Objections & How We Address Them
The consolidated register of objections raised against the platform, each with its current answer and status — tracked until fully addressed.
We track objections here until we fully answer them. We mark an answer done only when the design or evidence behind it is in place. This page is the single register. Where we have a solid answer, we state it plainly. Where we do not, we name the risk and leave its status open.
Status at a glance
The table below lists the most important objections and where each one stands. "Addressed" means the design or evidence answers it today. "In progress" means the answer depends on work still under way. "Open" means we accept the risk and watch it rather than close it.
| Objection | Current answer | Status |
|---|---|---|
| Client data must not go to an external model | Rule-based sanitization before any model, local-model routing for sensitive work, bring-your-own-model, explicit consent | Addressed |
| Regulated clients cannot use external AI at all | The core scan uses no AI and needs no AI approval; AI-assisted layers are optional and consent-gated | Addressed |
| Every line of AI output needs human review | Review is the product, not a cost to remove: dry-run preview plus senior-engineer approval on every change | Addressed |
| AI-generated code adds technical debt | Dry-run and engineer approval on every change; rule-based automation preferred where it fits | Addressed |
| Where is the durable IP versus "using AI better" | The curated, engagement-fed knowledge base plus measured benchmark data; tools are commodities, the corpus is not | Addressed |
| Nobody wants another portal | Vizzy surfaces inside existing workflows (Slack today, more surfaces planned); the platform is not an extra destination | Addressed |
| Compliance and consent block deployment | A concrete compliance package — sanitize-before-model, BYOM, explicit consent — proven with a design partner, then reused | In progress |
| The platform should be built with the tooling it sells | AI development on the platform codebase is paused pending a security assessment; the assessment is a tracked item, and the playbooks apply the tooling everywhere else today | In progress |
| Parts of the organization still question the value | The executive brief, measured margin instrumentation, and demo-first delivery prove value with evidence, not argument | In progress |
| Real ROI, pricing, and conversion are unproven | The first paid engagements exist to produce measured savings, real cost of delivery, and conversion data; until then ranges stay ranges | In progress |
| Platform vendors ship native AI that may absorb this | A vendor-neutral cross-platform position, the knowledge-base moat, and a dollars-quantified assessment vendors do not offer | Open |
| Experienced experts can get slower with AI | Keep AI on bounded, reviewable work and measure end-to-end delivery time before quoting savings | Open |
| Models may be near their capability ceiling | Accepted as a planning assumption; nothing in the plan requires models to improve | Open |
Security & data
Compliance & adoption
Economics & vendors
Reliability & quality
Market & defensibility
Internal
Nothing here asks anyone to lower a security or compliance standard. We built the design around these objections. Most of them are the reason the architecture looks the way it does. If an objection is missing from this register, that is an oversight. Raise it, and we add it.