Frequently Asked Questions
Pricing, IP, security, model choice, and how we run engagements. Written for the questions business leaders actually ask, not the ones that make good marketing copy.
How we structure our work, what we need from you, and how long things take.
We map your actual workflow before writing a line of code. Then we pick the single highest-leverage process, scope it tightly, and deploy. Once that system is running and producing measurable output, we expand, each new workflow builds on the integration work already in place. We stay involved: every system we build is something we continue to operate, tune, and improve.
Dev shops scope, build, and hand off. We stay. Our engineering team comes out of research, AWS AI Research Award recipients, NSF-funded neural networks work, and we treat your operation like a long-term research problem, not a project to close. We don't do templates. Every engagement is custom-engineered for your workflow.
Most clients work with us for twelve-plus months because the compounding value keeps expanding. Initial deployments typically go live within four to eight weeks depending on integration complexity. After that, the engagement is continuous: new workflows, model upgrades, guardrail tuning, and expansion into adjacent processes.
Less than you'd think. We need one executive sponsor for decision authority, and one subject-matter expert for the workflow we're automating, usually a few hours per week during scoping and deployment, tapering off once the system stabilizes. We handle everything engineering, integration, and ongoing operations.
Every engagement starts with success criteria you define: hours saved per week, cycle time reduction, error rate, throughput, or revenue impact. We instrument the system to track those metrics from day one and report them in a live dashboard. If the numbers aren't moving, we diagnose and fix before we expand scope.
Yes, and we recommend it. Every engagement starts with a single high-leverage workflow: one process, scoped tightly, with clear success criteria. Once that's running, expanding is fast because the integration groundwork is already in place.
Every engagement is a custom partnership scoped individually based on workflow complexity, integration depth, and operational scope. We typically start at engagements producing measurable ROI within ninety days. We don't take on work we can't defend.
How we structure pricing, what's included, and what you can expect on billing.
Every engagement is a custom partnership scoped individually based on workflow complexity, integration depth, and ongoing operational scope. Typical engagements run 60-80% less than a full-time hire for equivalent output, with no benefits, no ramp time, no attrition.
System operation, model hosting, ongoing guardrail tuning, model upgrades as new versions ship, integration maintenance, monitoring and alerting, a live metrics dashboard, and direct access to the engineering team. No per-seat licensing, no usage overages on standard workflows.
Yes. Our operational retainer bundles everything it takes to run your system, cloud infrastructure, model provider tokens, hosting, monitoring, and ongoing management. You get one predictable number instead of juggling AWS bills, Anthropic invoices, and surprise usage overages. If your workload grows beyond the envelope we scoped, we talk about it upfront rather than bury it in a line item.
Yes, discovery and initial build are typically billed as a one-time scoping and deployment fee, then the system transitions to a monthly operational retainer. All fees are disclosed up front. No hidden line items.
Scoping fees are non-refundable (you keep the research and artifacts), but operational retainers are month-to-month after the initial commitment, you can end the engagement any time with thirty days' notice.
Who owns what we build, what happens to your data, and what leaves with you if the engagement ends.
Yes. One hundred percent. Every line of code we write for you is yours, source code, infrastructure configuration, deployment scripts, and internal tooling. We don't retain any rights. If you ever part ways with PSV, you keep the repository, credentials, and everything else.
You do. One hundred percent. This includes prompt templates, fine-tuning data, evaluation suites, and the orchestration logic that connects it all. We use foundation models from providers like Anthropic and OpenAI, but the systems we build on top of them are yours.
Your data is yours. We don't use it to train shared models, we don't share it with other clients, and we don't retain rights to it. If we build evaluation datasets or fine-tuning corpora using your data, those artifacts are also yours and ship with the rest of the engagement.
Yes. Many clients start with us operating the system and later take over hosting, or host from day one on their own infrastructure. We deploy to your cloud account (AWS, GCP, Azure) when you prefer. The architecture is infrastructure-as-code, so the handoff is clean.
You keep everything. Code, data, models, workflows, infrastructure configuration, and documentation. We run a formal handoff that includes a runbook for ongoing operations, so your team or another vendor can take over without gaps. No lock-in, no hostage data.
How we protect your data, what certifications we hold, and how we handle regulated industries.
AES-256 at rest, TLS 1.3 in transit. Access is scoped to the specific engineers working on your engagement and logged. Production credentials are stored in a secrets manager, never in code. We follow least-privilege principles and run regular access reviews. Your data does not leave the infrastructure we build for you.
We are pursuing SOC 2 Type II certification. In the interim, we can provide our current security controls documentation, a completed SIG Lite questionnaire, and references from clients in regulated industries. We happily sign customer security addenda and participate in vendor security reviews.
