talk — usually viable, sometimes requires extra prep.
How do you handle code confidentiality?
NDA before access to anything. You can opt for analysis on your infrastructure, in an isolated environment of mine, or on yours with the team connecting in. There are never copies of the code outside what's agreed.
What KB sizes can you audit?
From 50k to several million lines. The engine is tuned for large KBs; very small KBs make less sense because the full service's return doesn't justify the cost.
How much does it cost?
Price depends on KB size and scope (specific modules or the whole KB). After the scoping kickoff I close a fixed price — no hidden variables. The first call is free.
Do you only audit, or do you do the refactor too?
The audit is delivered ready for your team to execute. If you'd rather I do the refactor, that's a separate contract I quote at the end of the audit — but it's never required, and I never tie recommendations to selling you extra work.
Other services
Have a KB asking for help?
A first call is enough to size the audit and give you an estimated quote. No commitment, no NDA up front.