Engagement process

From first call to production — what to expect.

A DehazeLabs engagement has five stages: discovery, scoping, delivery, operations, and handoff. Here's what each one looks like in practice — no surprises, no black boxes.

→ Stage 1: Discovery call

30 minutes. Direct.

The first call is a mutual qualification. We want to understand: what domain you're in, what you're trying to build, what's already in place, and what the timeline looks like. You want to understand: whether we've done this before, whether we're the right team, and whether the engagement economics could work.

We'll tell you at the end of the call if we think we're a fit. If we aren't — if the work is outside our core, or if another type of firm is better suited — we'll say so and recommend alternatives. Discovery calls are free. Nothing is confidential unless you want it to be; we sign NDAs before any substantive technical conversation.

→ Stage 2: Scoping

1–2 weeks. A concrete technical plan.

If the discovery call goes well, we move to scoping. Over 1–2 weeks, we produce a scoping document that covers: what we'd build, in what phases, with what team composition, at what cost range, and what the key technical risks are. This isn't a proposal deck — it's a working document with architecture diagrams, team sizing, and milestone definitions.

For qualified production builds, scoping is free. If we can't reach agreement on scope and economics, there's no engagement — and you keep the document. For teams that need technical diagnosis before committing to a full build, we offer a paid AI Infrastructure Assessment, typically $35K-$75K, credited toward the larger engagement if it proceeds.

→ Stage 3: Delivery

Phased. Milestone-driven.

Engagements run in 2–3 phases, typically structured around the natural architecture layers of the work:

Phase 1 — Data and infrastructure foundation. Ingestion, normalization, storage architecture, schema design. This phase often surfaces surprises about data quality and availability that affect phase 2 scope. Starting here before building the AI layer is how you avoid spending $500K on a model that has no reliable data to run on.

Phase 2 — AI and reasoning layer. The models, pipelines, and reasoning systems that sit on top of the data foundation. Whether that's a real-time SIEM, an agentic ops layer, a RAG platform, or a CV inference pipeline depends on the engagement. Phase 2 scope is confirmed after phase 1 based on what we learned about your data.

Phase 3 — Production hardening and pre-handoff. Load testing, SLA validation, monitoring and alerting, documentation, and the joint operations period that precedes formal handoff. This phase exists because production readiness is different from feature completeness.

Each phase has explicit milestones with defined deliverables. Scope changes within a phase go through a lightweight change order process. Surprises that affect timeline or cost are surfaced early — not at invoice time.

→ Stage 4: Operations

We run it while you learn it.

Most engagements include an ongoing operations period after the initial build — typically 3–12 months. We run the platform under SLA, monitor for drift and degradation, and handle incidents. This period exists because production AI systems require operational knowledge that can't be fully transferred through documentation alone. It takes time to see the failure modes.

During the operations period, your team shadows ours — joining incident response, reviewing the monitoring setup, running the platform jointly. By the time we hand off, the platform's operational history is understood, not just documented.

→ Stage 5: Handoff

Structured, not sudden.

Handoff is a process, not a date. It starts 4–8 weeks before the formal engagement close — your team takes primary on-call, we provide backup. We deliver runbooks, architecture decision records, a system health baseline, and a documented incident history. The final handoff meeting is a review of the platform's known edge cases, not an introduction to the architecture.

Some clients keep us on a retainer after handoff for specific components that require specialized expertise they don't want to hire for permanently. Others bring us back for the next build. Either is fine — we design engagements to end cleanly, not to create dependency.

[ FAQ ]

Engagement process — common questions.

How long does the scoping process take?
Scoping typically takes 1–2 weeks from the first call. For qualified production builds, scoping is free and produces a document covering what we'd build, in what phases, with what team, at what cost range. For teams that need technical diagnosis before committing to a build, we offer a paid AI Infrastructure Assessment, typically $35K-$75K, credited toward the larger engagement if it proceeds.
What happens if the scope changes mid-engagement?
Scope changes are normal. We structure engagements in phases so phase 2 can be scoped based on what we learn in phase 1. Changes within a phase go through a lightweight change order process. Major changes that affect timeline or cost are flagged early — we'd rather have an uncomfortable conversation in week 4 than a surprise invoice in week 16.
Do you sign NDAs before discovery calls?
Yes. We sign mutual NDAs before any substantive technical conversation. Send yours and we'll review and sign, or we can share our standard mutual NDA. We don't require sensitive disclosure in the initial email — just enough context to determine whether a first call makes sense.
What does knowledge transfer and handoff look like?
Handoff is a structured process, not a documentation dump. It starts during the final phase — we run the platform jointly with your team for 4–8 weeks, pairing your engineers with ours on operations and incident response. We produce runbooks, architecture decision records, and a system health baseline. By formal close, your team has been operating the platform alongside us for weeks.

Ready to start the first conversation?

Send a one-paragraph brief — what you're building, rough scope, timeline. We'll come back within two business days with a yes, no, or "let's talk."