If your system processes data about EU or California residents, we build in the required controls: data subject access request handling, deletion workflows, consent tracking, and data processing records. We sign DPAs and act as a processor under both frameworks. Specific implementation depends on your existing compliance posture, we integrate with it rather than duplicate it.
We minimize PII exposure to foundation models by default: redaction layers for sensitive fields, hashing where possible, and prompt design that avoids sending PII when the task doesn't require it. Where PII is unavoidable, we use model providers with enterprise-grade data handling agreements (no training on inputs, zero-retention modes where available).
Only what the system you hired us to build requires. We don't aggregate data across clients, we don't sell telemetry, and we don't retain operational data past the engagement. Logs and metrics are retained per a schedule you approve during scoping.
Yes. We work with regulated data in financial services, private equity, and commercial real estate contexts, including SEC-reported entities, PII, and financial transaction data. For federal government work requiring FedRAMP or IL4+ environments, we evaluate case by case and will tell you honestly if we're not the right fit. We don't pretend to certifications we don't hold.
Yes, when the compliance posture calls for it. We deploy to your own cloud account (AWS, GCP, Azure) or on-premise infrastructure for clients with data residency, air-gapped, or regulator-mandated isolation requirements. Open-source or self-hostable models (Llama, Mistral) cover workloads that can't touch a third-party API. We scope the architecture to your compliance envelope before we build.
The platforms we build on, how we handle model issues, and how we integrate with your stack.
We are model-agnostic. Anthropic Claude and OpenAI GPT for general-purpose reasoning, specialized models for vision, embeddings, and domain tasks, and open-source models (Llama, Mistral) where data residency or cost requirements demand it. We pick the right model for the workload, and swap as better options ship.
Whatever you run on. CRM (Salesforce, HubSpot, custom), ERP (NetSuite, SAP, Oracle), email, Slack, Teams, accounting platforms, data warehouses, legacy systems with SOAP or flat-file interfaces. If it has an API, we connect to it. If it doesn't, we build the adapter.
Hallucinations are a scoping problem more than a model problem. We design systems with hard guardrails: confidence thresholds, retrieval-grounded responses with citations, structured output validation, approval gates for high-stakes actions, and escalation paths when the system is outside its safe zone. Every action is logged and traceable. You always have full visibility and final authority.
We monitor model behavior against a versioned evaluation suite tied to your success criteria. When a provider ships a new model, we run it through the same suite, compare results, and either upgrade or hold. No silent upgrades that break your workflow. All changes are logged and reversible.
No. Never, unless you explicitly agree and we've written it into the engagement. We use foundation models with data-handling agreements that exclude training on inputs. Any fine-tuning we do on your data produces a model that's yours alone.
Providers publish deprecation schedules months in advance. We track them as part of ongoing operations, test successors against your evaluation suite, and migrate before the old model turns off. Migration work is included in the operational retainer, no surprise bills.
Who builds your systems, how selective we are, and what industries we know deeply.
Our engineering team. AWS AI Research Award recipients, NSF-funded neural networks researchers, and senior engineers with production experience at scale. No offshoring, no junior-heavy teams. The people you meet on the strategy call are the people who build, and they stay on the engagement.
Very. We take on a limited number of engagements each quarter because every client gets the senior team's attention. We turn down work we can't defend or don't believe we can ship meaningful value on. If we're not the right fit, we'll tell you on the strategy call.
Commercial real estate, private equity, and financial services are our three focus verticals. We know the workflows, the data, the regulatory constraints, and the operators. Each vertical has its own page with the specific playbook we run.
Fintech engagements use compliance-aware controls for KYC, AML, and financial data handling. Every action is logged, high-stakes decisions route to a human approver, and systems are designed around the specific regulatory regime you operate under. See the Financial Services industry page for the full architecture.
How to tell if you're ready, what the first step is, and what to expect.
Start with our AI Operator Maturity Assessment at /tools/ai-readiness-quiz. It takes five minutes and places you on our 7-stage framework, from Exploring through Native, so you know exactly which stage to cross next. If you'd rather talk it through, book a strategy call directly.
Book a thirty-minute strategy call. We'll learn your business, identify candidate workflows, and tell you honestly whether AI will move the numbers you care about. No pitch deck, no upsell. If there's a fit, we scope. If there isn't, we say so.
Four to eight weeks from signed scope to first production deployment, with the first measurable output typically within the first two weeks of build. Simpler workflows can ship faster; integration-heavy or regulated deployments take longer.
For custom partnerships, yes. Discovery is a fixed-fee engagement (one to three weeks) that produces a detailed scope, architecture, integration plan, and success criteria. If we build the system after discovery, the discovery fee is credited against the build. If we don't build it, you keep everything we produced.
Don't see your question? Tell us what you're evaluating and we'll give you a direct answer, usually same day.
Thirty minutes. We'll tell you whether PSV is the right partner for what you're trying to build.
Or reach us directly at aidan@pacificsoftwareventures.com