The process
From first call to production, phase by phase
Every engagement follows the same five phases. The durations vary by project; the structure doesn't.
Discovery Call
30 minutes - freeWe talk through your problem, your tech environment, your data, and your timeline. No sales pitch. An AI engineer (not a salesperson) is on the call. By the end, you know whether we can help and what approach makes sense.
Technical Scoping
3-5 business daysWe write a Technical Scoping Document: architecture decision record, data requirements, integration points, risk register, and a week-by-week delivery plan. You see the blueprint before any code is written. We sign an NDA before we start this phase.
Build Phase
Typically 4-12 weeksWe build in two-week sprints with working deliverables at each milestone. You get Slack/WhatsApp access to your engineer throughout. No 'black box' development - you see progress in real time. Every sprint ends with a demo and a written milestone report.
Deployment & Handover
1-2 weeksWe deploy to your infrastructure (or ours), run final QA, and document everything. The handover package includes: full source code, architecture docs, runbook, evals baseline, and a Loom walkthrough video. Your team can own and maintain the system from day one.
Ongoing Support
Optional - SLA-basedWe offer retainer-based ongoing support: model monitoring, performance optimization, feature development, and incident response. This is structured as a monthly SLA, not ticket-based. Most enterprise clients retain us for 3-6 months post-launch.
The first deliverable
What an Architecture Brief contains
The same outline the contact form promises. First version is free and arrives within one business day; the deep version follows discovery sessions.
- 01
Current workflow
The queue as it is worked today: inputs, exceptions, hand-offs, and who signs off. Written from what operators told us, not from the org chart.
- 02
Systems of record and integration points
Which system owns the data (WCS, MES, claims platform, core banking, CMS), what can be read, what may eventually be written, and by whom.
- 03
Trust phase to start in
Shadow or Assist, with the agreement metric that decides when to move. No engagement starts at Scale.
- 04
Stack
Named components per layer: orchestration, tools, models, retrieval, approval inbox, audit log. Chosen against what you already run.
- 05
Risks
The three things most likely to stall the build: data access, an unowned approval queue, a compliance review on the critical path. Each with a mitigation.
- 06
12-week plan
Week-by-week through shadow mode to the first assisted action, with the go/no-go metric and who owns each week.
- 07
Cost band
A range for the pilot and a monthly figure for the retainer that follows, with the assumptions that would move them.
After signature
Week one, day by day
What a fixed-scope pilot looks like in its first five working days. Dates shift; the sequence does not.
Day 1
Access and owners
Repo, cloud account, and read-only system access requested. Client-side owner for the approval queue named in writing. NDA and data-processing terms confirmed.
Day 2
Operator sessions
Two or three people who work the queue walk us through ten real items, including the ones that went wrong. Exceptions are logged.
Day 3
Integration spike
First read from the system of record. Data shape, latency, and permissions confirmed against the brief. Surprises go on the risk list.
Day 4
Eval set drafted
Fifty to two hundred real cases with the operator's decision attached. This is what shadow mode will be scored against.
Day 5
Plan re-baselined
The 12-week plan is updated with what week one found. You receive it with the risk list and the eval set. Weekly demos start the following week.
Engagement models
Four ways to be hired
Each has a defined scope, a named owner on your side, and a stop point you control.
Fixed-Scope Project
Best for: Well-defined problem, specific output
You define the outcome. We scope the build. One price, one delivery date. Best for MVPs, specific AI features, RAG systems, or automation pipelines with clear inputs and outputs.
Retainer / Fractional Team
Best for: Ongoing development or CTO-as-a-Service
Embed a Xenqube engineer or technical lead with your team on a monthly basis. Full access to AI/ML, full-stack, security, and Web3 expertise. Ideal for startups scaling their engineering, or enterprises running multiple AI workstreams.
51-Hour MVP Sprint
Best for: Validate fast, decide before committing
Idea to deployed prototype in 51 hours. Strict scope, fixed price, full ownership. Many sprint clients continue into a full project - with no obligation. Perfect for testing a hypothesis before a larger investment.
Enterprise Advisory
Best for: AI strategy, due diligence, architecture review
A Xenqube technical lead for your board presentation, AI strategy session, vendor due diligence, or architecture review. Independent technical opinion - no vendor bias, no upsell pressure.