Skip to main content

Consulting · Delivery

What actually happens if you hire us

Six stages, each with a decision that has to be signed off before the next one starts. This page is as much about what your team is on the hook for as what we deliver — engagements fail on adoption and unclear ownership far more often than on build quality.

Six engagement stages, we-do and you-do, gated between eachSwimlane diagram of six engagement stages — Assess, Baseline, Design, Build, Test, Hand off — each split into what Valentino AI does and what the client does, with a sign-off gate before every stage after the first.WE DOYOU DOSTAGE 1AssessInterviews, workflow map, auditMake owners availableSTAGE 2BaselineMeasure current-state numbersConfirm it’s a fair baselineSTAGE 3DesignSpecify workflow, action boundaryReview the boundarySTAGE 4BuildBuild against agreed scopeProvide least-privilege accessSTAGE 5TestRun real cases, exceptionsSupply the awkward casesSTAGE 6Hand offTrain, document, transferName the internal ownerGATEA signed-off decision before every stage after the first — not a status update.
Stage 1

Assess

We do

Interview the people who do the work, map the workflow end to end, audit data readiness, and identify the first project worth building. Delivered as the AI Operations Assessment.

You do

Make operations, pricing, finance, and systems owners available for about ninety minutes each. Name a business owner for the outcome — not just a technology sponsor.

Before the next stage

Agreement on which workflow goes first, and on the metric it has to move.

Stage 2

Baseline

We do

Measure the current-state numbers the same way we will measure them after launch, and write down the assumptions behind them.

You do

Confirm the baseline is a fair representation of how the operation runs today, including the seasonal or lane-level caveats we would not know to ask about.

Before the next stage

A signed-off baseline. Without one, no one can honestly say later whether the build worked.

Stage 3

Design

We do

Specify the workflow, the integration points, the data model, and the agent action boundary — what may be drafted, what may be recommended, what requires approval, what may execute, and what must escalate to a person.

You do

Review the action boundary with the team who will live with it. This is the stage where "the agent should never do X" has to be said out loud.

Before the next stage

Written scope with acceptance criteria, named source systems, and access method agreed.

Stage 4

Build

We do

Build against the agreed scope, in your accounts, with specialist partners named up front for data engineering, BI, telephony, or security work.

You do

Provide least-privilege access to the named systems, and a point of contact who can unblock questions about how the operation actually behaves.

Before the next stage

Working system against the acceptance criteria, demonstrated on your data.

Stage 5

Test

We do

Run the workflow against real cases including the exceptions, and check the agent boundary holds where the inputs are incomplete or contradictory.

You do

Supply the awkward cases. The ones your team works around are worth more than the happy path.

Before the next stage

Acceptance criteria met, exception handling agreed, and the measurement re-run against the baseline.

Stage 6

Train and hand off

We do

Train the people who will run it, document the system and its failure modes, and transfer ownership of the accounts, credentials, and code.

You do

Name the internal admin or operator who owns it after we leave, and put them in the training.

Before the next stage

Your team runs the workflow without us in the room.

Where our involvement ends

The most common fear about this kind of work is a consulting dependency that never ends. Here is the boundary, stated up front.

  • The system is yours at handoff — accounts, credentials, documentation, and code. There is no dependency clause that keeps you on a retainer to use what you paid for.
  • We stay on call for major upgrades and for questions the documentation does not answer. We do not bill for our absence.
  • Ongoing managed operations, 24/7 monitoring, and incident response are not included by default. Where you want them, they are scoped and priced separately, or delivered by a named partner.
  • Licences and platform costs — warehouse, BI seats, telephony, model usage — are yours directly and are never bundled into our fee.

When we say no, or not yet

A recommendation to wait is a legitimate outcome of the assessment, and we have given it. These are the cases where we would.

  • The data cannot support the workflow yet, and the honest sequence is to fix the foundation first.
  • There is no business owner for the outcome — only a technology sponsor. These engagements fail on adoption, not on build quality.
  • The requested scope needs a capability we do not hold and cannot cover with a named partner.
  • Nobody can name the metric the project is supposed to move.

It starts with stage one

The assessment is the only commitment you make up front, and the roadmap is yours whether or not the rest of this ever happens.

Book an AI Operations AssessmentWhat the assessment covers